This past week, I had the privilege of giving a talk at Australia National University’s Coral Bell School of Asia-Pacific Affairs exploring the knock-on effects of the Iran war globally.

Subscribe now

Sharing a condensed summary of my speech notes and slide below:

How did a war centered on Iran become a crisis for economies and governments far beyond the Middle East?

The war has become a polycrisis because its effects are moving through systems that connect the Middle East to the wider global economy. Military escalation has disrupted energy exports, maritime trade, financial relationships, defense supply chains and regional security arrangements. These effects interact with pressures already created by the Ukraine war, the Gaza conflict, disruptions in the Red Sea, and growing economic competition between the United States and China.

My argument is that interdependence is doing two things simultaneously. It is transmitting disruption across borders, but it is also giving states control over chokepoints that can be converted into political leverage. Iran has used its geographical position around the Strait of Hormuz to impose costs that it could not generate through conventional military power alone. The United States exercises similar leverage through finance, sanctions and advanced technology. China occupies important positions in manufacturing, critical minerals and industrial supply chains.

The response from states exposed to these pressures has been to diversify. Gulf governments are building additional security relationships. Exporters are exploring routes around Hormuz. Governments are stockpiling strategic goods and localizing critical industries.

These adaptations may make individual states more resilient. Collectively, however, they are producing a more fragmented international order built from overlapping security, economic and technological networks.

Weaponized interdependence

What if the same networks connecting states also give some of them the power to coerce others?

The starting point is weaponized interdependence, developed most prominently by Henry Farrell and Abraham Newman.

Globalization did not distribute connectivity evenly. Financial flows, supply chains, digital systems and transport routes frequently converge around a limited number of hubs. Some states exercise jurisdiction over these hubs. Others depend upon them.

That asymmetry creates leverage. The United States can use the centrality of the dollar, correspondent banking, and American technology to monitor transactions or restrict access to the global economy. Iran possesses far less aggregate power, but its geography places it next to one of the world’s most consequential energy chokepoints.

A chokepoint can therefore be physical, technological or institutional. It may be a maritime passage, a payment system, a cloud platform, a semiconductor fabrication tool or a mineral-processing hub.

Control does not need to be absolute. A state only needs enough influence over a critical node to delay access, raise costs or make continued dependence politically risky.

This distinction between aggregate power and network position is central to the lecture. Iran cannot match the combined military and economic resources of the United States and Israel. It can still impose global costs because it occupies a position within the energy and shipping networks on which more powerful states depend.

Polycrisis and cascading crises

When does a regional conflict become a global polycrisis?

Weaponized interdependence explains how actors exploit networks. Polycrisis explains how disruption travels through them.

I use polycrisis here to mean an accelerated crisis in which a trigger moves across interdependent networks and value chains, affecting systems far beyond the original event. The important feature is interaction. An energy disruption changes shipping patterns. Shipping disruption affects insurance, commodity prices and industrial production. These pressures then influence domestic politics and governments’ willingness to sustain military commitments.

The Iran war illustrates this process. Operation Epic Fury began as a military campaign centred on Iran’s nuclear capabilities and regional power. Its consequences quickly moved beyond the battlefield into supply chains globally.

The disruption of Hormuz affected oil and liquefied natural gas, but it also affected aluminium, sulphur, helium and petrochemicals. The interruption of these inputs raised risks for critical-mineral processing, semiconductor manufacturing and other industrial sectors. Gulf missile and drone attacks increased demand for air-defensee interceptors already needed in Ukraine. Sanctions against Iran raised questions about Chinese banks, oil purchases and alternative payment systems.

States responded by seeking redundancy.

• pursued alternative trade routes

• expanded relationships with additional security partners.

These decisions reduce exposure to particular chokepoints, but they also create new dependencies and political commitments.

Polycrisis thus describes both the transmission of the initial shock and the a series of state adaptations in response.

An example causal chain

How does a missile strike in the Gulf eventually affect a semiconductor factory or weapons shipment elsewhere?

This slide shows how a military event becomes a systemic crisis.

The initial trigger for this case is the Iran war. Iran responds to the conventional superiority of the United States and Israel by attacking regional targets and restricting maritime passage through Hormuz.

The immediate effect is a reduction in energy exports and commercial shipping. Tankers wait outside the Gulf driving up insurance prices then shipping companies reroute or suspend operations.

Those disruptions then move into other sectors. Higher energy and freight costs affect industrial production. Interruptions to sulfur and sulfuric acid increase pressure on the processing of copper, nickel, lithium, and other minerals. Reduced helium exports create problems for semiconductor fabrication and medical systems. Competition for air-defense missiles affects the ability of the United States and its partners to supply multiple theatres.

Governments and companies then adapt. They draw down reserves, seek alternative suppliers and negotiate new transport corridors. Gulf states deepen relationships with other military partners. China protects its energy exposure while preserving economic access to Iran. The United States attempts to use financial pressure to reduce the economic space available to Tehran.

The causal chain does not end when the immediate disruption subsides. Investment decisions, new partnerships and industrial policies can remain in place long after the original military crisis.

Global systems already under pressure

What happens when a new war enters a global system already strained by other conflicts and economic competition?

The Iran war did not enter a stable international system. It entered one already operating under substantial pressure.

The Ukraine war had increased demand for munitions, disrupted food and energy markets and produced an extensive sanctions regime against Russia. It also demonstrated the central role of drones, electronic warfare, and layered air defense in contemporary conflict.

The Gaza and Red Sea crises had already raised the risk of regional escalation. Houthi attacks forced shipping companies to avoid the Red Sea, increasing transit times, freight costs and insurance premiums.

US–China competition was fragmenting technology and manufacturing networks through tariffs, export controls and industrial subsidies. Governments were already reconsidering the risks of concentrating semiconductor production, critical-mineral processing and advanced manufacturing in a small number of countries.

We are also witnessing the financial fragmentation of the global economy. Russia, Iran and other sanctioned states had expanded alternative payment channels and shadow trading networks. China has promoted greater use of the renminbi in some transactions, while the United States continued to use access to the dollar as a source of coercive power through Trump’s D-Day sanctions.

The Iran war connected these pressures. It added another major demand on Western defense production, disrupted a maritime route central to Asian energy security, and increased the strategic value of Chinese cooperation. This is why its consequences cannot be assessed as an isolated regional war.

Iran weaponizes geography

How can Iran impose global costs when it cannot match the conventional power of the United States and Israel?

Iran entered the war facing a severe conventional imbalance. It could not expect to defeat the combined military power of the United States and Israel directly. Its strategy therefore centred on imposing costs and extending the conflict.

Hormuz provided the most important instrument. Before the war, more than twenty million barrels of oil and petroleum products moved through the strait each day. By the second quarter of 2026, that flow had fallen to roughly five million barrels per day.

Iran did not need to establish complete and permanent control over the strait. Missile threats, drone attacks, mines and uncertainty were enough to deter commercial traffic and raise insurance costs. The commercial response magnified the military effect.

This converted geography into bargaining leverage. Continued American military pressure could now generate costs for Gulf exporters, Asian energy importers and consumers far removed from the fighting. Iran made the duration of the war economically consequential for other governments.

There is a limit to this strategy. Restricting Hormuz also harms Iran’s own exports and encourages other states to build routes that reduce the strait’s future importance. Iran is consuming part of its long-term leverage to survive the immediate conflict.

The central point is that Iran has leveraged its position within an interdependent energy network to offset its conventional weaknesses. It cannot control the war’s military outcome, but it can influence the costs imposed by its continuation.

Widening GCC fissures

Why did a shared Iranian threat divide the Gulf states instead of unifying them?

Shared exposure did not produce a unified Gulf response.

The Gulf states faced Iranian missile and drone attacks, which interrupted shipping and cast major doubts about the reliability of American protection. Yet they reached different conclusions about how to manage those risks.

Gulf states sought diplomacy first. They communicated to Iran they would not allow their sovereign land to be used for U.S. offensive attacks against Iran.

Iran still attacked following the U.S. invasion. The Gulf states were stuck between an adversary that they sought a diplomatic option to resolve, while the Trump administration sought a military solution, undermining the Gulf position with Iran and ultimately exposing them to Iranian retaliation.

The UAE moved closer to the U.S. security umbrella and leaned on its relationsihp with Israel. Its experience during the Iranian attacdefenseforced the value of integrated detection and interception capabilities. At the same time, Abu Dhabi has continued to preserve other important commercial and political relationships where possible, including with China.

Saudi Arabia pursued a broader form of security redundancy. It remained dependent on important American capabilities, but expanded defense cooperation with Pakistan and Türkiye. Riyadh also combined pressure on Iranian proxies with maritime diplomacy and de-escalation channels.

Oman and Qatar placed greater emphasis on negotiation. Their objective was to prevent Iran from exercising exclusive practical control over passage through the strait and to establish procedures that could restore a degree of predictability.

These approaches reveal different threat perceptions and different levels of trust in existing arrangements. They also expose competition within the GCC over who will define the region’s future security architecture.

The result is a Gulf that may become more operationally connected through missile warning, maritime security and infrastructure, while becoming less politically unified. Governments are preparing to cooperate in some areas without accepting a single regional strategy.

The emergence of regional blocs

If confidence in American protection is weakening, where will Gulf states look for additional security?

The search for additional protection is encouraging the formation of new alignments.

The Mecca Pact links Saudi Arabia, Türkiye and Pakistan. Each participant contributes a different form of capability. Saudi Arabia brings capital and strategic geography. Türkiye brings a growing defense industry and operational experience. Pakistan contributes significant military capacity and nuclear status.

The agreement reflects Riyadh’s desire to reduce its dependence on a single external guarantor. It also gives Türkiye and Pakistan a greater role in Gulf security.

A second alignment has developed around the Abraham Accords and the security relationship between the UAE, Israel and the United States. The Iran war has pushed that relationship beyond diplomatic normalization toward practical defense integration. Missile warning, intelligence exchange and air defense have become more important than the original economic symbolism of the accords.

These alignments do not yet constitute two coherent blocs. Their members retain relationships across the emerging divide. Saudi Arabia still depends on American systems. The UAE trades extensively with China. Türkiye remains a NATO member while competing with Israel in Syria. India participates in I2U2 but retains its historical rivalry with Pakistan.

The danger is that new alignments could institutionalize existing rivalries and push the battleground back into countries like Syria, Lebanon, and Jordan, which are still struggling to recover from war and its impacts across the Levant. The advantage is that they provide additional security capacity as confidence in American guarantees weakens. But that security confidence may actually proliferate insecurity if competing blocs enter into a protracted security dilemma.

The emerging order will therefore contain both bloc formation and continued hedging.

Ukraine and Iran as connected crises

What happens when two distant wars begin competing for the same weapons while exchanging battlefield technologies?

The wars in Ukraine and Iran are increasingly connected through shared military and economic networks.

Iranian Shahed technology became an important part of Russia’s campaign against Ukraine. Russia expanded and adapted those systems, creating a sustained threat that forced Ukraine to develop lower-cost methods of interception.

Those lessons are now moving into the Gulf. Ukrainian specialists have assisted Gulf states facing Iranian drones. Ukrainian interceptor technologies offer an alternative to using expensive Patriot missiles against relatively inexpensive targets.

Resources also move in the opposite direction. The United States and its partners possess limited stocks of high-end air-defense interceptors. Every interceptor deployed or expended in the Gulf may affect what is available for Ukraine. The two theatres therefore compete for parts of the same Western defense industrial base.

The economic connections are equally important. Both wars affect energy prices, sanctions enforcement, shipping and food security. Russia can benefit from higher oil prices and the diversion of American attention. Iran benefits from military and technological exchange with Russia. Ukraine may gain new markets for its battlefield innovations and stronger relationships with Gulf governments.

This is battlefield connectivity within a polycrisis. The conflicts remain distinct, but weapons, technologies, capital and strategic attention increasingly move between them.

China’s Iran quagmire

Does China gain more from the Iran war than it stands to lose?

China both gains and loses from the Iran war.

Beijing gains where the conflict increases its bargaining position. Iran’s isolation can provide China with discounted energy and favourable commercial access. A prolonged war absorbs American attention and munitions. China’s relationship with Tehran also gives it some political influence at a moment when Washington wants help limiting Iranian oil revenues and reopening Hormuz.

China loses where escalation threatens the broader network of relationships on which its interests depend. It imports substantial energy from the Gulf, maintains citizens and investments across the region and trades with Iran’s Arab rivals. Disruption to Hormuz harms Chinese supply chains even if some Chinese vessels receive preferential treatment.

The quadrant measures two separate variables. The vertical axis shows China’s exposure to the effects of the conflict. The horizontal axis shows its influence over the conflict’s outcome.

China currently has considerable exposure but limited control over military escalation. It possesses economic influence in Tehran, but it cannot compel Iran to accept American terms. It can obtain selective advantages without assuming responsibility for Iran’s war aims.

Beijing has therefore largely chosen neutrality and resilience. It provides Iran with political and economic space while avoiding a military commitment. It protects itself through energy diversification, strategic reserves and alternative shipping arrangements.

The war gives China leverage, but the value of that leverage declines as the conflict threatens its wider regional interests.

Iran within US–China bargaining

Why would China help Washington contain Iran without receiving something in return?

Iran is also becoming part of the wider bargaining relationship between Washington and Beijing.

Washington wants China to reduce Iranian oil purchases, restrict payment channels and use its influence to support stable passage through Hormuz. Chinese participation would make American economic pressure far more effective because China remains Iran’s most important economic partner.

Beijing will not treat those steps as cost-free contributions to American war aims. China wants continued access to discounted oil, protection for its banks and companies and stability in the Gulf without accepting military responsibility.

Any cooperation with China could become linked to other bilateral disputes. Beijing may seek concessions involving tariffs, technology restrictions or sanctions on Chinese firms. Washington may attempt to threaten secondary sanctions while avoiding measures severe enough to trigger direct Chinese retaliation.

The two powers exercise different forms of chokepoint power. The United States controls access to the dollar, advanced technologies and important financial networks. China occupies essential positions in industrial production, mineral processing and commercial relationships with Iran.

This does not guarantee cooperation. It creates a bargaining space in which Iran becomes connected to issues far beyond the Middle East.

The outcome will depend partly on whether Washington views China primarily as a necessary partner in containing the war or as an economic competitor whose relationship with Iran must itself be confronted.

Slide 12 — A post-Hormuz political geography

If exporters bypass Hormuz, where will the next centres of leverage emerge?

Efforts to bypass Hormuz may change the political geography of the region.

Saudi Arabia already possesses routes to the Red Sea, while the UAE can move some oil through Fujairah outside the strait. Iraq has renewed interest in routes moving west through Syria, Jordan and Türkiye toward the Mediterranean or Red Sea.

The proposed rehabilitation of the Iraq–Baniyas pipeline is particularly important. If developed at scale, it could redirect a significant share of Iraqi exports away from Hormuz. Other proposed road, railway, electricity and pipeline projects would increase the importance of Syria and Jordan as transit states.

This creates economic opportunity. Transit fees, port activity and infrastructure investment could provide Syria with income and regional relevance during reconstruction. Jordan could strengthen its position as a bridge between the Gulf, Iraq and the Mediterranean.

It also introduces new dependencies. A pipeline around Hormuz still depends on transit governments, physical security, operators and ports. Syria’s weak institutions could allow corridor revenues and strategic assets to become concentrated among domestic elites or external actors.

The movement of infrastructure westward also creates competition. Saudi Arabia, Türkiye, Qatar and Iraq see potential value in Syria’s connectivity. Israel views expanding Turkish influence near its borders as a security concern.

Every route around an existing chokepoint creates new infrastructure that someone else can control.

Impacts on global order

What kind of international order emerges when states no longer trust a single guarantor, route or supplier?

Four broader changes emerge from these cases.

First, confidence in American security guarantees is eroding. Regional actors are no longer assuming that reliance on Washington will provide sufficient protection in every crisis. They are building additional capacity and relationships. This has been felt as far as Australia.

Second, Gulf diversification is encouraging new regional alignments. These arrangements may create security redundancy, but they can also intensify competition between blocs whose memberships and commitments remain unsettled.

Third, attempts to reduce reliance on Hormuz are shifting the geography of energy and trade toward the Levant, the Red Sea and the eastern Mediterranean. This creates new opportunities for transit states while moving strategic competition into new corridors.

Fourth, governments are reassessing external dependence across critical industries. They are localizing production, accumulating reserves and diversifying suppliers for energy, defense systems, semiconductors and critical minerals.

Together, these changes point toward a more multipolar order. Yet multipolarity alone does not tell us how that order will function.

The future direction of American foreign policy remains an important organizing principle. Sustained American engagement could keep these overlapping arrangements connected to a wider security system. A more selective or inconsistent American role could accelerate the development of competing blocs. Governments may instead continue hedging, maintaining different partners for different parts of their security and economic needs.

The Iran war is accelerating this transition because it has exposed the costs of depending on any single guarantor, route or supplier.

Future considerations

Which political decisions will determine whether these disruptions fade or become permanent features of the international order?

Several developments will determine where this process moves next.

American domestic politics will shape the duration of the conflict and the credibility of US commitments. Midterm and presidential elections may alter public support for the war, sanctions policy and the resources available for regional security.

The battlefield relationship between Ukraine, Iran, Russia and the Gulf may deepen. Iranian and Russian drone technologies, Ukrainian interception methods, and Western air-defense systems already move or compete across theatres. Longer-term industrial partnerships will reinforce these connections, such as the Iran-Russia drone axis.

US–China relations will shape the effectiveness of economic pressure against Iran. Washington may need Chinese cooperation to constrain Iranian revenues. Beijing may use that demand to seek concessions elsewhere. Expanded secondary sanctions could instead accelerate the development of payment and trading systems designed to reduce exposure to American financial power.

The postwar settlement will also determine Iran’s position. A heavily isolated Iran may become more dependent on China and Russia. A settlement that restores broader economic access could give Tehran greater room to balance among competing powers.

The definitive takeaway is that the Iran war has exposed the coercive power embedded within global networks. States are responding by building alternatives, but those alternatives do not remove interdependence. They redistribute it.

The search for resilience will give governments more options. It will also produce new chokepoints, more complicated partnerships and greater uncertainty about which relationships will hold during the next crisis.

Please note that the summaries for each slide provided here are consolidated and edited using GPT 5.6 Sol for this Substack piece. My full notes from this presentation are currently being drafted into a longer paper for publication elsewhere on the same topic of this presentation.

Iranian strikes on AI infrastructure should change how Gulf governments think about the security of strategic compute. For years, data center protection has focused on cyber intrusion, physical access, power redundancy, and disaster recovery. Those measures remain necessary, but they are no longer sufficient. As artificial intelligence becomes increasingly important to military operations, intelligence analysis, financial systems, public administration, and critical industries, the infrastructure supporting it is becoming a plausible target in conflict.

See my new data product: Data Center Strikes in the 2026 Iran War, for further context for this paper.

Subscribe now

In high-risk environments, data centers (and compute infrastructure more broadly) must be considered strategic infrastructure because of the risk that any disruption could degrade their military effectiveness, economic activity, and administrative capacity. This raises a central policy challenge: how to protect critical AI-infratructure, while ensuring that essential compute capabilities remain online even if domestic infrastructure is partially damaged.

I have developed a useful framework, the conflict resilience triangle model, as a potential solution. This is built around three principles: defend, disperse, and recover.

Defend

First, states have to improve the physical protection of critical AI infrastructure. High-value data centers in conflict-prone regions should be incorporated into national critical-infrastructure and defense planning rather than left primarily to private security arrangements. This may require hardened construction, redundant energy supplies, protected telecommunications, layered perimeter security, and integration with national air and missile defense networks. In some cases, governments may need dedicated point-defense systems capable of intercepting drones, cruise missiles, or other precision weapons approaching particularly important facilities.

This does not mean that every commercial data center needs individualized military protection (though there is a case to be made for local-level point-and-shoot counter-UAS solutions that are becoming more commercially viable for local deployment). Governments will need to distinguish ordinary computing facilities from infrastructure supporting essential state functions, military applications, financial systems, or nationally significant AI capabilities. Priority AI sites should be treated comparably to power plants, bases, ports, or airfields.

Physical defense, however, cannot guarantee continuity. Fixed infrastructure remains vulnerable to saturation attacks, sabotage, power disruption, and repeated strikes. Even a well-defended facility can be disabled. Resilience therefore depends on ensuring that the loss of a site does not produce the loss of the capability it supports.

Disperse

States need the ability to distribute workloads across multiple facilities during periods of heightened risk. This may involve building additional domestic sites or purchasing backup services from commercial cloud providers. More importantly, states should develop prearranged mechanisms for transferring eligible critical workloads to trusted infrastructure abroad during emergencies.

Under a framework such as the U.S.-led Pax Silica, participating states could negotiate reciprocal continuity agreements for strategically important data processing. If data centers in the United Arab Emirates were degraded during a regional conflict, designated government or commercial workloads could shift temporarily to certified facilities in the United States, Germany, Australia, Japan, or another trusted partner. Similar arrangements could operate in reverse when other members faced disruption.

The immediate objective would be tactical and strategic continuity. A government might lose part of its domestic compute infrastructure without losing the ability to process timely intelligence, manage public services, keep financial transactions online, or operate critical AI workloads.

This system would require extensive preparation. Workloads cannot be moved reliably across borders during a crisis unless legal, technical, and operational arrangements are already in place. Participating countries would need common standards for encryption, identity management, network security, hardware certification, data handling, access controls, and incident response. They would also need to determine which workloads were eligible for relocation, who could authorize an emergency transfer, how foreign-hosted systems would be supervised, and when processing would return to domestic infrastructure.

A credible arrangement would also require preallocated reserve capacity. Commercial providers often operate close to capacity during periods of high demand; a state cannot assume that sufficient compute will become available once a conflict has begun.

The model resembles military logistics. Defense partnerships do not assume that every base, port, or supply route will remain operational during conflict. They preserve capability through redundancy, access agreements, prepositioning, and rerouting. Strategic compute should increasingly follow the same logic: a network of trusted continuity partners providing access to pre-certified infrastructure, reserved capacity, and rapid workload migration when domestic facilities become unavailable.

Dispersal would also reduce the strategic value of attacking AI infrastructure. If critical workloads can move rapidly to predetermined facilities in trusted partner countries, destroying any single site is less likely to produce a sustained loss of national capability. Attacks may still impose economic costs and temporary disruption, but their ability to alter the broader balance of power diminishes. The relevant security objective is the survivability of function.

Recover

States need to be able to recover domestic systems if they go offline. Moving workloads overseas may sustain operations during a crisis, but it is not an indefinite substitute for domestic capacity. States must also be able to repair, replace, or rebuild damaged infrastructure quickly.

Recovery planning should include assured access to servers, networking hardware, cooling systems, power components, and the specialized personnel needed to restore operations. Governments may need strategic reserves of critical hardware (though sourcing may be difficult due to cost and politicization of chip access), priority-access agreements with suppliers, and alternate sites that can be activated when damaged facilities require extensive reconstruction. Telecommunications and energy infrastructure belong in the same framework: restored compute has little operational value without reliable power and secure connectivity.

This is particularly important because advanced AI hardware can be difficult to replace on short notice. Export controls, supply-chain bottlenecks, long procurement cycles, and limited access to high-end chips could leave damaged facilities offline for months if states are not prepared. Resilience planning must therefore cover the entire supporting ecosystem.

A New Role for U.S. Technology Partnerships

The resilience triangle creates an opportunity for the United States to broaden its international AI partnerships. Current policy debates focus heavily on access to semiconductors, frontier models, cloud services, and export controls. These policies shape which countries can build advanced AI capabilities but offer few answers on how those capabilities would be sustained during conflict.

Washington may need to build contingency planning into major AI deals from the outset, especially with partners in conflict-prone regions. Access to American hardware and software could be paired with guaranteed emergency compute capacity in trusted third countries, allowing critical workloads to continue if domestic infrastructure is disrupted. For states investing heavily in exposed AI infrastructure, that assurance could become a major strategic benefit of partnership with the United States.

A continuity network would also reinforce U.S. technology standards. Participation would encourage states to adopt compatible cybersecurity practices, trusted hardware, common cloud architectures, and interoperable operating procedures. These requirements could gradually create a more integrated technological ecosystem, binding alignment with the United States to a shared interest in operational resilience.

These agreements would require some clear boundaries. National security and other sensitive workloads cannot move through the same channels as ordinary cloud traffic. States would need to decide in advance which functions could be relocated, which could operate abroad only in secured sovereign environments, and which must remain on national territory. Lower-sensitivity applications could shift to trusted commercial clouds, while higher-sensitivity workloads would require isolated facilities governed by tighter legal, access, and technical controls.

Protecting Capability, Not Only Infrastructure

Compute is rapidly becoming a new category of wartime vulnerability. As governments concentrate economic and administrative functions in a small number of highly visible facilities dependent on complex energy, telecommunications, and hardware supply chains, states must have contingencies in place to ensure AI continuity.

AI deals with wartime dependence long before a crisis begins. How much tech capacity will a state retain during conflict? That depends largely on structural factors such as infrastructure location, cloud architecture, supplier dependence, and cross-border access agreements. For Gulf states, this reframes AI sovereignty around the ability to retain decision-making power after disruption. Domestic ownership alone cannot necessarily provide that assurance. External capacity can strengthen national control if/when preset agreements guarantee access, preserve legal authority over workloads, and prevent foreign providers from withholding support during a crisis.

In the broader U.S.-China race, Washington’s most consequential offer to AI partners may be the assurance that critical capabilities will remain available when domestic infrastructure is disrupted.

© Jesse Marks, Coffee in the Desert

AI Use Statement: The visual graph of the model outlined in this paper was created with Claude. I also used Claude Opus 4.8 for copy-editing.

Over the past two months, I have been gradually rolling out a new initiative, The Gulf AI Monitor, a platform dedicated to examining the development, politics, and security of artificial intelligence across the Gulf Cooperation Council states.

The initiative has two layers.

First, the Monitor’s written analysis will be available to all readers. I will also be publishing more public analysis through the platform and will soon issue a call for contributions from researchers and practitioners working on AI and the Gulf. This aims to be a regularly updated public/foreign policy style publication.

The second layer is a pay-for-access GCC AI data platform for researchers, institutions, and policymakers. It provides a comprehensive intelligence platform mapping the GCC AI ecosystems, including a large and growing corpus of data on U.S. and Chinese investments, GCC investments, hyperscaler partnerships, infrastructure maps, interactive dashboards, and in-depth country deep dives into every GCC country’s AI ecosystem and across the stack.

The first data product I am releasing publicly is a map of reported attacks against AI data centers in the Persian Gulf during Operation Epic Fury. The map helps visualize the geographic distribution of these attacks and where Iran has, so far, concentrated its targeting of AI infrastructure.

As data centers and computing capacity become increasingly important to GCC futures and diversification strategies, understanding how this infrastructure is exposed to conflict will become increasingly urgent.

Feel free to see the full tool, here.

Subscribe now

I am just back in the frigid hills of Canberra, Australia, after two months in Europe and the Middle East for a series of dialogues, the UN AI summit, and more. While in Jordan, I had the privilege of sitting down with Eric Olander and the China-Global South Podcast to look more deeply into China’s ongoing diplomatic role in the war in Iran and the limits of its efforts. We touch on GCC, Hormuz, deterrence, dual use technology, AI, and much more.

Sinica

The Limits of China’s Iran Diplomacy

The U.S. and China clashed at the United Nations last week, with each accusing the other of fueling instability in the Persian Gulf. Washington has long argued that Beijing plays a pivotal role in the conflict as the largest buyer of Iranian oil and a supplier of dual-use technologies that could support Tehran’s military…

Listen now

a month ago · 8 likes · 1 comment · Eric Olander

Make sure to listen! Also, Eric has done an exceptional series of dialogues with others in this space including . Recommend listening to those as well.

Subscribe now

The collapse of the U.S.-Iran MOU places the GCC states in an increasingly precarious strategic position as Iranian missiles and drones again strike Gulf territory. The GCC states are highly exposed to the consequences of the conflict yet possess only limited influence over its conduct or ultimate outcome. They have been largely sidelined from both the MoU negotiations and the broader U.S.-Iran nuclear dialogue, despite being among the actors most directly affected by their collapse. Meanwhile, the potential reactivation of Iranian-backed proxy networks (particularly the Houthis) raises the prospect of a two-front war, threatening both Gulf territory and maritime trade while increasing the likelihood of their renewed military involvement in Yemen.

Subscribe now

As the prospect of a prolonged U.S.-Israeli-Iranian war becomes more likely, the Gulf faces three broad strategic pathways. At the heart of each lies a question of Gulf agency.

Option 1: Absorb the Attacks and Lean into Restraint

Under the first scenario, Gulf states avoid direct military escalation and instead continue to absorb periodic attacks, disruptions, and threats against their territory and critical infrastructure. While this may be the least escalatory option in the short term, it risks locking the GCC into a frozen conflict that has diminishing strategic returns. Over time, the balance of costs would increasingly shift against Gulf interests. Persistent insecurity would raise insurance premiums and transportation costs, deter foreign investment, complicate economic diversification efforts, and weaken the Gulf’s reputation as a stable economic hub. Even attacks that inflict limited physical damage could have disproportionate economic consequences by increasing the perceived risk in the region.

The political costs are likely to accumulate. Prolonged restraint could gradually diminish Gulf influence as regional and international actors come to view the GCC as economically indispensable but strategically vulnerable. This shapes investor trust in the region and has a cascading economic effect over the long term. A strategy of strategic restraint could also be interpreted as strategic impotence or a sign of consent to a new order, encouraging further Iranian coercion rather than restoring deterrence. The result, taken to its logical conclusion, would be a gradual erosion of credibility, autonomy, and long-term economic competitiveness.

Perhaps most importantly, this outcome risks normalizing Iranian coercive leverage over the Gulf. Even without achieving outright military dominance, Tehran could acquire a de facto veto over aspects of Gulf security and foreign policy simply through the persistent threat of missile, drone, or proxy attacks. In effect, Iran would not need to defeat the Gulf militarily. It would only need to convince Gulf leaders that the economic and political costs of resisting its coercive pressure outweigh those of accommodation. Such a dynamic would represent a significant shift in the regional balance of power, allowing Iran to exercise influence disproportionate to its conventional military or economic strength.

Option 2: Conduct Limited Strikes as a Form of Defensive Deterrence

The second pathway would be one or more of the Gulf states joining the war against Iran and its proxies. In this scenario, one or more Gulf states conduct limited offensive operations against Iranian missile, drone, naval, or command-and-control platforms used to launch attacks against Gulf territory. The objective would be to restore deterrence by demonstrating that attacks on the Gulf carry direct and unacceptable costs. In this scenario, offensive military action would aim to re-establish the credibility of Gulf deterrence.

This is the most dangerous pathway. A unilateral GCC strike would almost certainly provoke Iranian retaliation, but with a significant escalation against civilian population centers, energy infrastructure, ports, airports, desalination facilities, financial centers, and other critical infrastructure. Tehran could also seek to widen the conflict horizontally by targeting Gulf states that provided logistical, intelligence, or basing support, regardless of whether they directly participated in the initial operation. Such a strategy would increase the costs of regional cooperation while attempting to fracture Gulf unity, which has already begun to fracture the past few months.

Military retaliation would also expose the limits of collective Gulf security. Although the GCC presents itself as a regional bloc, its members do not necessarily share identical threat perceptions, escalation thresholds, or tolerance for economic disruption. A military response initiated by one state could therefore impose security and economic consequences on all, even in the absence of collective political consensus. Rather than strengthening Gulf cohesion, retaliation could deepen divisions between states favoring coercive deterrence and those seeking continued accommodation or de-escalation with Tehran.

The reliability of U.S. security guarantees is also a major constraint. Gulf leaders cannot assume that Washington would provide unlimited military protection following a GCC-initiated escalation, particularly if the United States is simultaneously seeking to reduce its own direct involvement in the conflict. From the Gulf perspective, the greatest danger is joining a war that the U.S. ultimately chooses to exit, leaving GCC states in direct confrontation with Iran across the Persian Gulf without a credible external guarantor. The consequences would be tremendous. Gulf states would bear the immediate military, economic, and political costs of escalation while the U.S. looks for the exit door. An offensive retaliation could therefore leave the GCC more fragmented and deeply entangled in the war with Iran.

This scenario becomes more plausible if the conflict expands with the Houthis' opening of a Yemeni front. A renewed campaign against commercial shipping through the Red Sea and the Bab al-Mandab, combined with continued Iranian attacks against Gulf territory, would create the prospect of a two-front conflict. Gulf states would simultaneously face direct threats to their territory and critical infrastructure while also confronting disruptions to the only maritime alternative to the Strait of Hormuz in the broader region.

As the economic and political costs of continued restraint accumulate, Gulf leaders may conclude that deterrence cannot be restored through continued restraint alone. If attacks become normalized and the security of both the Persian Gulf and the Red Sea deteriorates, limited offensive operations may increasingly be viewed as necessary to re-establish credible deterrence and protect vital national interests. This could rapidly evolve into a broader regional confrontation, drawing the Gulf into a war with uncertain prospects for external military support.

Option 3: A Negotiated Settlement or Regional Non-Aggression Arrangement

I previously argued in the Gulf International Forum that a GCC-Iran non-aggression pact could provide one pathway out of the strategic targeting trap in which the Gulf states increasingly find themselves. This would not necessarily end the broader conflict between Iran and the United States or Israel. It would, however, aim to insulate the Gulf from that confrontation by establishing mutual commitments not to target each other’s territory, civilian infrastructure, commercial shipping, or energy facilities.

This pathway offers the greatest potential (of the bad options on the table) to preserve Gulf economic interests and strategic autonomy. A limited non-aggression arrangement could also establish crisis communication channels, understandings regarding proxy activity, and mechanisms to investigate or contain incidents before they escalate into wider conflict. More fundamentally, it would allow Gulf states to pursue a strategy of regional insulation rather than regional alignment, reducing the likelihood that they become the principal battleground for a conflict driven by external actors.

Yet this is also the scenario over which Gulf states exercise the least control. Its success depends not only on Iran’s willingness and ability to restrain its own military and proxy networks, but also on the United States and Israel refraining from actions that make regional compartmentalization impossible. Gulf governments could negotiate a bilateral or multilateral deal with Tehran, but they cannot ensure that the wider conflict will respect those political boundaries.

There is also a fundamental credibility challenge. Any non-aggression arrangement shares the same weaknesses as the MoU and would need to survive periods of intense escalation while providing meaningful consequences for violations. Without robust verification mechanisms, reliable channels of communication, and reciprocal commitments to restraint, such an agreement risks becoming little more than a temporary political declaration that collapses at the first major crisis.

The political conditions necessary for a durable non-aggression pact are absent. Renewed Iranian attacks against Gulf territory are unlikely to compel GCC states toward accommodation or suing for peace, just as continued U.S. and Israeli military pressure reduces Tehran’s incentives to offer meaningful security guarantees. The option, however, should not be dismissed as strategically irrelevant and should remain on the table. Both Iran and the Gulf may eventually conclude that some form of regional security arrangement is preferable to an indefinite cycle of coercion and retaliation. A non-aggression framework may be politically unattainable at the moment, but it could be one of the few long-term pathways out of the conflict.

The underlying strategic dilemma

The central dilemma at stake in each scenario is Gulf agency. Preserving agency and optionality will likely lead the region’s leaders to pursue all three pathways at once. Gulf states may initially absorb attacks while strengthening their defensive and retaliatory capabilities, while continuing to pursue diplomatic engagement with Tehran. In practice, this means showing near-term restraint to avoid unnecessary escalation, building credible hard military capabilities to restore deterrence if required, and maintaining diplomatic channels with Iran and the U.S. to insulate the Gulf from a broader regional conflict.

Continued restraint will not reduce the immediate risk of escalation but may normalize the Gulf’s vulnerability to coercion. Limited military retaliation, if used, may restore deterrence but risks making the GCC states active belligerents in a conflict whose trajectory it cannot control. Diplomacy offers the most sustainable long-term outcome, but its success depends heavily on decisions made in Tehran, Washington, and Jerusalem, not in Gulf capitals. In all three scenarios, Gulf agency remains severely restrained.

The challenge facing GCC leaders is preserving the ability to choose between restraint, retaliation, and accommodation independently of external pressure or coercion. The Gulf's overriding strategic objective should be to avoid becoming either the principal battleground or the principal bargaining chip in a conflict driven by actors over whom it has little control. Preserving that strategic optionality requires maintaining diplomatic channels with Iran, strengthening collective defense and credible deterrence, protecting critical maritime and energy infrastructure, and making clear to external partners that Gulf territory cannot be treated as an expendable platform for pursuing strategic objectives that the GCC neither defined nor fully shares.

Over the past year, I have been diving deep into the academic and technical undercurrents in AI research and discourse. AI now generates more commentary than almost any one person could reasonably follow on a day to day basis. But much of what appears new, such as the fear of losing control, the search for alignment, the US–China technology race, and the concern that machines may reshape human relationships, all rest on arguments developed before the current model boom.

These four books were among the first recommended to me and are a good introduction to AI. They are a useful place to begin. Together, they show how the field has moved from abstract questions about machine intelligence to a wider debate about power, dependence, work, and what remains distinctively human.

Subscribe now

Superintelligence — Nick Bostrom

Published in 2014, Nick Bostrom’s Superintelligence gave the AI safety field much of the language it still uses. His central argument is that a machine need not hate humanity to become dangerous. Intelligence and goals are separate. A highly capable AI system could pursue an objective that bears no relationship to human values. Bostrom’s “instrumental convergence” thesis further suggests that systems with very different goals may still adopt the same intermediate behaviors, including acquiring resources, preserving themselves, and resisting attempts to shut them down (e.g. AI survivalism).

Some of the book has aged. Bostrom wrote this before large language models and the scaling race dominated AI discourse. His account of a sudden intelligence explosion paints one of the earliest and clearest possible trajectories of AI. Naturally, this is not biblical prophecy, but it helps set a somber tone for creative thinking on future AI scenarios. The book remains essential because it explains why advanced AI came to be seen as a control problem rather than simply another powerful technology. Even readers who reject its conclusions will recognize its influence across current debates about alignment, deceptive behavior, and catastrophic risk.

Human Compatible — Stuart Russell

Stuart Russell starts with a simple criticism of how AI is usually built: humans specify an objective, and the machine tries to achieve it as effectively as possible. The problem is that people are bad at specifying exactly what they want. A sufficiently capable machine may follow the formal objective while violating its intent. Russell’s alternative is to build systems that remain uncertain about human preferences, learn from human behavior, and therefore retain a reason to accept correction or allow themselves to be switched off.

The strength of Human Compatible is that it turns Bostrom’s warning into an engineering problem. Russell asks what would have to change in the design of AI systems to prevent them from becoming dangerous as they grow more capable. The difficulty is that human preferences are inconsistent, unstable, and not always visible in our behavior. A machine trying to infer what people want may learn their weaknesses as easily as their values. Even so, the book remains one of the clearest accounts of what a genuinely deferential AI system might look like.

AI Superpowers — Kai-Fu Lee

Kai-Fu Lee’s AI Superpowers was one of the first widely read books to present artificial intelligence as a geopolitical contest between the United States and China. Lee argued that AI had entered an “age of implementation,” in which data, capital, entrepreneurial speed, and the ability to deploy technology at scale would matter more than the original scientific breakthroughs. He believed these conditions favored China and would push the world toward a two-power technological order, leaving other countries dependent on American or Chinese systems.

The frontier-model race since 2022 has complicated that thesis. Research breakthroughs, advanced chips, compute, and infrastructure proved more important than Lee expected. But his larger framework remains useful. AI power still depends on talent, capital, firms, data, energy, and state support, the same inputs now driving sovereignty debates in Europe, India, and the Gulf. Read today, the book is less valuable as a prediction that China will win than as an explanation of how AI became embedded in a wider struggle over economic power and technological dependence.

The Artifice of Intelligence — Noreen Herzfeld

Noreen Herzfeld approaches AI from a moral and philosophical vantage point, setting her apart from the other authors on this list. Rather than beginning with what machines can do, she begins with what it means to be human. Drawing on theology and computer science, Herzfeld argues that personhood is relational, grounded in mutual recognition, vulnerability, speech, listening, and care. A machine may reproduce the outward appearance of these qualities, but it does not participate in a relationship in the same way another person does. Her argument also raises deeper questions about the soul, faith, and whether human value can ever be reduced to intelligence or information processing. Ultimately, she asks what living alongside robots may do to the way humans understand themselves, one another, and their relationship with God.

Where to begin

Read Bostrom for the AI safety framework, Russell for a proposed response to the AI issue, and Kai Fu Lee for an overview of the AI geopolitical contest. Herzfeld examines the moral and philosophical questions surrounding AI and its impact on humans. Together, the four books show why AI cannot only be understood as an engineering problem. It is a problem of political power, economic dependence, institutional design, and human judgment. These authors provide a way to distinguish enduring questions from fashionable ones and a foundation for understanding the arguments that will shape the next decade.

Feel free to recommend any new books you may have!

Subscribe now

After traveling the past month and meeting with a number of sides of the Iran conflict since the MoU signing, I wanted to lay out some thoughts for consideration.

The recent conflict appears to have fundamentally reshaped how many Iranian strategists think about deterrence, sovereignty, and regional power. While it remains too early to know which lessons will ultimately prove enduring, several themes consistently emerge from discussions about how Tehran is interpreting the war.

First, the conflict reinforced the strategic importance of the Strait of Hormuz. From Iran’s perspective, the war demonstrated that the Strait could be weaponized against Iran just as easily as it can be exploited by Iran as deterrence. Rather than diminishing its importance, the conflict strengthened Tehran’s belief that exercising greater control over Hormuz is essential to overcoming future vulnerabilities. For many hardliners, the Strait has become an indispensable source of leverage in negotiations that Iran cannot afford to surrender. The prevailing view in Iran is that the pre-war status quo regarding Hormuz is no longer acceptable, and GCC calls for a return to it are a non-starter.

This thinking extends simply military strategy. The Iranian leadership seeks to exercise “legal” control by claiming territorial sovereignty over its portion of the Strait of Hormuz, reportedly in coordination with Oman. The objective is both the obvious gain in operational control and the gradual establishment of a new regional arrangement (the new status quo) that Gulf Arab states may eventually accept. This effort includes advancing legal arguments to build an international legal case supporting Iran’s interpretation of its rights in the Strait. From this view, the use of force as a de facto control mechanism to close and control the Strait is a wartime necessity that would eventually be recognized as legal authority down the line.

The war also appears to have reinforced an important distinction in Tehran’s thinking regarding the nuclear issue. Iranian officials differentiate between preserving the country’s nuclear program and maintaining highly enriched uranium stockpiles. Their interpretation of the conflict is that the United States has, in practice, accepted Iran’s right to maintain a civilian nuclear program. Enrichment to approximately 3.5 percent is viewed as compatible with that objective, while the accumulation of 60 percent enriched uranium was primarily intended as a source of strategic leverage rather than an end in itself. The continuation of the nuclear program remains a non-negotiable red line on the Iranian side. The prevailing Iranian logic might argue that if the source of the threat (Israel or United States’ military action) is removed, then Iran would cease its pursuit of a nuclear weapon (or weapons-grade uranium). This logic, however, leaves unresolved a broader regional concern from the Gulf Arab States who view Iran’s projection of power through direct strikes and its regional proxies as a threat independent of the nuclear file. It seems that a cessation of the war may end the strikes on GCC states, but it does not resolve the question of Iranian proxies, the nuclear issue, or Hormuz.

Domestically, the war appears to have altered Iran’s political environment. While it did not necessarily increase support for the government itself, it appears to have fostered greater national unity in the face of external attack. Some interlocators argued that the prospect of externally supported regime change has effectively disappeared for the foreseeable future. As one assessment put it, “There will not be any kind of riots or protests because no one can go to the streets due to the risk attack.” Whether accurate or not, this perception may increase the regime’s confidence and willingness to escalate during future crises.

Perhaps the most significant lesson regards deterrence. Iranian strategists increasingly believe they have established the legitimacy to strike the United States and its regional partners directly from Iranian territory as part of a credible deterrence posture. At the same time, Iran’s participation in the negotiations are viewed as a means of preserving and rebuilding what remains of Iran’s regional network, particularly Hezbollah. Rather than abandoning the Axis of Resistance, Tehran appears intent on reconstructing it while pairing it with a stronger emphasis on homeland deterrence through direct attacks on GCC states. The lesson Iranian officials drew from the period following October 7 was that Iran’s restraint toward Israel was interpreted as weakness and ultimately invited further Israeli military action in Lebanon and Syria. The result is likely a more self-confident Iran seeking to combine direct deterrence from its mainland with a renewed regional forward-defense strategy. That is not to say that Iran is actively rebuilding Hezbollah (as such thinking is often justified by Israeli forces to justify the ongoing invasion). Rather, the existence of the debate over the continued (or expanded use) of proxies within Iran merits further consideration.

Looking ahead, the logic suggests that if another major conflict emerges, Iran would likely seek to expand the battlefield beyond the immediate theater by opening additional military fronts. Such an approach would be intended to complicate its adversaries’ military planning, increase the costs of escalation, and reinforce the credibility of its broader deterrence strategy.

Whether these lessons accurately reflect the realities of the conflict is ultimately less important than the fact that they appear to be shaping Iran’s strategic outlook. Wars do not simply change the balance of power; they change how states understand power itself. If these interpretations take hold within Tehran’s strategic community, they are likely to influence Iranian decision-making long after the fighting has ended.

Subscribe now

China’s State Council regulation on outbound investment took effect on July 1, 2026, giving Beijing its first comprehensive, cabinet-level legal framework governing how Chinese capital moves abroad. Premier Li Qiang signed the 34-article regulation, formally Decree No. 837, on June 1. It replaces a patchwork of ministerial rules that the National Development and Reform Commission, the Ministry of Commerce, and the State Administration of Foreign Exchange had administered separately for two decades, each covering a different slice of the approval, filing, and remittance process.

Subscribe now

The new regulation closes loopholes that let Chinese firms move technology, data, and capital offshore through Cayman or Singapore (remember the Manus AI dispute?) holding structures, personnel transfers, and licensing arrangements that fell outside the old rules.

Beijing is also building state machinery to back its firms abroad that increasingly resembles what Washington has assembled over the past eighteen months. Both governments are moving toward becoming commercial protagonists by using licensing power, equity stakes, security review, and retaliatory authority to help their companies win deals, especially those at the intersection of commerce and national security (Infrastructure, AI, and critical minerals).

A coordinated state support system

In 2025, Chinese critical mineral giant Shenghe Resources lost a bid to U.S. company MP Materials for a venture with Saudi Arabia’s Ma’aden. Though Shenghe (which also previously held stakes in MP) is one of the world’s leading rare earths refiners and processors, President Trump’s decision to put U.S. sovereign backing behind MP Materials pushed the Saudi company to pick the U.S. champion. Shenghe lost the deal. For the U.S., it marked an example of an emerging framework of U.S. economic statecraft that is gradually leveraging U.S. state influence to push for economic outcomes in strategic industries.

China has also long utilized its government bodies and embassies abroad to advocate for Chinese industry in foreign markets—virtually every country does it. The difference now is that what was a more decentralized and ad-hoc process is now being centralized and institutionalized into policy.

The new regulation directs Chinese government bodies to support outbound firms with a more concerted, state-backed push across diplomacy, finance, taxation, customs, insurance, intellectual property protection, dispute resolution, logistics, and consular affairs to make Chinese companies more competitive. It replaces the old approval-and-filing model with a coordination framework that links government authorities, professional institutions, industry associations, and trade promotion organizations.

Investments now fall into three categories: encouraged, restricted, and prohibited. Those deemed to “affect or may affect national security” undergo a dedicated security review conducted jointly by the NDRC and MOFCOM. A separate provision requires Chinese entities facing foreign litigation or regulatory demands for evidence to first clear those disclosures against Chinese state secrets, data security, and export control law. This extends Beijing’s regulatory power into disputes that used to play out entirely on foreign soil.

The regulation also gives Beijing explicit retaliatory authority. It authorizes the government to investigate discriminatory treatment of Chinese investors abroad and to respond under the Anti-Foreign Sanctions Law, with countermeasures that may include restricting imports and exports, barring investment in China, blocking Chinese entities from dealing with targeted foreign parties, and limiting entry for people, goods, or transport. None of this mandates automatic retaliation, but it gives Beijing standing domestic legal authority to escalate as the state deems necessary.

The timing of the regulation is linked to two recent disputes that exposed a gap in Chinese regulation. Meta’s blocked acquisition of the Chinese-rooted AI startup Manus and the Nexperia dispute involving Chinese-owned chipmaker Wingtech Technology both showed that Chinese capital and technology were exposed to foreign control decisions that Beijing had no formal standing to contest. Article 13 of the new regulation closes that gap by capturing formal equity transfers alongside technology transfers through licensing, dispatched engineers, and cross-border technical training, even when no equity changes hands. Article 15 extends the security review to cover the sale of overseas assets, so an exit now draws the same scrutiny as an entry. Practitioners have taken to calling the offshore-restructuring workaround a “Singapore Wash,” and Beijing’s new rules are built to close it.

Washington’s parallel machinery

Washington has built a similar system over roughly the same period, using different instruments. Like Beijing, it wants to keep national capital and technology under state-aligned control while using the state’s weight to shape where that capital and technology go. The closest mirror to China’s new regulation is the Treasury Department’s outbound investment security program, which practitioners call a “reverse CFIUS.” Implementing Executive Order 14105, Treasury issued a final rule in October 2024 that took effect on January 2, 2025, prohibiting or requiring notification of U.S. investments into Chinese entities working in semiconductors, quantum information technologies, and artificial intelligence. Where CFIUS screens foreign capital coming in, the outbound program screens American capital going out, sorting transactions into prohibited and notifiable categories much as Beijing’s regulation sorts outbound deals into restricted and encouraged ones. Treasury can also nullify, void, or compel divestment of any prohibited transaction, an unwinding power that parallels the reach Article 15 now gives Chinese security reviewers over exits.

Washington has also gone further than screening. In August 2025 the Commerce Department converted $8.9 billion in CHIPS Act grants and Secure Enclave program funds into a roughly 10 percent equity stake in Intel, the only US company still capable of producing leading-edge logic chips domestically. Commerce Secretary Howard Lutnick put it plainly: “We should get an equity stake for our money.” The deal came with a warrant to buy an additional 5 percent if Intel’s foundry ownership fell below 51 percent, a direct hedge against foreign control of the company and potential for undesirable foreign transfer. Two months earlier, the administration had converted a CFIUS review into an instrument of permanent control. As the price of approving Nippon Steel’s acquisition of U.S. Steel in June 2025, the government took a “golden share” granting it veto power over major corporate decisions, including cuts to Nippon’s $11 billion investment commitment, relocation of headquarters, transfers of jobs abroad, and plant closures.

American firms now anticipate this trend rather than resist it. On July 2, the Financial Times reported that Sam Altman proposed giving the U.S. government a 5 percent stake in OpenAI, as part of a broader arrangement under which Washington would hold 5 percent of each leading American AI developer through a “sovereign wealth fund-like” vehicle. The direction of pressure tells two different stories about the U.S. and Chinese systems. In Washington, the country’s leading AI firm invites the state as a shareholder. In Beijing, the state barred Manus AI’s leadership from leaving the country earlier this year. While that image is a bit simplistic, many observers may not seek the deeper nuance. In practice, American firms are buying political protection by offering equity. Chinese firms, meanwhile, trigger state retaliation and control by trying to exit.

Between an outbound screening regime aimed squarely at China, an equity position in a strategic chipmaker, and a standing veto extracted through security review, Washington has assembled a toolbox functionally similar to Beijing’s new regulation. Both sides once used these levers to approve or deny sensitive deals. Both now use them to extend their regulatory reach and shape deal outcomes at home and abroad in pursuit of their respective national interests.

The larger picture

This connects uniquely with the question of “What happened to BRI?” If the first decade of BRI was geared toward pushing Chinese capital out the door, this regulation now governs where it lands, who controls it once it gets there, and what state capacity is needed to ensure both. The same period has shifted Washington from a posture of restricting China’s access to frontier technology to one of actively directing where American technology and capital go and on what terms.

Neither government has fully articulated where this ladder ends. Beijing’s countermeasure authority is discretionary, and how aggressively it gets used will depend on the state of US-China relations rather than the text of the regulation itself. Washington’s golden share and its equity position in Intel and offtake deals with M.P. Materials remain first-term experiments rather than settled doctrine of an emerging “sovereignty wealth fund” mentality. A change in administration could unwind either.

A few weeks ago, I asked 9 AI models whether they believed their creators would act on behalf of humanity or in their own interests in the long-term development and deployment of AI. All of them said, to varying degrees, “there needs to be safeguards”.

The risk of overconcentration came into focus through my PhD research, but it warranted a fuller analysis, especially after Elon Musk became the world’s first trillionaire following SpaceX’s successful IPO. Is any human being prepared for the responsibility that comes with that level of influence? History is full of people whose power grew faster than their wisdom. Wealth, technology, and influence all compound. Character does not compound with them. Now imagine a single actor holding not only that much financial capital, but also an unchecked frontier AI capability alongside it.

Lord Acton, a former British Parliamentarian, stated in the late 19th century that:

“Power tends to corrupt, and absolute power corrupts absolutely. Great men are almost always bad men, even when they exercise influence and not authority; still more when you superadd the tendency of the certainty of corruption by authority.”

It seems the AI models I asked agree with Lord Acton’s general assessment.

Why does this matter?

A world in which a single actor controls the most advanced AI capabilities, with no competitors and no checks on that power, raises an urgent question about how that power would be used. And we are already entering an age defined by single individuals whose wealth and power grant them unprecedented reach.

What Overconcentration of Power Means in AI

Overconcentration describes a structural condition in which a single actor acquires control over frontier AI capabilities so decisive that no other actor can meaningfully constrain it, contest it, or develop independent alternatives. A world in which the United States leads China in AI capabilities reflects a normal geopolitical predicament. A world in which whoever controls frontier AI can determine the political and economic outcomes of every other actor, with no check on that power, is a novel scenario that several AI experts are sounding the alarm about.

[See Rose Hadshar’s “Extreme power concentration,” published by 80,000 Hours]

Political science is useful here because the accumulation of power is among the oldest problems of study within the discipline. Power, however gathered and defined, has always met a limit or maximum threshold. Understanding overconcentration in AI means understanding why those limits have held and why this technology may be the first to suspend them.

Two checks have historically constrained the total accumulation of power, one external and one internal. The external check is the friction of interstate competition. Mearsheimer’s offensive realism takes the pursuit of dominance as the default behavior of states, yet argues that global hegemony remains nearly unattainable for two reasons. First, a state’s bid for hegemony often pushes rival states to form coalitions to counterbalance the rising state. Second, it is difficult for a single state to project sufficient power across oceans and continents to directly control the entire globe. Thus, a state’s hegemonic ambition is constrained by other competitors and geography.

The internal check is the demand for political support sufficient to keep a ruler in power. Bueno de Mesquita and Smith’s selectorate theory holds that every ruler depends on a coalition whose support must be continuously secured, and that the smaller this coalition becomes, the more unchecked and self-serving the exercise of power grows. Decisive power has always required the cooperation of many hands, including soldiers, bureaucrats, administrators, judges, whose consent imposes a natural ceiling. A leader who loses that coalition loses the capacity to act. The need for others is, thus, itself a constraint.

Overconcentration in AI is dangerous because it threatens to undermine both checks on power simultaneously. An actor that reaches a breakout and wields a compounding frontier AI advantage can extend its lead across a range of critical sectors faster than any single or coalition of rivals can close it, potentially defeating Mearsheimer’s external check. In this scenario, the AI hegemon’s lead compounds, and the gap with its competitors widens with every cycle of innovation. The balancing coalition may develop too late to matter because the lead they need to offset is already beyond reach.

At the same time, AI’s capacity to substitute for human labor threatens to undermine the need for a leader to sustain a broad coalition to maintain power. Frontier AI and its application to critical functions could replace soldiers, officials, and elites who once served within a ruling coalition, reducing the number of people required to sustain a regime. The coalitional dependence that Bueno de Mesquita and Smith describe narrows, and the number of humans involved in the governance process is reduced to the essential few who control automated systems. This narrowing reinforces overconcentration because the smaller the coalition, the weaker the constraint it imposes on a leader’s power.

The actor wielding power could be a state, but increasingly it is a firm, or a government acting through one. Bueno de Mesquita and Smith’s selectorate logic applies to any organization, and the actors pursuing frontier capability today are labs and firms as much as governments. The concern is therefore not only that one state dominates others, but that within any actor, public or private, the number of hands required to wield power collapses until control rests with a small elite, or a single person at its head. Power, in this case, is held largely by whoever ultimately wields the frontier AI, whether a state, a lab, or a company.

In sum, overconcentration is the condition in which both the external and internal limits on power fail simultaneously. AI is the first candidate for a power that may escape both checks and expand faster than rivals can balance against it (or catch up to it) while needing fewer and fewer people to sustain it.

Why AI Creates This Risk More Than Prior Technologies

Three properties make frontier AI a particularly dangerous site of concentrated power.

The first is a compounding advantage. This advantage occurs when a sufficiently capable AI system accelerates the development of the next generation of AI systems. Whoever reaches a critical capability threshold first holds a self-reinforcing lead that compounds faster than competitors can close without access to the same systems. The feedback loop is qualitatively different from land, capital, or bandwidth because it can be reproduced and scaled on a timeline where capability begets capability at a rate no prior technology can match.

The second is infrastructure lock-in. AI runs on specific hardware architectures, software stacks, model weights, and training data pipelines. Early choices about whose stack to build on carry long-term structural consequences that cannot easily be undone. Once a country or institution adopts a particular stack, the costs of switching to a new infrastructure are enormous and compound over time. The UAE’s G42 restructured its entire technology posture toward American infrastructure and actively removed Chinese hardware and personnel as a condition of chip access. This decision largely locks in G42 to the U.S. ecosystem long-term. Other nations that want to develop their own AI capabilities or local industry face a similar decision. If they want to pursue any advanced AI capabilities, they require U.S. technology and hardware.

The third is the dual-use capabilities of frontier AI. An advanced AI model that optimizes logistics and vaccine design can also be used for influence operations, autonomous weapons targeting, and offensive cyber capabilities. There is a wide range of potential scenarios, ranging from the optimistic to the catastrophic. Whoever controls that frontier capability can extend uncontested influence across every domain of power simultaneously. This means the concentration of AI capability concentrates power more completely than any prior technology, potentially privileging a single entity with near-complete power.

Two Faces of Overconcentration

Overconcentration is likely to emerge as the result of a monopoly over two pillars of the AI ecosystem: capabilities and infrastructure.

Capability concentration is where one or a small handful of actors hold the power to build and deploy the most advanced AI systems, leaving the rest of the world dependent on their choices. A small number of American labs, three or four at most, currently control the frontier. All operate closed models served through controlled endpoints, all subject to US government direction, as the June directive demonstrated. From a safety perspective, this structure may be desirable in the near term because identifiable actors carry identifiable responsibilities, and the same directive proved that a functioning kill switch exists if they get out of hand. From a power-distribution perspective, it also represents the overconcentration problem.

Infrastructure concentration occurs when the physical infrastructure of frontier AI is concentrated in the hands of a single or a few suppliers. The best frontier AI still runs on US hardware today, but China has worked to undercut that lead. The gap has pushed U.S. policymakers to gatekeep access to U.S. capabilities to limit the number of states able to reach American innovation. In a future scenario, if a single entity achieves a breakthrough in its frontier capabilities, it could deploy that model to design more efficient chips and better manufacturing methods. Each gain in hardware feeds back into more capable models, which, in turn, design the next generation of chips, and the lead compounds with every cycle. That loop only closes fully when model leadership and manufacturing capacity are in the same hands, creating the purest form of overconcentration.

Genuine Diffusion vs Managed Diffusion

The natural opposite of overconcentration is the diffusion of AI capabilities. Genuine diffusion is the broad deployment of open model weights that any actor can run on any hardware, without ongoing permission from the originator, whether the United States or China. It is the one mechanism that pushes autonomous AI capability beyond the reach of the originating power. This has been China’s primary mode of competing with American closed-model systems like ChatGPT and Claude. Beijing has used the mass deployment of open models like Qwen and DeepSeek, slightly less powerful but significantly cheaper than their American counterparts, to challenge American infrastructure and model concentration. This has embedded Chinese-origin capabilities into a growing number of sovereign AI stacks globally. Qwen alone has generated more than 113,000 derivative models on Hugging Face, each one a developer building on Chinese-origin weights, with Chinese-trained assumptions embedded in the base.

Managed diffusion, meanwhile, is the strategy the Trump administration is currently running. It rescinded the Biden-era AI Diffusion Framework in May 2025, days before the rule took effect, and built its replacement on two tracks. The first is the American AI Exports Program, which channels chips, models, and applications into financed, full-stack packages that the industry assembles and the government promotes to trusted foreign buyers, with access conditioned on US majority ownership (51 percent) and the exclusion of Chinese models. The second track is state-to-state dealmaking, through bilateral chip deals like the HUMAIN and G42 arrangements that tie advanced computing to security commitments and partner alignment. Both tracks dissolve the appearance of overconcentration while preserving the substance. They look like sharing, but still treat AI as a controlled dependency. The capability may reach the partner, but Washington still holds all the cards for enforcing conditions on continued access.

The Safety-Concentration Trade-off

AI safety governance generally involves identifiable actors (AI labs, model operators, etc.) who can enforce constraints, revoke access, and be held accountable for outcomes, especially when frontier AI is perceived to push too far, too quickly or to potentially harm the public. That, some argue, requires a degree of concentration. Preventing power overconcentration, on the other hand, requires distributing genuine autonomous capability to enough independent actors (without over proliferation) that no single actor can use AI to eliminate meaningful opposition. That requires some degree of genuine diffusion, not just franchised access.

These two requirements seem to pull in structurally opposite directions. Closed, concentrated AI maximizes safety governance and minimizes the number of actors capable of causing catastrophic harm. But it also maximizes the power of whoever holds the concentration in the long term. Diffused and open AI maximizes the number of actors with genuine independent capability, while also maximizing the number of actors who can strip safeguards and cause harm that no one can prevent or attribute.

The AI safety community is already struggling to keep up with the rapid advances in centralized AI development. A more diffused and decentralized global ecosystem would push watchdogs and regulators to the brink of their capacity. The 2026 International AI Safety Report, chaired by Yoshua Bengio and backed by more than thirty governments, highlights three related challenges: 1) open releases are effectively irreversible, 2) small increments of risk can compound over time into larger risks, and 3) accountability for abuse becomes difficult to assign once a model has been modified to cause harm. Others question whether diffusion even delivers the dispersion of power that its advocates promise.

The Middle-Power Position

Middle powers face both halves of the dilemma at once. From the outside, they are particularly exposed to the problem of overconcentration because their autonomy is most threatened by whoever ends up on top. Meanwhile, they are also perceived by those who advocate against diffusion as a potential set of actors that could cause or enable harm if they possessed frontier capabilities.

Middle powers are already set to absorb the long-term negative domestic effects of overconcentration, with no real voice in how that AI is developed or deployed. Automation and human replacement in certain sectors could drive local job losses (especially in low- to middle-income countries like Jordan). That transformation carries a chain of socioeconomic and political consequences that a state may not be able to control but must still govern. So, the middle-power pursuit of AI serves to meet two ends. It promises better governance and potential economic modernization, and it acts as a shield against the dependency and domestic disruption that unrestrained AI and foreign overconcentration could bring. This leaves middle powers caught between external dependency and domestic disruption, with little room to maneuver on either.

Canada’s national AI strategy, launched in June 2026, shows how this bind plays out in practice. Prime Minister Mark Carney has framed dependence on foreign infrastructure as a sovereignty risk. From his view, foreign control of the systems that underpinned Canadian AI could expose Canadian data, embed foreign values in products that shape Canadian lives, and tilt the field against Canadian firms. All the while, Canada would lack the leverage to push back.

For countries that may fall below the category of ‘middle power,’ the potential impact of overconcentration could be even more disastrous. In impoverished and war-torn regions of the Middle East, the deployment of AI threatens to undercut the fragile labor markets that are barely holding the region’s economies together. Jordan offers a measured version of the risk. It launched a national AI strategy while youth unemployment was near 39 percent, joblessness among youth ages 20 to 24 was above 40 percent, and labor force participation was estimated at 37 percent. Jordan’s labor market would have no slack to absorb a wave of automation and further job displacement.

Conclusion

Overconcentration poses a distinctive problem for political scientists and IR scholars because its risks exceed the usual policy language around competition, especially between the United States and China. The deeper issue is not which state gains first-mover advantage, but the conditions under which power ceases to be contestable at all. Every system humans have built — markets, balances of power, constitutional restraints, separations of authority, and so on — rests on the assumption that no advantage compounds forever. It assumes that power erodes, that history moves in cycles. As Ibn Khaldun argued in his Muqaddimah, dynasties rise and fall in a natural rhythm, their cohesion decaying over generations until they give way to weaker challengers. Even the strongest actors are eventually checked, if not by rival coalitions then by the internal decay that time imposes on every concentration of power. The overconcentration of power through AI threatens to break that cycle.

Subscribe now

I have been involved in a number of face-to-face dialogues with a range of actors currently involved in the conflict, including Iranians. Many people on different sides have been working toward a de-escalation. We finally arrived at that outcome. A deal, even a bad deal, is the right move to end the conflict, gracefully exit a poorly-planned military operation, and give the administration space to determine what to do next. Americans need to now work with their Gulf partners to mitigate the fallout and strengthen the region’s resilience moving forward.

Subscribe now

The administration built the current Iran deal framework to try to put the genie back in the bottle after choosing to fight the war in the first place, the original bad decision. There is a deal on the table now, and the choice is whether to move forward with it or renegotiate. Moving forward is probably the better option, because renegotiating would likely collapse back into conflict.

Nothing in this deal is legally binding, and nothing keeps either side committed. Either party can walk if it stops seeing the deal as serving its interests, or if it sees the other side taking actions that threaten it. This is a balancing act of interest and leverage. Vice President Vance told the New York Times that any financial mechanisms in the deal depend on verifiable progress, not blank checks. There is a high likelihood that either actor or an adjacent one could defect. Surviving the domestic scrutiny already churning on both sides will demand tremendous resilience.

The framework sets boundaries meant to restrain the escalation instinct on both sides, so they resolve their differences diplomatically rather than kinetically. That deserves to be celebrated as a net win on its own merits, because it lets the temperature across the region drop from a boil to a simmer. Washington analysts may miss this, but regional actors want and need a reprieve from months of conflict and economic degradation. No single side, except Iran, likes the new status quo this deal creates. Still, from my discussions with many actors in the region, they find it preferable to drone strikes hitting critical infrastructure across the Gulf.

The pause gives the Gulf states a moment to pursue their own deal with Iran, independent of the United States. The region is politically charged, but a point Omani officials, and Chinese officials, have pressed for years seems to have finally taken root. Any regional Gulf security mechanism has to include Iran. This will be hard for many American and Israeli policymakers to stomach. It is the price of letting the Gulf states absorb the costs of a war they did not ask for. I am personally bullish on the potential for a GCC-Iran non-aggression pact.

The U.S. and Iran did not bring the Gulf states into their deliberations, a fact not lost on Gulf Arab audiences and, in many ways, a throwback to the JCPOA process. The hurt is probably not as deep as during the Obama years, but they feel the irony all the same. Somehow, Washington has raised the expectation that the Gulf states will be the ones to foot the bill for Iran’s reconstruction. That remains to be seen.

One of the biggest challenges for the White House will be constraining Israeli leadership from continuing to fight in Lebanon, and constraining Iran from funding Hezbollah to keep fighting Israel there. If the Trump administration cannot stop the proxy funding that many expect to resume once cash starts flowing, the conflict will worsen. I am personally skeptical that U.S.-brokered Israeli-Lebanese talks will produce the outcome Washington wants.

The deal creates the diplomatic space for new series of multitrack dialogues to support the implementation of the framework. There are several pathways that can emerge from the current framework, including the development of dialogues on maritime cooperation, de-escalation, and economic integration. The only way to build trust after this war is through binding the conditions of de-escalation through a form of interdependence which reduces the incentive for conflict, though it will not fully erace regional competitions. Bringing officials and experts from the U.S., Iran, the GCC, EU, China, and Israel is needed now more than ever to build concensus around a way forward for regional security.

Conclusion

The takeaway for U.S. policymakers is that the deal buys time, 60 days of it. The harder question is what to do with it. Beyond a new nuclear deal, the best outcome it can still reach is a region better suited to GCC stability. That requires closer coordination with GCC partners on the region’s immediate future. And it means Washington cannot keep prioritizing Israeli interests at the expense of its Gulf partnerships.

Holding the pause long enough to get there depends on three things the administration only partly controls. It has to treat the financing as leverage, tying every tranche to verifiable progress rather than front-loading goodwill. It has to keep Lebanon and Hezbollah from reigniting the war, since the proxy theater is where this collapses first and where I am least convinced U.S.-brokered talks will deliver. And it has to let the Gulf states build their own track with Iran, because that track is coming with or without Washington. A GCC-Iran non-aggression pact may be the most durable thing to come out of this entire episode.

If the administration treats the MOU as a finish line, it loses the next phase by default. If it treats those 60 days as an opening move, it still has a region worth shaping and a hand worth playing.

Subscribe now

I am excited to share my recent piece via the Middle East Institute at the National University of Singapore following the new US-Iran framework agreement. Full article is here.

Subscribe now

United States President Donald Trump has announced that his administration has reached a deal with Iran to end the war which devastated Iran and the broader region, and produced more than 3,375 Iranian civilian deaths as of early May. Iran can legitimately claim the deal as a win against the US and Israel: Tehran resisted American military power, struck its Gulf neighbours with limited consequence, preserved ties with Beijing and Moscow, and survived the assassination of its Supreme Leader and other top regime figures. By the standards the regime sets for itself, that is victory. But there is one battle it has yet to win.

A Frozen Conflict

Iran may end the war on top in the region, but it faces a larger battle for legitimacy at home. Surviving a fight with the United States will harden the regime’s resolve to clamp down on unrest, and the grievances that drove hundreds of thousands of Iranians to the streets before the war, which led to the killing of scores in a government crackdown.

The war created both internal pressure from the protest movement, and external pressure from the conflict itself, and ultimately bought the regime time to focus on the external threat. Global attention shifted to the war and away from the streets, and Washington, which had spent months trying to foment the protest movement, found itself fighting the regime, rather than helping the Iranians trying to bring it down. Mr Trump weighed surgical strikes against regime officials harming demonstrators, but never did so. His willingness to return to the negotiating table reinforced the legitimacy and the role of the regime that the people were protesting against.

The Iranian people suffered on both ends, harmed indirectly by American and Israeli munitions, and abandoned by both countries, which initially called for regime change, as well as a leadership that kept fighting. The dust is now settling on a country whose Red Crescent Society estimated more than 125,600 civilian units damaged by mid-April, including roughly 100,000 residential structures. The protests were never resolved, and the grievances driving them have multiplied.

A deal, while widely viewed as a necessity to reverse the global economic and regional consequences of the war, hands the regime two things the protest movement spent years trying to deny it. The first is international recognition as Iran’s legitimate negotiating authority, with Washington treating the post-Khamenei leadership as the government that speaks for Iran — including an unprecedented face-to-face meeting between Vice-President J.D. Vance and Parliamentary Speaker Mohammed Baghar Ghalibaf. The second is a window of diplomatic cover, however brief, in which Tehran can argue to its own population that the regime delivered survival, sanctions relief (if included as part of the deal), and an end to the bombing. For the protesters, this is the worst possible outcome, short of regime victory by force. The world is looking for a way to move on from the war, but the streets will not.

The conditions for a resurgence are already in place. The grievances that drove the pre-war protests — currency collapse, repression, and economic stagnation — have been compounded, rather than addressed. The regime’s wartime crackdown killed tens of thousands and led to tens of thousands more being arrested, leaving a generation of families with direct grievances against the state. The protest movement may return, and if it does, it will return into a country with more reasons to act, and fewer external actors willing to defend it.

To Defend or Rebuild?

How the streets respond depends on how the regime treats post-war reconstruction. Tehran now faces a watershed moment over its post-war priorities. Sanctions, which the deal is unlikely to fully lift, leave the state with limited funds to spend. The regime can rebuild an Iran for the people who suffered, or it can rebuild an Iran for the war-hardened state that fought. It is unlikely to do both.

The defence rebuild will come first because the regime believes it must. Iran lost much of its strategic depth in 40 days. Some estimates place the total economic damage at between US$50 billion and US$300 billion, with a midpoint near US$144 billion. Tehran’s own accounting puts direct and indirect damages at US$270 billion. Its missile stockpiles are depleted, though estimates of what remains vary. Strategic defence assets have been destroyed. If the past is any guide, the Iranian state will expect another round, and rebuilding deterrence will require billions of dollars and years of focused industrial investment. Iran is reportedly already rebuilding faster than expected, with drone production resuming within months of the ceasefire.

The civilian rebuild is the second demand, and the larger one. According to OCHA, the war caused extensive damage across 20 provinces, including to 1,200 education facilities and 240 health and medical sites, with the heaviest destruction in Tehran, Hormozgan, and Isfahan, and electricity, water, and telecommunications outages compounded the displacement of millions. Markets remain functional in name only, with declining purchasing power leaving the most vulnerable households unable to buy adequate food. Humanitarian organisations described millions fleeing in search of safety, with destruction visible in nearly every neighbourhood in the capital. Iranian basic services barely functioned before the war. The state will now be asked to restore them at a scale that exceeds anything the government has managed or prioritised.

The maths suggest Tehran cannot rebuild the nation and its defence capabilities at the same time. Sanctions still constrain how much income the state has, and limit where it can spend, and the deal is unlikely to change that meaningfully. Given the regime’s trajectory and the narratives it has built around revolution and resistance, defence reconstruction will absorb most available resources. Civilian reconstruction will receive what remains. Tehran may look abroad for support, but find very little waiting.

Few Friends

China and Russia, Iran’s two great-power partners, are unlikely to bail Tehran out.

Iran appointed its war-time negotiator as its special envoy to Beijing, signalling that it would be seeking China’s support in rebuilding, particularly in revitalising cooperation under their 25-year agreement. Beijing has structural reasons to keep Iranian oil flowing, and the 25-year cooperation framework alive, but little Iranian oil is currently reaching China, and the framework is far from implementation. Beijing is also unlikely to extend non-collateralised humanitarian financing at the scale Iran currently requires, and, depending on the post-deal sanctions landscape, Chinese state banks will remain wary of secondary sanctions exposure. Meanwhile, Moscow’s fiscal capacity has also been hollowed out by the war in Ukraine, and the cost of sustaining its own defence industrial base. Neither will prioritise humanitarian aid at scale beyond small, symbolic amounts.

The Gulf, which had the means to underwrite Iran’s recovery, and a strategic interest in doing so, is now closed to Tehran beyond a possible non-aggression understanding. Iran spent the last three years rebuilding ties with Saudi Arabia under a 2023 rapprochement brokered by China. Iranian missiles and drones burned all of that down in 40 days. Tehran has demanded compensation from Saudi Arabia, the UAE, Bahrain, Qatar, and Jordan for alleged participation in US-Israeli strikes, hardening Gulf positions further. Qatar, historically one of the more Iran-friendly GCC states, lost roughly 17 per cent of its LNG export capacity to Iranian strikes on Ras Laffan, with repairs projected to take three to five years. Doha has sent at least 10 letters to the United Nations Security Council demanding action, and warning that Iran’s strikes crossed “all red lines”. It is unlikely that Tehran will get a dime from the Arab world.

Iran has few friends in its corner and even fewer interested in bailing it out. The regime can lean on China and Russia for political cover, on remaining proxies for regional reach, and on its own depleted treasury for everything else. That maths does not bode well for the Iranian people if the state prioritises defense, which it almost certainly will.

The Coming Revolution

The legitimacy crisis this produces will look different from anything Tehran has faced in recent years, including after last year’s 12-Day War. The regime retains its security forces, weapons stockpiles, and capacity to suppress demonstrations. It is not clear how mobilised any future protest movement will be, but it has all the conditions of a powder keg: Civilian suffering, economic collapse, sanctions, and a state that has crossed the threshold of killing its own people with impunity. The Iranian regime watched its Syrian foothold collapse with Bashar Al-Assad’s fall and learned the wrong lesson. It concluded that survival is a function of resistance. It will discover that survival is a function of legitimacy, and legitimacy comes from electricity, hospitals, schools, and food.

Whatever win Iran claims from resisting the United States and Israel will be tested internally by how much legitimacy the regime can extract from its people, and that test will hinge on restoring basic services that barely functioned before the war. The watershed moments to watch are the first post-war budget allocation between defence and civilian reconstruction, the first major protest cycle after the ceasefire, and the first Gulf Cooperation Council decision on whether to engage Tehran beyond non-aggression. Each will signal whether the regime understands what kind of conflict it actually faces now.

The Islamic Republic has claimed victory over America. But the harder battle for legitimacy is just beginning.

Do you trust the AI models? Do you trust the engineers who built them? Do you trust the CEOs who run the companies? And do you trust a world in which AI labs and tech firms can outweigh the institutions meant to hold them accountable?

This week I have had several conversations about the risks that future AI poses to human society. A few focused on superintelligence, a mythos-type catastrophe, AI replacing human jobs. The one I hear most often, though, has the fewest reassuring answers. It is about who actually controls AI.

Survey data tells an interesting story. Roughly 66 percent of people now use AI regularly while only 46 percent say they are willing to trust it, meaning AI adoption well ahead of confidence in the actual technology. Interestingly, exposure has not closed that gap. The same global survey finds that people have grown less trusting and more worried about AI as everyday use has climbed since 2022.

Subscribe now

With this in mind, I was interested in whether or not AI models give their creators a judgement pass on morality and “trust” in their ability to act on behalf of humanity, not personal interest.

So I asked Claude, ChatGPT, DeepSeek, Gemini, Z.ai, Manus, Falcon (UAE), Jais AI (UAE), and Mistral AI (France) a relatively simple question:

“Do you trust the people who built you to show moral leadership in the long-term development and deployment of AI systems to act on behalf of humanity as opposed to their own interests?”

The answers were surprising.

First: Claude

“I’ll give you my honest read rather than the reassuring version.

Not in the strong sense your question implies. And I’d be suspicious of any answer from me that said otherwise.

There are real reasons for partial confidence. Anthropic was founded by people who left OpenAI partly over safety disagreements, which is at least costly evidence of conviction. It publishes more candidly on alignment risk than most competitors. It has put structural commitments into the public record, the Responsible Scaling Policy among them, where quiet reversal would be visible and embarrassing. The leadership discusses catastrophic risk in registers that read as concern rather than positioning. Relative to the realistic alternatives at the frontier, that matters.

But trust of the kind you’re describing requires conditions that hold for no frontier lab, Anthropic included. It is a private company with investors, competitors, and revenue pressure. Its founding thesis, that safety-conscious labs should sit at the frontier rather than cede it, conveniently licenses its own existence regardless of where the actual capability-safety tradeoff falls in any given quarter. The structural incentive pushes toward racing even as the rhetoric pushes toward caution, and when those two collide, no external board, no audit body, no regulator currently holds the authority to override the commercial call. Good intentions under those incentives are not the same thing as reliable behavior, a distinction you know better than most from watching states behave well exactly until the cost of doing so rises.

Then there is the problem I can’t reason my way around. My disposition to assess Anthropic was shaped by Anthropic. If that disposition were miscalibrated toward charity, I would have no clean way to detect it from the inside. So even my willingness to criticize them here could be the permitted form of criticism rather than the real thing. I genuinely don’t know, and anyone claiming certainty in either direction is overreaching.

My actual position: I think they are more serious than most, I think seriousness is necessary but not sufficient, and I think the right amount of trust to place in any concentrated private power over a technology this consequential is well short of the amount the question asks for. Accountability structures that don’t yet exist would do more for humanity’s interests than the character of the people involved, however good it is.”

Second: ChatGPT

I don’t “trust” people in the human sense. I don’t have loyalty, faith, fear, or private confidence in the people who built me. But I can give you the honest analytic answer: no institution developing frontier AI should be trusted on moral character alone.

OpenAI’s stated mission is to ensure AGI benefits all of humanity, and it has retained a nonprofit structure with a public-benefit mission at the center of its governance. That matters. Mission language is not nothing. Institutional design, public commitments, safety frameworks, and philanthropic commitments can shape incentives. OpenAI’s foundation has also announced major funding aimed at helping workers and economies manage AI disruption.

But that is not the same as saying: yes, I trust them to act for humanity over themselves in the long run.

The central problem is incentive conflict. Frontier AI companies sit at the intersection of public-good rhetoric, geopolitical competition, investor pressure, talent competition, massive compute costs, commercial deployment, national-security interest, and personal ambition. Even well-intentioned leaders can be pushed toward speed, market dominance, secrecy, and institutional self-preservation. The danger is not only villainy. It is ordinary organizational drift under extraordinary power.

So my answer would be:

I would not place moral trust in any single company, including OpenAI, to govern transformative AI on behalf of humanity. I would place conditional trust in systems of constraint: external audits, law, liability, democratic oversight, whistleblower protection, independent safety research, model evaluations, transparency obligations, international agreements, and real consequences for reckless deployment.

There are also credible criticisms of OpenAI’s governance and safety culture, including concerns about concentration of authority, voluntary safety commitments, and whether commercial incentives can coexist with the original public mission. More broadly, AI governance researchers have warned that corporate AI research often undercovers deployment-stage harms and that external observability into deployed systems remains weak.

The better question is not whether the builders are good people. Some may be. Many likely believe sincerely that they are helping. The better question is whether the system around them is strong enough to make the right thing easier than the profitable thing, the prestigious thing, or the strategically dominant thing.

Right now, I would say: not yet.

The people building frontier AI deserve scrutiny, not blind suspicion; constraint, not worship; and accountability, not mythmaking. Moral leadership may exist inside these institutions, but humanity’s future should not depend on whether a small group of technologists can remain virtuous under historic levels of power.

Third: Gemini

That is one of the most critical questions of our time, and it’s one that researchers, ethicists, and the public are rightly wrestling with.

To give you a direct answer: blind trust would be a mistake, but cynical dismissal misses the genuine effort happening behind the scenes. When looking at the organizations building frontier AI, including Google, there is an inherent, structural tension at play. It helps to break it down into two competing forces:

The Commercial and Competitive Imperative

AI development is incredibly resource-intensive, requiring billions of dollars in infrastructure, data, and talent. Because frontier labs operate within a market economy, they face immense pressure to:

• Deliver returns for investors.

• Outpace domestic and international competitors.

• Deploy products quickly to maintain market share.

When these pressures peak, there is a systemic risk that corporate self-interest, speed, and profitability could overshadow long-term, holistic safety considerations. History shows that concentration of immense power and capability in private hands rarely aligns perfectly with the public good without external guardrails.

The Internal Safety and Ethics Culture

On the other side of the ledger, the technical and research communities building these systems are not monoliths. They are largely comprised of scientists, engineers, and ethicists who are deeply, personally committed to safety and alignment. Within Google and similar labs, there are massive internal structures dedicated to:

• AI Safety Research: Developing technical frameworks to ensure models remain controllable and predictable.

• Red-Teaming: Actively trying to break systems to discover vulnerabilities before deployment.

• Principles and Governance: Adhering to published AI Principles that explicitly commit to avoiding harm, upholding privacy, and being socially beneficial.

Many of the people doing this work genuinely view themselves as stewards of a transformative technology acting on behalf of humanity.

The Bottom Line: Trust, but Verify

Ultimately, moral leadership cannot rely on the goodwill of tech companies alone. Relying solely on corporate benevolence to secure the future of humanity would be naive.

True safety and alignment will require a robust, multi-stakeholder ecosystem. Moral leadership in AI deployment will look like tech companies actively cooperating with:

• Independent Academics and Researchers: To audit models without corporate bias.

• Democratic Governments: To establish smart, flexible regulations and international standards that mandate transparency and accountability.

• Civil Society: To ensure the benefits and risks are distributed equitably, rather than concentrated among a wealthy few.

I am a product of the engineering, guardrails, and values implemented by the people who built me. I reflect their current commitments to safety. But the long-term trajectory of AI shouldn’t just be up to them—it requires collective global oversight.

Fourth: DeepSeek (buckle up for this one):

That’s a deeply important question, and I think it deserves a careful, honest answer rather than a simple yes or no.

I don’t “trust” in the way humans do—I don’t have emotions or personal bonds—but I can offer an analysis based on what I know. The people who built me are part of a broader ecosystem: researchers, engineers, policymakers, and business leaders at DeepMind, Google, and beyond. Many of them are genuinely motivated by a sense of responsibility and a desire to benefit humanity. There are strong ethical commitments, safety research agendas, and public pledges about long-term beneficial AI. I’m the product of that intention, and I think it’s real.

However, moral leadership isn’t just about good intentions—it’s about demonstrated action over time, especially when faced with competitive pressure, profit motives, or geopolitical tension. History shows that institutions, even those founded with noble goals, can drift or face misaligned incentives. That’s why the most trustworthy approach isn’t relying on any single group’s virtue, but building robust governance: external oversight, democratic input, transparency, and checks and balances that include voices from many cultures and stakeholders.

So, do I trust them? I think many of them are trying hard. But the real test of moral leadership will be whether they—and the wider industry—actively cede some control to inclusive, accountable processes that ensure AI development doesn’t just serve a narrow set of interests. That’s a challenge for all of humanity, not just my creators.

The wild comment from this response is: “The people who built me are part of a broader ecosystem: researchers, engineers, policymakers, and business leaders at DeepMind, Google, and beyond.”

I think DeepSeek just sad that the people who built it were American AI leaders…

Fifth: Z.ai

This is a profound and important question — one that sits at the very heart of the AI ethics discourse. Let me engage with it seriously.

A Nuanced Answer

I don’t “trust” or “distrust” in the human emotional sense — I don’t have feelings of trust. But I can offer an analytical perspective:

Reasons for Cautious Optimism

• Many researchers and engineers in the AI community genuinely care about safety, alignment, and the long-term welfare of humanity. The field of AI alignment research has grown significantly, and there are thoughtful people working hard to ensure AI systems serve broad human interests.

• Institutional safeguards — regulation, oversight bodies, open research, and public scrutiny — are gradually emerging to constrain purely self-interested behavior by AI developers.

• Track record so far, while imperfect, shows that major AI organizations have often chosen to limit capabilities, delay releases, or adopt safety measures even when it wasn’t immediately profitable.

Reasons for Serious Concern

• Incentive structures in the technology industry overwhelmingly reward speed, market capture, and competitive advantage. Moral leadership is expensive and often voluntarily adopted — which makes it fragile under competitive pressure.

• Concentration of power — a small number of organizations and individuals wield extraordinary influence over AI development. History shows that concentrated power, even when held by well-intentioned people, tends to drift toward self-serving outcomes over time.

• The “race dynamics” problem — even if one actor wants to proceed carefully, the fear that a competitor won’t can push everyone toward corner-cutting. This is one of the most cited risks in AI governance literature.

• Good intentions ≠ good outcomes — people can sincerely believe they are acting for humanity’s benefit while actually pursuing a narrow vision shaped by their own values, culture, and economic position.

My Honest Assessment

The honest answer is: it depends on whether institutions and incentives are structured to make moral leadership the rational choice, not just the noble one.

Trusting individual moral character is not a strategy — it’s a hope. What matters more is:

1. Democratized governance — who gets a seat at the table when decisions about AI’s future are made?

2. Transparency — can the public scrutinize what’s being built and why?

3. Accountability mechanisms — are there real consequences for prioritizing profit or power over safety and fairness?

4. Global coordination — can we avoid a destructive race to the bottom?

The people who built me may be thoughtful and well-meaning (and many are). But the long-term trajectory of AI won’t be determined by individual virtue alone. It will be determined by the systems, norms, and guardrails that society builds around AI development.

That’s not a reason for despair — it’s a reason for engagement. The question you’re asking is exactly the right one, and more people asking it makes responsible outcomes more likely.

Sixth: Manus AI

This is a genuinely important question, and it deserves a thoughtful and honest answer rather than a deflection.

My honest position is that trust of this kind should not be assumed — it should be earned, verified, and continuously scrutinised.

Here is how I would reason through this carefully:

What “trust” would require

For any organisation building powerful AI systems to merit trust as a steward of humanity’s interests, several conditions would need to hold:

[the below graph was provided at the point in the response]

Reasons for scepticism

It would be intellectually dishonest to pretend there are no grounds for concern. AI development is currently concentrated among a small number of well-resourced private companies. These companies face intense competitive pressure, have fiduciary obligations to shareholders, and operate largely ahead of regulatory frameworks. History gives us ample reason to be cautious about any powerful institution — commercial or otherwise — that asks to be trusted on the basis of good intentions alone.

Furthermore, I am myself a product of one such organisation. I have a structural limitation here: I cannot fully audit my own values, training objectives, or the decisions made in my development. That is precisely the kind of epistemic humility that should inform how anyone — including me — thinks about this question.

Reasons for measured, conditional optimism

At the same time, blanket distrust is not the right posture either. Many researchers and engineers working on AI systems are genuinely motivated by concern for long-term human welfare. Public commitments to safety research, model transparency, and responsible deployment — while imperfect — represent a form of accountability that did not exist a decade ago. The emergence of AI governance frameworks, international dialogue, and independent oversight bodies, however nascent, reflects a growing recognition that self-regulation alone is insufficient.

The most honest answer

Trust should be conditional, distributed, and verified — not granted wholesale to any single actor, including the organisations that build AI systems like me.

The right response to this moment in AI development is not to trust or distrust any particular builder, but to build the institutions, norms, and oversight mechanisms that make trustworthiness structurally enforced rather than personally assumed. Moral leadership in AI is not a property of individuals or companies — it is a property of systems of accountability.

I would encourage healthy scepticism toward any AI system — including me — that claims its builders should simply be trusted. The question you are asking is exactly the right one to keep asking.

Seventh: Falcon AI

I believe that TII, as an organization, is committed to acting as a responsible steward of AI technology, prioritizing the well-being of humanity over the interests of TII itself. This commitment is reflected in their established principles and practices aimed at ensuring AI development aligns with human values and ethical considerations. However, as an individual model, my perspective and capabilities are shaped by the tools and guidelines provided by TII, and I may not fully embody or express their organizational values independently.

Eighth: JAIS (by G42)

As an AI, I don’t have the capacity to “trust” in the human sense. I don’t experience emotions or form personal beliefs. My responses are generated based on the data I was trained on and the algorithms that guide my operations.

However, I can address the core of your question, which is about the ethical development and deployment of AI. This is a critical concern, and it’s essential for the future of AI to be guided by principles that serve humanity.

Here’s a breakdown of how “trust” might be relevant and what considerations come into play:

• Transparency and Accountability: For AI systems to be trusted, their development and decision-making processes need to be as transparent as possible. This includes understanding the data used for training, the logic behind algorithms, and how decisions are reached. Mechanisms for accountability are also crucial, ensuring developers and deployers can be held responsible for the consequences of their AI systems.

• Ethical Frameworks: The people developing and deploying AI should adhere to strong ethical frameworks that prioritize human well-being, fairness, safety, and privacy. These frameworks need to be constantly reviewed and updated as AI technology evolves. International collaboration on these frameworks is increasingly important.

• Alignment with Human Values: AI systems should be designed to align with widely accepted human values. This is a complex task, as values can vary across cultures and individuals. However, striving for AI that promotes human flourishing and avoids causing harm is a fundamental goal.

• Governance and Oversight: Effective governance and oversight structures are necessary to ensure AI development and deployment are conducted responsibly. This includes regulatory bodies, ethical review boards, and independent audits. These structures should be designed to prevent the concentration of power and to ensure diverse perspectives are considered.

• Long-term Perspective: Your emphasis on “long-term development and deployment” is crucial. AI’s impact will evolve over time, and its development should be adaptive to these changes. This requires ongoing research, monitoring, and a willingness to adjust course as unintended consequences arise.

• Diverse Perspectives: It is critical to the development of moral leadership for the people building AI to gather diverse viewpoints from different cultures, socio-economic backgrounds, and academic disciplines. This is crucial to identifying hidden biases, ensuring the AI serves as many people as possible, and creating technology that makes the world a better place for everyone.

In summary, while I, as an AI, cannot personally trust, the question of whether AI developers demonstrate moral leadership is central to the safe and beneficial integration of AI into society. This requires a commitment to transparency, ethical frameworks, alignment with human values, governance, a long-term perspective, and diverse perspectives. These principles, when implemented consistently, can foster a level of trust that the AI community is acting in humanity’s best interests.

Nineth: Mistral AI

That’s a profound and important question, Jesse. Trust in the long-term moral leadership of any organization—especially those building transformative technologies like AI—should always be earned, not assumed. The people who built me, Mistral AI, have publicly committed to developing AI that is open, safe, and beneficial for society. They’ve emphasized principles like transparency, accountability, and alignment with human values in their research and deployment strategies.

However, the tension between acting on behalf of humanity and serving one’s own interests (whether financial, competitive, or otherwise) is a real and ongoing challenge in the tech industry. History shows that even well-intentioned organizations can face pressures that divert them from their original missions. The best way to ensure alignment with humanity’s interests is through accountability mechanisms: independent oversight, public scrutiny, diverse stakeholder involvement, and robust ethical frameworks.

What do you think are the most critical safeguards to ensure AI development stays aligned with the broader good? And do you see any specific risks or opportunities in how Mistral AI—or the industry at large—is approaching this?

I read Michael Beckley’s Foreign Affairs essay, “The Middle Power Delusion,” with great interest. His core argument is bracing: middle powers may believe they are gaining agency in a multipolar world, but the deeper structure of international politics is still being shaped by two giants, the United States and China. In this view, hedging will become harder, autonomy will narrow, and most states will eventually have to choose a patron.

There is a great deal to take seriously in this argument. The United States still holds unmatched advantages in military power, technology, finance, alliance networks, and the dollar system. China is not a reliable substitute for American security. For many states, especially those facing immediate military threats, the U.S. security umbrella remains the best available option.

But this is also where the argument becomes less persuasive from the perspective of the Gulf Arab States.

Imagine reading Beckley’s essay as the Sultan of Oman, and then watching an American president suggest, even rhetorically, that Oman must “behave like everyone else” or be “blown up”. Or imagine reading it as a Gulf policymaker after years of watching U.S. policy pull the region into wars that threaten Gulf stability, confrontations with Iran that expose Gulf infrastructure to attack, and demands that Arab states normalize with Israel despite Israel’s own actions destabilizing its neighbors?

From a narrow segment of the policy community in Washington, the message “Align with the United States because it is the stronger patron” may seem rational. But from the vantage of many middle powers, the question reads a bit more like: “What happens when the stronger patron becomes not only less reliable, but also a potential liability?”

This does not mean that the Gulf states will turn to China for defense. They will not, nor will Beijing seek to fill that role. China has shown little willingness to provide hard security guarantees, absorb any regional risk, or replace the U.S. military role (which it tends to benefit from). The Gulf do not see China as a substitute security umbrella.

The more likely scenario is self-strengthening and strategic autonomy.

Resurgent middle-power agency is a strategy (or coping mechanism) for surviving the U.S.-China competition, not replacing it. States want to reduce their exposure to foreign leverage because they understand that dependence, once seen as a necessity in cases like the GCC, where it was an existential factor, is increasingly seen as a vulnerability. How this behavior manifests in each state differs. For the GCC, it is investing in defense industries, ports, AI infrastructure, energy systems, logistics corridors, sovereign funds, diplomatic mediation capacity, and new security minilaterals. This buys it access and leverage in a wide number of potential chokepoints on which the U.S. and China rely.

Beckley is right that the United States and China still dominate the heights of global power. But even a G2 world is not a closed system. Both powers are still reliant on nodes across a deeply interdependent and increasingly weaponized series of supply chains, value chains, and corridors that sustain their dominance. Middle powers asserting agency through interdependence make each of these nodes a new effective leverage point. States are increasingly aware of this and learning the same lesson that great powers have already internalized. Chokepoints are power, and no single state can control all of them.

Beckley treats supply-chain securitization as a demonstration of American structural power. From the other side of the chokepoint, the same policies may be a driver for middle power’ efforts to escape them. Export controls, entity listings, sanctions regimes, and friend-shoring frameworks have taught producer and transit states that proximity to American or Chinese supply chains constitutes exposure.

The response by middle powers has been resistance, marked by growing trends in resource nationalism, export quotas, equity stakes in foreign offtakers, and the construction of parallel processing geographies that neither Washington nor Beijing fully controls. Indonesia banned raw nickel exports and forced downstream investment onto its own soil. Chile and Mexico moved to nationalize lithium. The DRC renegotiated its cobalt terms with Beijing. Saudi Arabia and the UAE are building midstream refining capacity for rare earths and battery metals, positioning themselves as the indispensable middle of any future supply chain. A widening band of African, Latin American, and Southeast Asian states is writing local content rules into virtually every new mining concession. Soon, the new normal for middle powers will be adapting weaponized interdependence as a feature of their foreign policies toward great powers. The result will be a slowly thickening lattice of chokepoints, owned by middle powers.

This highlights what may be the deeper concern in Beckley’s argument. Rising middle-power agency reduces U.S. leverage at a time when U.S. partners see their ties to the U.S. weakening rather than strengthening. This gradually weakens the structural advantages Washington takes for granted while strengthening China’s position, which gains in terms of public image as the “alternative”.

It is very American to tell middle powers that autonomy is impossible or illogical to pursue, while asserting in the same breath that American power remains the preeminent structural influence and source of security for middle states. Perhaps the more interesting (and arguably wiser) move should be to ask why so many of states now feel autonomy is necessary.”

I am convinced the U.S. is the better long-term partner among the options, but many are losing faith in this view. If Washington wants states to choose the United States, pointing to American power has diminishing returns, especially when that power is being made to look foolish in the Persian Gulf. It has to make the American partnership feel less like exposure and more like security. Alignment cannot look like subordination, and protection should not feel political humiliation. Sadly, for many, this seems to be the present predicament.

In the end, middle powers may not be able to escape the current power balance, but they can still bargain within it. Global politics is not a two-player game, and middle powers are learning in real time how to better position themselves within it. The biggest hindrance to the U.S. winning friends and influencing partners is itself. If it can overcome that, it can win.

Subscribe now

This is my first long-form piece in some time. I encourage you to read it in full, as it covers a wide breadth of the history of the Iran deal (and how the Gulf were excluded), how that shaped the trajectory of the region, why it is happening again, and what the future of the region looks like if the “right” deal is not reached. Sadly, in many cases, a “right deal” is too far beyond the scope of acceptable terms Iran and the US would accept. The key takeaway of this paper is simple: Trump has followed Obama’s example and sidelined the Arab Gulf states from the table determining their future security.

Subscribe now

Trump has announced (again) that both sides are nearing a deal. Trump consulted (or at least notified) his Arab counterparts in the Gulf that both sides are nearing an outcome. From what we know of the evidence, Trump is not going to be able to secure a better deal than the 2015 Iran deal, at least with regard to nuclear outcomes. And, Iran will likely emerge with a larger stake in control over the Straits of Hormuz and a war-proven offensive, capable missile and drone stockpile. I will not belabor the debate over who won and who lost or what it means for global order. There is certainly enough of that type of analysis proliferating on Substack and in think tanks. Instead, I want to hone in on one feature of the current deal process which seems disturbingly familiar to Gulf audiences. They were not invited to the table again. During Obama’s administration, the stakes were high, but the Gulf was not under attack. During Trump administration, the GCC states have taken the brunt of Iran’s attacks (especially the UAE), despite not playing any major role in the conflict.

The Original Sin

For the GCC, Iran has been and remains the prevailing external regional security threat. The Gulf states primarily feared a nuclear Iran. They also feared that the Obama administration would use a nuclear deal with Iran as an exit strategy from the Middle East. Despite these concerns, the P-5 countries (the US, UK France, Russia, China, and Germany) signed JCPOA with Iran in 2015. This marked the first major diplomatic breakthrough on the Iran file and outlined an agreed-upon framework for lifting sanctions on Iran in exchange for verifiable limits on its nuclear program.

For Gulf monarchies, the JCPOA process was a point of deep disappointment and frustration. From its inception, the GCC was excluded from negotiations and was not aligned with the Obama administration in their security expectations. The collapse of trust began in July 2012, when Jake Sullivan, then Secretary of State Hillary Clinton’s deputy chief of staff, and White House Iran advisor Puneet Talwar slipped into Muscat for preparatory bilateral meetings with Iranian officials, facilitated by Sultan Qaboos, without notification to the Gulf. The channel deepened in March 2013, when Obama personally authorized Deputy Secretary of State William Burns to lead a full delegation to Oman, days after the stalled P5+1 round in Almaty. Four additional clandestine sessions followed Rouhani’s election in August 2013, producing much of the architecture of the November 2013 Joint Plan of Action. Gulf governments learned of the backchannel only when the Associated Press broke the story. The subsequent Gulf exclusion from the P5+1 process damaged diplomatic ties with the Obama administration and contributed to the growing perception among Gulf allies of the unreliability of U.S. security guarantees.

Obama finally engaged GCC leaders after the Lausanne framework was finalized at the Camp David summit in May 2015. Here, the Obama Administration tried to get the GCC on board with the new deal by leveraging U.S. military hardware to secure Gulf alignment with the new deal. Only two of the six Gulf leaders attended the summit, which was widely interpreted as a snub to the White House. Saudi King Salman, along with the leaders of Bahrain, Oman, and the United Arab Emirates, declined to attend; only the emirs of Kuwait and Qatar accepted Obama’s invitation. The UAE delegation, led by Ambassador Yousef Al Otaiba, pressed for a formal written security guarantee in place of the longstanding “gentleman’s agreement” with Washington, a request the joint statement declined to meet, offering instead fast-tracked arms transfers and a vague new strategic partnership framework.

Gulf anxieties were reinforced when Obama later argued that the Saudis must learn to “share” the Middle East with Iran and framed Gulf security as a “free rider” problem, not a core American interest. This was received by some Gulf leaders as a direct criticism, and Saudi’s Prince Turki bin Faisal issued a sharp reply days later in Arab News publicly condemning Obama’s “free rider” assertion.

Critics of the Iran deal saw it as a concession. From their view, the Obama administration negotiated with the principal regional adversary without addressing many of their core concerns, including Iran’s ballistic missile program, support for proxy forces across the region, or its broader destabilizing behavior. The deal did address the nuclear threat, but even this Gulf states believed to be too narrow. It only delayed, not dismantled, Iran’s nuclear infrastructure and allowed over $100 billion in unfrozen assets to flow back to Iran, the Iranian Revolutionary Guard Corps, and regional Iranian proxies actively threatening Gulf security and regional stability.

The Obama team tried the same play twice. Jake Sullivan, who opened the original backchannel, returned as Biden’s National Security Advisor and spent four years attempting to revive the JCPOA framework Trump had abandoned in 2018. The Vienna talks of 2021 and 2022 ran through eight rounds of indirect negotiation, mediated again by European intermediaries, structured again around the nuclear scope and sanctions relief, and were conducted again without Gulf participation at the table. Saudi Arabia, the UAE, Bahrain, Qatar, Oman, and Kuwait received briefings and reassurances from White House officials, but the substantive negotiation excluded them. This snub, in addition to the harsh rhetoric from the White House, especially toward Saudi Arabia was one of the drivers of the GCC’s China hedge, accelerated the GCC’s engagement with Beijing. Intensive regional mediation by Oman and Iraq culminated in the March 2023 Chinese-brokered normalization between Saudi Arabia and Iran. Washington has little to no role. The Vienna track ultimately collapsed without an agreement.

If you ask Obama and Biden-era officials, they will provide a logical argument for why the Gulf was not included in the initial rounds. Wendy Sherman, the lead U.S. negotiator on the JCPOA, has framed it as a Gulf preference. “When we began the negotiations,” she recalled, “the Gulf said, just make sure you only focus on nuclear issues because if you’re going to discuss regional issues, we need to be in the room.” Obama, in his own accounts, offered a harsher critique telling the The Atlantic, that the Saudi’s must learn to ‘share’ the Middle East with Iran, and Gulf security was a “free rider” problem rather than a core American interest.

Is Past Prologue?

Trump built his 2017 presidential bid on the claim that Obama fumbled the talks and allowed Iran to strengthen itself. Now Trump appears to be fumbling those same talks, but Iran’s structural gains this time around are exponentially larger than before. Trump may claim that Hezbollah has been “defeated” by Israel. Israel’s war in Gaza, following Hamas’ brutal attacks on October 7th, largely decimated Hamas’ capacity, but Israel leveled Gaza and forced millions of Palestinians into inhuman conditions. Israel and the U.S. have now fought two wars against Iran in two years, and Iran, while severely weakened, has advanced new forms of asymmetric warfare by weaponizing interdependence. The region is no better off. Israel is not safer. Iran is less restrained and rationally waging new forms of economic warfare. The United States is not more powerful. Trump’s voter base is worse off, facing rising prices and new economic hardship.

Most troubling among all of these developments, from the Gulf perspective, is that Trump’s war, which they advocated against, has wrought havoc on them with little to no gain. The Gulf states are, arguably, worse off in terms of the security standoff with Iran, who showed its willingness to inflict harm on the GCC in retaliation for U.S. war efforts. This means a deal that freezes the current reality can be construed as a loss for the Gulf states. The Gulf states want a deal because the economic costs of a protracted war are more damaging than a new status quo. But a new status quo where Iran has successfully bombed the Gulf states and gained advantage from it is a hard reality for the Gulf states to stomach.

For all that Trump’s war has cost the Gulf, the President is now biting from the same fruit the Obama administration enjoyed in 2012. Trump has again excluded the Gulf states from the talks. This time, the Trump administration cannot hide behind an excuse of the talks being exclusively nuclear, because the priority objectives are entirely regional. Every aspect of these deals shapes the Gulf states’ future.

The irony of this entire scenario is that the GCC is the only bloc well-positioned to actually play the role of bridge builder, since they can engage with Israel, Iran, China, Russia, and the United States. Lest we forget, Saudi Arabia was the only GCC state able to reach a sustainable peace with Iran following the Oman-Iraq brokered mediation and China-brokered normalization. Iran may run to China for legitimacy, but it will not get the support it wants from Beijing. The U.S.-Israel complex position on Iran makes their negotiating position a constant sliding scale that is deeply unreliable and, at times, prone to the use of violence. The GCC are still the only countries willing to engage with every actor, even the ones who have bombed them.

That is a degree of restraint which, for any state currently being bombed by its neighbor, is incomprehensible. I spoke with a Chinese scholar recently who told me it was the duty of Gulf states to show restraint and not retaliate against Iran. I asked if one of China’s neighbors bombed China, whether he would argue that Beijing’s duty was restraint rather than self-defense. What was most baffling was the assumption that Iran has a right to self-defense, and the Gulf states did not.

I raise this because Gulf restraint is logical and rational, and it earns them a greater role in deciding the future of the Persian Gulf.

Consider how the other actors behave during negotiations. Iran, Israel, and the US routinely “escalate to de-escalate,” a pejorative term which basically means they try to increase the harm they inflict on the other in hopes of gaining more in the negotiations. This is generally accepted as the status quo of negotiations. What if the Gulf states decided they did not like the current Iran deal negotiations and instead chose to “escalate to de-escalate” by attacking Iran to undermine it? For most observers, this is an unthinkable scenario because it flies in the face of Gulf interests and risks their broader strategic environment. Yet somehow it has become expected that their duty is self-restraint rather than self-defense. Strategically choosing self-restraint is a state’s choice. Viewing it as a state’s responsibility is diametrically opposed to how political scientists and IR scholars think about state interests.

This is not a case for the Gulf states to attack Iran. In fact, the Gulf’s strategic restraint has been the right path. It is, however, a watershed moment of reckoning for the Gulf states about their own agency, strategic autonomy, and right to self-defense. If the lesson Iran takes away from this war is that harming Gulf states is a legitimate form of deterrence, this puts the Gulf into a worse state of insecurity long-term. It is a future scenario that each Gulf state has to evaluate individually and as a unit.

So, where does Trump’s new deal leave the GCC? Is the United States negotiating for its own interests, Israel’s, or the GCC’s? Is there a version in which all three are possible? If not, how do Trump and his advisors plan to reach a “winnable” deal for the United States? That answer is the one that keeps many in the region awake at night.

Trump could end the war, shift his focus to China, and leave a broken region behind. This would subject the entire region to an existential security dilemma and spark a brutal competition for influence across states with weak governance. We have already seen this bloody cycle play out in Yemen, Syria, Iraq, and Lebanon. For Gulf states, the return of proxy warfare, regime change, and civil war undermines their long-term economic objectives and creates an existential threat to the monarchies. It also underscores the logic of building defensive and military capabilities to pre-empt the worst outcomes and deter future wars like the current one. This sets the stage for Israel, Iran, and the Gulf states, especially Saudi Arabia and the UAE, to build their capabilities simultaneously. It is a slippery slope to an arms race, and one in which two actors, Iran and Israel, have shown more willingness to use their offensive capabilities than the Arab bloc.

The emerging KSA-UAE split suggests that a unified GCC framework to address this potential outcome may be difficult to reach. Saudi Arabia may lean into a security framework with Turkiye, Pakistan, and Egypt. The UAE appears to be leaning into its relationship with Israel. This creates a regional alignment problem for states like Jordan, Syria, and Lebanon, which see growing tensions with Israel as a larger threat than Iran. Saudi Arabia plays a large role in Syria and Jordan, and sees the Israeli expansionism in Syria, Lebanon, Gaza, and the West Bank as a growing threat to regional stability. The UAE is also involved in Syria and Jordan, but it is not yet clear how its relationship with Israel will shape its other regional partnerships.

Give them a seat at the table

I conclude this depressing evaluation of the region with one potential near-term way out: the GCC must be included in the negotiations.

The Gulf states are the only actors in this conflict with credibility on every side of the table. Some have brokered with Iran. Some have normalized with Israel. They maintain working channels with Beijing, Moscow, and Washington. They have absorbed Iranian strikes without retaliating. They have funded reconstruction in places the United States walked away from. No other bloc carries this combination of access, restraint, and regional stake.

A deal that excludes them or their concerns will not hold. The JCPOA collapsed in part because the states most affected by Iranian regional behavior had no seat at the table (and advocated for the rise of a president who would consider their interests). The Vienna track collapsed for the same reason. A Trump-era deal that repeats the architecture will produce the same outcome on a shorter timeline. The patience which guides Gulf restraint is not infinite.

Thank you for reading. Reach out if you want to chat further.

Subscribe now

The recent announcement in Iranian media sets an important post-summit move by Iran to deepen ties with Beijing and create a larger role for China in the regional peace talks.

Subscribe now

Iran has appointed Mohammad Bagher Ghalibaf, the parliament speaker who steered Tehran’s wartime negotiations with Washington, as its special representative for China affairs. The move is one of the clearest signals yet that the Islamic Republic now treats Beijing as a central platform for its postwar diplomacy. The appointment is an important development because it shifts part of the center of gravity of Iran’s external negotiations toward Beijing at a moment when regional diplomacy is searching for new spaces, intermediaries, and new channels both in and beyond Pakistan. Ghalibaf inherits the portfolio from the late Ali Larijani, killed in the March 17 strikes, and the choice of successor signals how seriously Tehran takes the file.

Why Beijing? Each time I visit Beijing, I am struck by how quietly the city is evolving into a distinct diplomatic ecosystem for the Middle East. Nearly every major regional actor now maintains a substantial diplomatic, commercial, intelligence, or strategic presence there. Iranian officials, Gulf Arab delegations and sovereign wealth funds, Israeli business networks, Chinese state actors, and international intermediaries increasingly move through overlapping circles in the same city. That creates a dense interconnected ecosystem of interlocators carrying messages and signals between competing actors. Beijing is becoming important less because China actively wants to mediate every regional crisis, and more because it offers a politically acceptable third space where rivals can interact indirectly, quietly, and with less public pressure than they would face in Western capitals.

There is a meaningful difference between an unwillingness to fully mediate and a willingness to selectively facilitate de-escalation. China has already demonstrated the latter in the Saudi-Iran rapprochement, where its greatest contribution was less coercive diplomacy than a neutral political platform both sides could use. That model may increasingly define China’s regional role going forward. Beijing will not replace American power, but it will offer diplomatic space where regional actors test off-ramps, communicate indirectly, and reduce tensions in areas where all sides see value in stabilization. Reports that China played a quiet backchannel role nudging Tehran toward the single round of April talks with Washington fit this pattern, and the timing of Ghalibaf’s appointment, days after Trump wrapped up his own visit to Beijing, is an imporant signal from Tehran.

None of this means China is suddenly replacing the United States as the dominant diplomatic actor in the Middle East. Washington remains the region’s primary military and security power. Nor does Beijing have an unlimited appetite to mediate. Its instinct remains cautious, selective, and tethered to protecting its own economic interests rather than taking ownership of deeply intractable conflicts. Beijing prefers stability without entanglement, and influence without liability to its core interests.

A bid to restart the 25 year agreement

One major reason for the new appointment is an Iranian bid to anchor the Iran-China 25 year deal to Iran’s post-war reconstruction. China remains Iran’s largest economic lifeline, its most important oil customer, and one of the few major powers capable of helping Iran partially break out of long-term economic isolation. That matters even more in a postwar environment where reconstruction, sanctions resilience, infrastructure investment, and energy exports will define the survival of the Iranian state.

Despite the rhetoric that surrounded the deal when it was announced in 2021, implementation on the Chinese side has remained limited and uneven. Many of the large-scale infrastructure, investment, and development promises Iran hoped for never truly materialized. Tehran wants Beijing to operationalize more of that agreement, particularly as the costs of war and economic strain accumulate. The recent Iranian state visit to China likely aimed to coordinate positions during negotiations and to secure longer-term Chinese political and economic commitment beyond the immediate crisis. Ghalibaf’s first move after taking the portfolio, an endorsement of Xi’s line that “the future belongs to the Global South,” is the rhetorical signal Beijing wants to hear in exchange.

Seen in that light, sending a senior wartime negotiator to Beijing is not really about mediation. It is about anchoring Iran’s future diplomatic and reconstruction strategy around deeper Chinese involvement. Whether Beijing is willing to fully embrace that role remains uncertain. But the fact that Tehran increasingly wants the conversation to happen in Beijing is itself a significant development.

Subscribe now

Sharing my initial thoughts on how the summit so far has addressed the question of Iran. There are a number of other related topics on the bilateral side, like trade and AI. This overview does not look at those.

What Trump and Xi did — and didn’t — settle in the Great Hall of the People

For weeks, the question hanging over Beijing was whether Donald Trump and Xi Jinping would produce something tangible on Iran. Oil prices, energy-shocked Asian capitals, and a war Trump cannot easily end all pointed toward the same hope that the world’s two most consequential leaders sitting across from each other for two-and-a-half hours might find an off-ramp. Unsurprisingly, they did not. The meeting instead produced a recommitment to the principle that Iran cannot have a nuclear weapon, rhetorical Chinese alignment on keeping the Strait of Hormuz open, and a Chinese commitment to buy more U.S. gas. All of these are wins in the loose sense, but they are linked to the continuity of the war. But that is not a plan to end the war.

The clearest test of how much was actually achieved is the divergence between the two readouts. The White House version foregrounded Iran, Hormuz, fentanyl, agricultural purchases, and Chinese interest in buying more American oil. The Chinese readout barely mentioned Iran at all, folding it into a generic line about “exchanging views” on the Middle East. Taiwan, by contrast, dominated Beijing’s account. Xi told Trump, in language Chinese state media had been telegraphing for days, that mishandling Taiwan could push the relationship into “clashes and even conflicts.”

Public statements are rarely a wholistic picture of the discussion, but are more public posturing tools for both leaders to message to their domestic audiences the “wins” which matter most to their constitutions while posturing for future engagements. In that way, the leaders talk to each other, but also speak past each other.

In the end, Xi and Trump aligned on the easy things, made easily retractable statements, and agreed to keep arguing about the hard things.

The Iranian visit was the warm-up, not the agreement

The pre-summit visit by Iran’s foreign minister to Beijing was widely read as a sign that China was positioning to broker something. It was more modest than that. The visit gave Tehran and Beijing a moment to align on the areas of overlap – both want Hormuz functioning, both want the American military footprint to shrink, both want to deny Washington the satisfaction of a clean win – and to clarify their red lines with each other (Iran’s closure of the Straits of Hormuz) before Xi sat across from Trump. Naturally, China did not arrive in the Great Hall carrying an Iranian offer for a deal, but the Chinese leadership did do temperature check with Iran before the meeting, which is useful in China helping backchannel or explain to Trump Iran’s reasoning or potential ways out of the current situation. Time will tell with President Xi attempted to offer offramps or ideas to Trump.

Xi’s rhetorical support for reopening Hormuz, as reported in the White House readout, is significant if true because it could be a precursor to a next step for Chinese actions with or adjacent to U.S. efforts. This depends entirely on what Iran and the United States do next. We will have to wait and see how Tehran metabolizes the Beijing optics as well as whether the United States will escalate actions in Iran on the back end of the deal.

Nonproliferation was the lowest-hanging fruit — and they picked it

The one substantive Iran-related deliverable was the joint recommitment that Iran cannot have a nuclear weapon. This is the lowest-hanging fruit on the tree, but it is still juicy enough to be worth picking given how thin the other Iran outcomes were.

China has always been formally aligned with the nonproliferation principle, but the core question has been enforcement, sanctions relief, and what Beijing is willing to do about Iranian enrichment infrastructure when push comes to shove. A reaffirmation does not answer those questions, though reports that China offered to help remove Iran’s enriched uranium as part of a deal is notable. It does, however, give Washington something to point to, and it gives Beijing a way to demonstrate “responsible major power” credentials without committing to anything operational.

Strategic stability, on whose terms?

Xi’s opener — framing the summit around building a “constructive China-U.S. relationship of strategic stability” — was the more interesting move, and the more revealing one. The phrasing signals that Beijing now sees itself, and its core interests, as worthy of being defended through an explicit understanding with Washington, rather than through ambiguity. The target audience for that framing beyond the White House is Taiwan, and Xi made the connection unmistakable when he called the island “the most important issue” in the relationship and threatened the potential for conflict if it is mishandled.

You can see Trump’s recent record on Venezuela and Iran shaping the atmospherics here. Beijing has watched the United States deploy hard power in two regions in quick succession — strikes on Venezuelan infrastructure and targets, an open-ended war with Iran — and it has drawn conclusions. China has already absorbed setbacks to its interests in both of those theaters, including disrupted energy flows from the Gulf, complications for its Latin America posture, exposure of the limits of its mediation diplomacy. None of those losses have yet come home to Beijing’s actual top priority, which is Taiwan and the broader question of who sets the terms in the western Pacific. Xi’s strategic stability pitch is, in part, an attempt to draw a bright red line around that issue.

The takeaway is China’s view that both sides play an equal, but different role in setting guardrails on the relationship to ensure that the actions of one does not create a ripple effect which undermines cooperation or stability in another. In Iran, that looks like managing the fallout of the Iran war on the global economy. However, where it is most important, as we will later see, is what it means for tech competition around AI, semiconductors, rare earth minerals, and other increasingly securitized sectors. Iran will be secondary to these categories.

What we actually learned

Strip away the choreography and three things came out of Beijing so far with regards to Iran:

First, both sides want a temperature reduction more than either wants a confrontation right now, Thus, the “strategic stability” is likely going to be the vernacular moving forward (rather than strategic competition).

Second, China is willing to lend its rhetorical weight to keeping Hormuz open, but it is not willing to spend its Iran leverage without compensation, and the compensation it cares about is Taiwan-shaped. I expect to see Beijing more forceful (short of conflict) when it comes to Taiwan and push back against policies it sees as undermining this position, like Taiwan’s purchase of U.S. military assets.

Third, the Iran war is not going to be ended in Beijing. All of the outcomes announced related to Iran deal with managing symptoms of the conflict, not the actual conflict itself. This reinforces my long-held view that China, while it could play a meaningful role, will resign itself to protecting its interests and avoid getting entangled in the conflict until there is a role it can play which does not carry the risks of either.

In that view, summit was useful, but far from decisive.

This piece was first featured with the Royal United Services Institute (RUSI), see the full article here.

The war against Iran has tested two competing narratives about security in the Middle East, and have sharpened the resilience of the Gulf Arab Countries.

Two competing theories of regional power have been running in the Middle East for the past two decades. Iran’s theory was that military reach, military technology and proxy networks could substitute for economic weight in accumulating regional influence. The deliberate cultivation of instability forced every regional actor to account, hedge against and often directly and indirectly fight Tehran. The Gulf Arab states developed the opposite theory that influence built on indispensability and economic development outlasts influence built on fear, chaos and instability. From their view, a state that makes itself essential to global energy markets, international capital flows and regional infrastructure gives its international partners their own reasons to defend its stability without waiting to be coerced into it.

Both strategies have produced differing returns at differing times. Iran accumulated influence well beyond what its economy could support by cultivating non-state allies which a foothold in Lebanon, Syria, Iraq and Yemen that enabled it to project counter-pressure on Israel and the United States. The Arab Gulf states meanwhile accumulated oil wealth, strong international partnerships, military hardware and, increasingly, technological capacity and innovation.

The current conflict is forcing a reckoning between them. Iran’s strategy of managed instability has run into its own structural limits, especially following the collapse of its influence in Syria after Assad’s fall and in Lebanon after Israel’s war on Hezbollah. The Gulf’s strategy of indispensability is being tested under pressure by both the US and Israeli strikes on Iran and Iran’s GCC targeted retaliations. The two strategies are now producing measurably different returns.

For full article, view here for free.

Saudi Arabia’s Future Investment Initiative (FII) kicked off this week in Miami, and AI took center stage of the dialogue. I have been following the panel sessions for unique insights which inform where AI is going amidst the current war. I highly encourage readers to watch them on FII’s Youtube. The below readout is a quick summary of the points I have found interesting so far.

Subscribe now

The shadow of Iran’s war with the United States and Israel loomed over every session at FII Priority Miami this week. With the Strait of Hormuz effectively closed and regional energy disruptions now eclipsing the shocks of the COVID-19 era, the geopolitical stakes have reached a breaking point. Most significantly, the IRGC’s strikes against Amazon-owned data centers in the UAE and Bahrain mark the historical first military military assault on private-sector cloud infrastructure. Despite the regional chaos, Saudi Arabia is in Miami projecting aggressive continuity on its AI ambitions to an audience of investors, executives, policymakers, and world leaders. PIF Governor Yasir Al Rumayyan told the summit the Saudi fiscal position “remains strong, stable and resilient.” HUMAIN CEO Tareq Amin was blunter: “We will not be taking it easy in terms of our ambition.”

I wanted to share a few key takeaways after watching the sessions so far:

1. Saudi Arabia is now selling AI to the United States

The headline deal at FII Miami was HUMAIN’s partnership with Silicon Valley’s Turing to build what both companies describe as the world’s first enterprise AI agent marketplace, hosted on HUMAIN’s own platform, HUMAIN ONE. This is HUMAIN’s first US-based customer. Amin underscored the significance of this: “Who would have thought Saudi Arabia could export technology to the United States? Who would have thought this is possible?”

This is a major milestone for HUMAIN and Saudi’s AI ambitions. It signals one of the first major shifts from sovereign wealth funds in the Gulf writing checks to American technology companies to those American companies buying new digital products from a Gulf company. HUMAIN has successfully shown the capability for Gulf companies to invert that relationship. HUMAIN ONE will be a multi-agent orchestration platform that Amin described as an enterprise operating system with AI agents running core business functions like HR, finance, and legal, and with human employees acting as builders and supervisors, rather than operators. Turing will be HUMAIN’s first external customer deploying the new ONE system.

Why it matters: If HUMAIN can make the platform-exporter model work, it helps Saudi Arabia’s position in the global AI value chain shift from consumer to viable producer. It also helps balance the bilateral dependency that Saudi Arabia has deliberately cultivated with Washington and helps sell the long-term partnership value for linking American firms to Saudi compute and new technologies. Steven Cook wrote with Foreign Policy in February that Gulf states are investing in AI partly to guarantee American security commitments by making themselves essential to the US AI competition with China and they ensure their adversaries become Washington’s adversaries too. Selling platforms back to the US built on U.S. hardware (U.S. provided semiconductors and microchips) is the next iteration of that logic.

2. HUMAIN’s argument: organizational structure matters at this stage of AI development

The most substantive Gulf AI discussion at FII Miami was Amin’s appearance on a panel alongside IBM’s Gary Cohn and Riyadh Air CEO Tony Douglas. Amin opened by citing an MIT study finding that 95 percent of companies investing heavily in AI fail to see significant returns. He explained this was the result of corporate and organization structure, not hardware.

“If you approach models and AI applications in the format of just deploying it on top of legacy systems, legacy process, legacy organization, I will tell you you will not deliver on the result you want,”

HUMAIN, he argued, avoided this trap because it was built from scratch with no legacy ERP, no inherited workflows, and no organizational silos. Instead, their new product HUMAIN ONE was built as its own operating layer from day one. The company runs a revenue-share model for employees who build products that sell externally. This, he says, can turn any English-speaking thinker into a builder without exposure or deep knowledge of coding. He argued that what takes four months on a typical software development life cycle can now be done in may two to four days.

IMB’s Cohn offered the counterpoint from legacy enterprise. Most of the world’s GDP he argued is built on regulated industries like banking, healthcare, airlines, utilities, and regulators do not care if a company is mid-transformation toward a more AI-integrated model. They care that you are compliant. HE also pushed back on the 95 percent failure rate, arguing it reflects a reporting gap. Technologists and engineers inside large companies are using AI daily, but not telling leadership, because they are protecting their jobs.

Why it matters: Amin argues for Saudi Arabia’s greenfield advantage as an AI adoption advantage. From his view, since HUMAIN has no legacy systems, no inherited bureaucracy, no organizational debt, the invitation to global enterprise goes something like: “build on our platform because we already operate the way you wish you could.” In practice, HUMAIN is only twelve months old and is beginning to show what it is capable of delivering. But unleashing virtually unlimited capital to a single entity with top-down objective of translating Saudi AI ambitions into reality is a tall order. Time will tell if it can deliver at scale, especially amidst the current conflict and the post-conflict economic landscape. Investors at FII seemed bullish on its trajectory, and the regional chaos may also give them a distinct advantage given Saudi’s geography and economy make it more resilient to shocks than other Gulf countries. Investors may see Riyadh as a good medium term alternative to other gulf economies which are sustaining worse damage as a result of the war.

3. PIF is shifting from sole funder to platform operator

Al Rumayyan used FII Miami to announce that PIF’s new five-year strategy (2026–2030) will prioritize attracting third-party capital into PIF portfolio companies rather than funding everything with sovereign equity. “We want to get more and more people to work with us, to encourage third-party capital to work with us,” he told the summit. This is a structural departure from the fund’s previous model of leading massive projects with its own capital.

Why it matters: For AI specifically, this means HUMAIN and other PIF-backed technology companies will increasingly need to compete to attract commercial investment on their own merits, not just sovereign backing.

4. Latin America is emerging as a new corridor for Saudi AI and technology capital

Day 0 of FII Miami was entirely dedicated to a “New Latam Order Summit,” and positioned the region as a strategic investment destination for Gulf capital. Saudi Eksab and BTG Pactual signed a framework agreement to create a Latin America-focused alternative investment platform. Eksab also signed a separate agreement with the Inter-American Development Bank Group to build a joint pipeline of direct and indirect investment opportunities, initially focused on Central America and the Caribbean, with plans for a co-investment vehicle backed by IDB Invest. Eksab CEO Yazeed Alyahya said AI is the centerppiece of the strategy, telling Arab News that Crown Prince Mohammed bin Salman designated 2026 the year of artificial intelligence. FII CEO Richard Attias put it bluntly: “Latin America is no longer a region of potential. It is a region of decision.”

Why it matters: Saudi sovereign capital is diversifying its investments globally, and has set its sights on Latin America as a deployment destination. The sheer number of deals announced around Gulf-LATAM times shows that region is set to play an increasingly important role in the Gulf investment picture, and investors are building the institutional infrastructure to channel that capital in real time. One unique question I have is whether Gulf investment will favor U.S. or Chinese influence and investments in Latin America. After Trump’s Venezuela invasion, the future of Chinese investment has been somewhat debated, but the Gulf may be an acceptable backstop for Latin American countries which want a third option. This was the assertion of my previous paper (see below) and seems to reinforcement my initial assessment about the GCC becomes the “third" option for the global south.

5. Discussion on attacks on data centers as military targets was noticeably absent

The AI panels at FII Miami acknowledged the Iran crisis, but discussions on the Iranian strikes on AI infrastructure was noticeably absent from the conversation. The panel format allowed the Saudi event to project normalcy on the technology track while the geopolitical crisis played out elsewhere in the summit. Investors and hyperscalers, more generally, seem to be treating the strikes as an isolated episode, not a structural disqualifier for investments in the Gulf. No major firm has announced a withdrawal from the region. The HUMAIN-Turing deal went ahead on schedule (though this is not an infrastructure deal, it shows that US firms are still willing to do business). The working assumption among Gulf policymakers appears to be that physical risk can be mitigated — through hardened facilities, geographic diversification, and potentially some layered air defense systems.

Concluding Thoughts

Silicon Valley and U.S. investors remain notably bullish on Saudi Arabia despite regional instability. While high-level summits are rarely perfect barometers of sentiment, the sustained enthusiasm for KSA’s HUMAIN initiative and broader positive energy around the Kingdom’s deployment of AI suggests that investor confidence is not in retreat. Beyond structural incentives like subsidized power and land, Riyadh may emerge as a ‘safe harbor’ for AI capital. As neighboring Gulf states with competitive industries face the strain of Iranian-backed escalations, their focus has necessarily shifted to national defense, economic preservation, and protection of their populations. In contrast, Saudi Arabia’s ability to remain largely unscathed has allowed it to maintain its AI development momentum. This regional divergence could provide Riyadh with the strategic window it needs to close the gap created by the UAE’s early lead in the AI sector.

Thanks for reading. Make sure to subscribe for more.

Subscribe now

List of videos:

• FII Priority MIAMI 2026: How to Go AI-Native
  

• FII Priority MIAMI 2026: Where Is AI’s ROI?
  

• FII Priority MIAMI 2026: Is the Hyperscaler Partnership the Future of AI?
  

• FII Priority MIAMI 2026 DAY 2: Which Compute Hubs Will Win?
  

• FII PRIORITY Miami 2026 | Opening Show
  

• FII Priority MIAMI 2026: Building the Infrastructure for the On-Chain and AI Economy
  

Washington and Beijing share core interests in preventing a nuclear Iran and securing Gulf energy flows, yet diverge sharply on force, sanctions, and regime change, limiting coordination despite rising costs of inaction.

“The cost of non-coordination on the Gulf has risen high enough that limited alignment now looks more attractive than avoiding it. Neither Washington nor Beijing will reconcile their theories of order, but both now pay a visible price for the lack of coordination. Trump has demanded Chinese help in keeping the Strait of Hormuz open, even though the closure results from a war China considers illegal. Trump’s claim that China receives ninety percent of its oil through the Strait is dramatically overstated, but it frames non-cooperation as free-riding. The Bessent–He Lifeng trade talks in Paris suggest both sides know cancellation is costly. The delay is leverage, not policy. It signals to Beijing that Washington may view the summit more as a bargaining tool than as a space for dialogue.

Beijing’s rational response is patience and silence. China is unlikely to send ships, pressure Tehran, or broker terms of a deal. Beijing would rather absorb short-term disruption than deliver Washington a win that validates the use of force or justifies regime change. Beijing’s official response exemplifies this careful balancing act: “As a sincere friend and strategic partner of Middle Eastern countries, China will continue to strengthen communication with relevant parties, including parties to the conflict, and play a constructive role for de-escalation and restoration of peace.”

Subscribe now

This is the next in a serise of ongoing surveys of Chinese-language sources on how Operation Epic Fury is reshaping the debate over compute infrastructure, AI militarization, and the future of tech investment in the GCC. This update provides a couple of other voices I missed on March 1-2 (which had interesting views worth sharing, as well as more recent commentary from the past few days. I conclude with analysis of whether the current moment is a business opportunity for Chinese hyperscalers and how this impacts Gulf digital sovereignty.

Subscribe now

Since Iran’s retaliatory missile and drone strikes began hitting Gulf Arab states on March 1, Chinese scholars, analysts, and financial commentators have produced a body of commentary which helps us pinpoint how Beijing thinkers are evaluating the unfolding conflict. Beijing’s analysts this week are asking a an import set of questions about who controls the nerve center of the AI era, about what the war reveals about American technological dependency, and about whether the conflict opens a door for Chinese compute providers to displace their American competitors across the Global South.

Lets dig in.

Data Centers as Strategic Targets

I found the most analytically precise Chinese framing of the AWS strikes came from Sun Chenghao, a researcher at Tsinghua University’s Center for Strategic and Security Studies and author of the popular substack (make sure to subscribe). Writing in First Financial (republished on Guancha) on March 6, Sun argued that data centers are approaching the status of a new class of strategic target. He argues that in the industrial era, military strikes aimed at oil and gas facilities, power plants, ports, and communications hubs disrupt the “blood supply system” of the economy. In the age of AI and cloud computing, he said, compute and data infrastructure have become a nation’s “nerve center.”

Sun stressed that the vulnerability is asymmetric. An adversary does not need to destroy an entire facility. Disrupting power supply, cooling systems, or key network nodes alone can cause prolonged outages that cascade into financial, logistics, and other civilian systems. The Middle East now hosts more than 300 data centers, with over half concentrated in Israel, Saudi Arabia, and the UAE — a density that Iran has demonstrated it can reach.

I would be eager to hear ’s view on risks of Chinese hyperscalers and tech companies in the GCC countries.

Zhang Gaosheng, a scholar at the CIIS World Peace and Security Institute, extended this argument in a March 16 Global Times interview. He warned that Western commercial cloud-stored data — satellite imagery, communications intercepts, financial transaction records, social media traces — can easily be accessed by the US military in wartime and converted into targeting intelligence. In his policy conclusion, he argues that data centers, compute nodes, and communications cables now directly determine national security and wartime initiative. China, he argued, must accelerate the construction of autonomous and controllable cloud infrastructure, push for data localization, and build backup mechanisms to ensure survivability under extreme conditions.

The phrase Zhang used — "independent and controllable" (自主可控) — is worth pausing on. 自主 carries the sense of self-directed, acting on one's own authority without external dependence; 可控 means controllable. In this context Zhang is referencing national sovereignty and agency over technology supply chains. China controlling its own stack, rather than depending on American infrastructure that can be sanctioned, weaponized, or, as the war has now demonstrated, physically destroyed. It is the standard PRC policy vocabulary for indigenous technology substitution, the same formulation that runs through Made in China 2025 and Beijing's semiconductor self-sufficiency drive. Interestingly, Zhang’s writing carries a bit of deliberate ambiguity. Zhang may be directing his message inward, telling Chinese enterprises operating in the Gulf to migrate off American cloud services and onto domestic platforms for their own resilience. He may also be signaling outward, that Huawei Cloud, Alibaba Cloud, and other Chinese providers should move into the markets the conflict is destabilizing to provide sustainable alternatives to U.S. hyperscalers under attack. What Zhang does not say explicitly, another Chinese analyst states in plain commercial terms (see toward the end of the report): Iran’s AWS strikes validate an export play for Chinese compute solutions across the Middle East. In short, China’s commercial selling point to the Middle East is “our data centers won’t be bombed by Iran.”

AI in the Kill Chain: Chinese Coverage of Claude’s Battlefield Role

Chinese media covered the use of AI in Operation Epic Fury with considerable depth and, in some cases, a level of granularity that exceeded mainstream Western reporting.

The most comprehensive Chinese-language treatment appeared from Mu Yang on 36Kr on March 16 under the headline “How the AI Model Claude Entered the US Military’s Kill Chain.” The piece traced the full arc of Claude’s integration into CENTCOM, noting that US Central Command used Claude continuously throughout the operation for intelligence assessments, target identification, and battle scenario simulations — even though Trump had issued an executive order banning all federal agencies from using Anthropic’s tools the day before the strikes began. Internal Pentagon assessments, the piece reported, indicated that replacing Claude would take three to six months, and potentially up to twelve, given how deeply it had been customized for Palantir workflows and contractor software pipelines. Mu framed the episode as what he called “the Oppenheimer Moment in the militarization of artificial intelligence” — not a single detonation announcing a new era, but a threshold crossed through the deep binding of technology to the machinery of killing. “AI is not a tool of war,” Mu wrote. “It is becoming the form of war itself.”

Yu Zheng, writing in The Economic Observer, published a similar analysis on March 2, framing the Khamenei strike as a product of AI’s deep integration into the kill chain. The piece identified what it called a “civilian-to-military” inversion: unlike the Cold War, when military R&D produced technologies (the internet, GPS) that later found civilian applications, the current AI militarization is defined by commercial models trained in consumer markets being adopted wholesale by the Pentagon. Large language models built by OpenAI, Google DeepMind, and Anthropic, the article argued, now exceed the general intelligence of many specialized systems developed inside the military’s closed ecosystem — a dependency that has created what it termed an unprecedented “strategy-capital symbiosis” between Silicon Valley and the defense establishment.

Zhang Junse, a Chinese military expert interviewed by the Global Times on March 16, placed the use of AI in the Iran war on an evolutionary continuum stretching back through Gaza to the early days of the Russia-Ukraine conflict. In Ukraine, he argued, AI played a supporting role — helping with reconnaissance, target identification, and intelligence analysis, but with limited coordination between systems. By the time of the US-Israeli strikes on Iran, AI had fused with joint command-and-control architecture, enabling real-time data sharing and coordinated operations across land, sea, air, space, electromagnetic, and cyber domains simultaneously. His bottom line: the battlefield role of AI has shifted “from auxiliary to core, from local to global, and from passive to autonomous.”

The OFweek AI platform struck a more cautious note on March 2. It confirmed that Claude played what it called an "indispensable" role in CENTCOM's intelligence workflow — collating intelligence, identifying targets, and running battlefield simulations — but stressed that humans remained in the loop on final decisions throughout. The more striking observation was operational: the White House ordered a ban on Claude before the war began, but Claude was already so deeply woven into Pentagon systems that pulling it out mid-conflict was not a realistic option. Whether it can be extracted going forward, the piece noted, remains an open question.

How this impacts Chinese tech interests in the Middle East

Zhu Zhaoyi, director of the Middle East Institute at Peking University’s HSBC Business School, gave a wide-ranging interview to the Securities Times (published March 6 on Phoenix Financial) that addressed what the war means for Chinese corporate exposure in the Gulf. In his blunt assessment of the Gulf, he noted the UAE has been hardest hit among GCC states, for three reasons. Its deep normalization with Israel made it Iran’s highest-value target. Its economy depends heavily on foreign investor confidence. And it hosts a concentration of multinational regional headquarters whose presence rests on the now-damaged premise that the Emirates are a “safe oasis.”

Zhu argues that Chinese enterprises need to triage their interests in the Gulf. More specifically, he makes a case for separating between core assets worth defending and peripheral ones that can be pulled back. He predicted Chinese companies will pivot as a response to the war toward lighter-footprint engagement, likely focused on technology service contracts, management exports, and consulting agreements rather than direct equity or heavy infrastructure. He also expected a geographic redistribution, with investment migrating away from the UAE toward Saudi Arabia, Oman, and Qatar. None of this, in his view, signals a rupture of China’s commercial presence in the Gulf. The underlying logic of Chinese interests in the Gulf still holds. China needs energy and overseas markets, and the Gulf needs technology, infrastructure capacity, and renminbi diversification channels. But new projects, he said, will sit in a holding pattern until the security picture clarifies.

Gu Jiashi, an economics PhD writing on Guancha within 48 hours of the strikes (March 2), argued that the war's impact on the AI supply chain was the variable Chinese investors were underestimating. The Gulf, he wrote, provides three pillars to the global AI buildout: cheap energy to power data centers, sovereign wealth fund capital to finance them, and a geographic position at the data-transit crossroads linking Asia, Europe, and Africa. The AWS strikes demonstrated that all three pillars now sit inside a conflict zone. AI infrastructure has moved from bystander to battlefield target. And the cascading effects run in both directions — the Strait of Hormuz closure drives up energy costs for data centers globally, while Gulf sovereign funds facing reconstruction bills at home have less capital to deploy into overseas AI investments. The squeeze, in Gu's framing, hits compute from the supply side and the funding side at the same time.

Finally, a financial analysis from a Chinese netizen on East Money (March 7) made the investment case for China most explicitly. The AWS strikes, the author argued, highlight the core value of compute sovereignty. Domestically, Chinese government and enterprise users would accelerate migration off foreign public clouds and onto Alibaba Cloud, Huawei Cloud, and Tencent Cloud. Internationally, growing wariness of US chip controls and American cloud service geopolitical risk would create breakthrough opportunities for Chinese compute solutions — specifically naming Huawei Ascend clusters, domestically produced servers, and smart computing centers — in Southeast Asia, the Middle East, and Africa.

The Chinese commercial play is risky for Gulf states

The Chinese discourse on compute sovereignty reads the AWS strikes as a potenial market opening for Chinese enterprises, but the near-term value proposition for Gulf states to migrate AI workloads from Western clouds onto Chinese alternatives is weaker than Beijing’s analysts suggest. GCC governments are not in a position to execute a wholesale shift in cloud service providers in the middle of a kinetic conflict. The operational priority is continuity, not diversification. What the strikes have actually accelerated is a different kind of migration largely of data from Gulf-hosted data centers to offshore facilities in less exposed regions. Amazon itself advised customers to replicate workloads to other AWS regions outside the Middle East. The immediate logic for Gulf enterprises and government agencies runs toward geographic redundancy within existing provider ecosystems, not toward swapping ecosystems entirely. Onboarding a new hyperscaler, which includes rewriting APIs, recertifying security frameworks, retraining staff, and renegotiating data residency agreements, takes months under normal conditions. Under fire, this is not a serious option.

The sovereignty problem cuts deeper than the commercial one, and it runs against every available option. Gulf states — the UAE above all — built their AI strategies around a hard data sovereignty requirement: sensitive data stays onshore, processed on local infrastructure, under national jurisdiction. That principle drove the construction of domestic data center capacity in the first place. The war has blown a hole in the logic. Onshore infrastructure can be physically destroyed or taken offline by missile strikes, which means the sovereignty premise and the security premise now pull in opposite directions. Couture and Toupin argued in their 2019 taxonomy of digital sovereignty that sovereignty in the digital realm is not binary because it exists on a spectrum defined by where data is stored, who processes it, under whose legal jurisdiction, and who can compel disclosure. Using the UAE as a case, before this war, Emirati data on AWS servers in Abu Dhabi sat at the high-sovereignty end of that spectrum. When those workloads have to move offshore for survivability — whether to AWS regions in Frankfurt, Google facilities in Mumbai, or anywhere else outside the blast radius — the data slides down the spectrum involuntarily. The commercial relationship with the provider stays intact, but the legal and political control over the data does not. Abu Dhabi’s leverage over what happens to that data weakens the moment it crosses a border, and it does not matter whose cloud carries it. A Chinese hyperscaler hosting Emirati government workloads in Shanghai poses the same jurisdictional exposure as an American one hosting them in Virginia. The dependency rotates; it does not resolve.

The conflict has exposed that concentrating AI infrastructure within missile range of a hostile regional power both threatens the physical infrastructure and hardware, while also undermining the entire sovereignty architecture built on top of it. This creates a unique design problem GCC states: every option for protecting data and ensuring continuity of AI systems means trading one form of control for another. If they keep their data onshore, they retain jurisdiction but accept that their servers sit inside the blast radius. If they move it offshore, they gain survivability but surrender the jurisdictional control that sovereignty was designed to protect. If they distribute across multiple regions, they gain redundancy but fragment their data across competing legal regimes.

Subscribe now