• Nvidia is reportedly buying Hugging Face for $12.9 billion, extending its AI dominance from chips into the ecosystem built around them.

• Vanguard’s Altruist acquisition is a direct run at Schwab and Fidelity—and a rare deal from a company that almost never buys anything.

• AI infrastructure M&A is moving downstream into power distribution, where the bottlenecks are physical and the revenue is real.

• Anthropic considered a $7 billion chip acquisition before shifting toward a partnership.

• Two would-be megadeals died after boards and shareholders decided the strategic logic did not justify the price or complexity.

Subscribe now

Deals & Rumors

Nvidia → Hugging Face (reported $12.9 billion)

Nvidia reportedly agreed to acquire Hugging Face, the central repository for open-source AI models and datasets. The deal would give Nvidia control of an important developer platform just as major AI labs are trying to reduce their reliance on Nvidia’s chips. At roughly 86 times Hugging Face’s reported $150 million of annualized revenue, Nvidia is paying for ecosystem control—not current earnings. (Reuters)

Vanguard → Altruist (reported approximately $4–4.6 billion)

Vanguard agreed to acquire Altruist, the technology and custody platform for independent financial advisers. Vanguard gains modern adviser infrastructure and a stronger position against Schwab and Fidelity, while Altruist gets the backing to keep scaling. The strategic fit is clean; preserving Altruist’s independence will be the harder part. (Vanguard, Axios)

nVent Electric → Maverick Power ($1.75 billion, plus up to $550 million in earnouts)

nVent agreed to buy data-center power-equipment maker Maverick Power, adding switchgear, switchboards, and modular power-distribution systems. It is a bet that one of AI’s most durable bottlenecks will not be models or chips, but getting enough electricity to the building without setting anything on fire. (Reuters)

Tata Consultancy Services → Porsche’s MHP (€320 million / approximately $373 million)

TCS will acquire Porsche’s automotive consulting and IT subsidiary as part of a broader five-year, €1.25 billion partnership. Porsche simplifies its portfolio and secures a long-term technology partner; TCS buys specialized capabilities with a major customer already attached. The contracted revenue may be more valuable than the acquisition itself. (Reuters)

Anthropic → MatX (abandoned discussions at approximately $7 billion)

Anthropic reportedly considered buying AI-chip startup MatX before the talks shifted toward a partnership. The discussions show how urgently AI labs want alternatives to Nvidia—but also why buying a chip team does not automatically solve years-long development cycles and enormous manufacturing costs. (Reuters)

Stripe and Advent → PayPal (abandoned reported $53 billion pursuit)

Stripe and Advent reportedly ended their pursuit after PayPal’s board considered their $60.50-per-share offer inadequate and raised financing and regulatory concerns. PayPal’s shares subsequently traded above the bid. “Cheaper than its pandemic valuation” turns out not to be a complete acquisition thesis. (Reuters, Axios)

Deal of the Week

The Deal

Nvidia reportedly agreed to acquire Hugging Face for $12.9 billion. Hugging Face hosts open-source models, datasets, and tools used throughout the AI ecosystem and was valued at $4.5 billion in 2023.

The reported price represents roughly 86 times Hugging Face’s $150 million of annualized revenue.

Why It Matters

Nvidia already dominates the hardware used to build AI. Hugging Face would give it influence over how developers find, test, and deploy the models running on that hardware.

That makes this an ecosystem acquisition. Hugging Face sits between model creators, developers, and infrastructure providers, giving Nvidia valuable distribution and visibility across the market—even as Anthropic, OpenAI, and others develop competing chips.

It would also put the leading open-source AI platform inside the company with the strongest financial incentive to remain the industry’s default infrastructure provider. Regulators and developers will notice.

Sara’s Take

The strategy makes sense. The price requires imagination.

Hugging Face gives Nvidia a powerful developer community, strategic data, and a front-row seat to emerging models and workloads. Those advantages could reinforce its chip business in ways that will never appear neatly in Hugging Face’s standalone revenue.

But paying nearly 86 times revenue creates pressure to monetize a community that values openness and neutrality. If developers start viewing Hugging Face as an Nvidia-controlled distribution channel, the asset’s strategic value could erode precisely because Nvidia bought it.

Nvidia is not buying another AI company. It is buying the town square—and hoping everyone keeps showing up after the landlord changes.

Market Trends & Takeaways

• AI buyers want control points. Chips, model repositories, power distribution, and custody platforms all provide leverage over broader ecosystems.

• Distribution is driving premiums. Hugging Face and Altruist matter because developers and advisers already use them, not simply because their software is good.

• Vertical integration is moving in both directions. Nvidia is expanding from hardware into software while Anthropic considers moving from models into chips.

• Shareholders are willing to say no. PayPal, Solstice, and Element show that scale and strategic logic do not excuse weak economics.

From the Data Room

• Technology is carrying the U.S. deal market. From May through July, U.S. technology deals worth at least $100 million generated approximately $324 billion in value, up 161% year over year, while volume increased 32%. (EY)

• Asset-management M&A has reached a record $53.8 billion in 2026. Vanguard–Altruist and Victory Capital’s $7 billion First Eagle acquisition illustrate the two main strategies: buy technology and distribution, or buy more assets. (Financial Times)

• Nvidia has $18 billion committed to equity investments through fiscal 2027. Its reported Hugging Face deal is part of a much broader effort to invest across—and exert greater influence over—the AI ecosystem. (Reuters)

Hit reply with the deal I missed, the multiple that needs adult supervision, or the “strategic partnership” everyone knows began as an acquisition pitch.

Share

• Stripe is buying OpenRouter for a reported $7.5 billion—a roughly sixfold markup from its May valuation.

• Cybersecurity buyers are no longer just securing AI models. They’re buying protection for agents, prompts, tools, and whatever else the robots touch.

• Private equity is still willing to buy traditional software, but AI anxiety is giving it considerably more negotiating leverage.

• Smaller software deals are clustering around the same thesis: own the workflow, add AI, call it a platform.

• AI infrastructure is becoming so capital-intensive that the financing structures increasingly resemble project finance with a GPU problem.

Subscribe now

Deals & Rumors

Stripe → OpenRouter (reported $7.5 billion)

Stripe agreed to acquire OpenRouter, the gateway that routes AI requests across more than 400 models from over 80 providers. The strategic logic is bigger than adding another developer tool: Stripe already helps companies optimize how money moves, and OpenRouter lets it optimize how tokens are spent. If payments are the economic plumbing of the internet, Stripe is betting that model routing becomes the plumbing of the AI economy—and it would very much like to own both pipes. (Stripe, The Information)

Ridgeview Partners → Pinewood Technologies (£545 million / approximately $739 million)

Ridgeview reached a recommended agreement to take automotive software provider Pinewood private for 448 pence per share, a 43% premium to its undisturbed price. The fun part: Apax previously offered £575 million, then walked away as investors panicked about AI’s impact on software. Ridgeview is now buying essentially the same asset for less while promising to invest in—you guessed it—AI. Disruption for public shareholders; buying opportunity for private equity. (Reuters, Financial Times)

Fortinet → Virtue AI (undisclosed)

Fortinet acquired Virtue AI, which provides automated red-teaming, governance, validation, and runtime protection for AI systems. The attack surface has expanded from networks and endpoints to prompts, models, autonomous agents, and MCP tools, creating an entirely new category of things CISOs are expected to lose sleep over. Virtue fills that gap and gives Fortinet a more credible end-to-end “security for AI” story. (Fortinet)

Lattice → Pando (undisclosed)

Lattice acquired performance-management platform Pando and named its founder, Barbra Gago, chief marketing and strategy officer. Pando adds career progression, evidence-based calibration, and real-time performance signals to Lattice’s people platform. This is part product acquisition, part talent acquisition, and part attempt to answer the uncomfortable HR question of the moment: if AI is changing everyone’s job, how exactly are companies supposed to evaluate whether anyone is doing it well? (Lattice)

Brinqa → PlexTrac (undisclosed)

Exposure-management provider Brinqa acquired PlexTrac, adding penetration-testing workflows and the ability to verify that identified vulnerabilities were actually fixed. The combined company says it serves more than 3,000 customers across 57 countries, including over a quarter of the Fortune 500. Security teams do not need another dashboard telling them they have problems; Brinqa is buying the workflow that proves someone solved them. (Brinqa announcement, Covington)

osapiens → Nasdaq Metrio (undisclosed)

German compliance-software company osapiens acquired Nasdaq’s Metrio sustainability-reporting platform and customer base. Metrio brings carbon-accounting capabilities and a foothold in North America, while Nasdaq sheds a specialized software product that fits more naturally inside a dedicated compliance platform. It is a tidy example of two recurring deal theses: software vendors expanding through vertical tuck-ins and larger companies pruning products that never quite became core. (osapiens, ESG Today)

Deal of the Week

The Deal

Stripe agreed to acquire OpenRouter for a reported $7.5 billion. OpenRouter was valued at approximately $1.3 billion in May, meaning a deal at the reported price would deliver one of the faster, and richer, AI infrastructure exits of the current cycle.

OpenRouter sits between developers and model providers, directing requests based on price, speed, reliability, and task complexity. Stripe plans to combine that capability with products such as Token Billing, giving customers tools to manage both the revenue generated by AI applications and the compute costs required to run them.

Why It Matters

Stripe is quietly expanding its definition of “financial infrastructure.” It no longer wants to process only the dollars flowing through AI businesses; it wants to manage the underlying unit economics.

That makes OpenRouter a control-point acquisition. As businesses use more models, routing becomes valuable precisely because no single model wins every workload. Owning the layer that decides which model receives each request gives Stripe insight into usage, cost, and performance across the ecosystem—while remaining neutral enough to work with nearly everyone.

The valuation also shows where buyers think durable AI value may accrue. Models are expensive to build and prone to leapfrogging one another. The routing, billing, governance, and security layers can benefit regardless of which model is fashionable this quarter.

Sara’s Take

This is a big swing, but it is not random empire-building. Stripe understands metered usage, optimization, and complex infrastructure better than most companies—and OpenRouter turns tokens into something Stripe can manage like payments.

The risk is paying tomorrow’s control-point valuation today. A reported $7.5 billion for a company valued at $1.3 billion three months ago requires heroic growth, limited disintermediation, or preferably both. Model providers will keep improving their own routing, and large customers may build the capability internally.

Still, the thesis is: in an AI market full of competing models, do not bet on a single winner. Buy the tollbooth.

Market Trends & Takeaways

• The new AI stack is forming through M&A. Routing, security, billing, governance, and workflow orchestration are becoming distinct acquisition categories around the models themselves.

• Cybersecurity consolidation is moving up the stack. Buyers are shifting from protecting infrastructure to monitoring how autonomous agents behave once deployed.

• AI fear is helping PE find entry points. Public software valuations are getting punished for disruption risk even when the underlying businesses remain attractive—and sponsors are happy to arbitrate the panic.

• Workflow still matters. Pando, PlexTrac, Metrio, and Lupl all reinforce the same point: AI features are easier to monetize when attached to a workflow customers already use.

From the Data Room

• The recovery remains top-heavy. U.S. M&A value reached $1.2 trillion in the first five months of 2026—nearly twice the prior-year level—but deal volume declined 4%. Thirty-nine transactions worth at least $5 billion accounted for much of the increase. Translation: the market is back, provided your definition of “market” is “very large things.” (PwC)

• AI capex is creating a shadow deal market. Broadcom is reportedly seeking more than $60 billion—and potentially as much as $100 billion—of debt financing through a special-purpose vehicle to fund AI-chip infrastructure. These structures are not conventional M&A, but they are becoming strategically similar: enormous, bespoke capital commitments that lock together chipmakers, hyperscalers, AI labs, and private-capital providers. (Reuters)

• Cybersecurity’s AI land grab has real numbers behind it. Gartner projects the market for securing AI ecosystems and agents will grow from $2.8 billion in 2026 to $16.4 billion by 2030. Fortinet’s Virtue AI acquisition is unlikely to be the last time an incumbent buys rather than builds its way into the category. (Fortinet)

• Even the bankers are trading. JPMorgan hired Bank of America veteran David Fishman to lead North American technology M&A as part of a broader push around key tech clients. When banks start acquiring senior dealmakers, they are telling you where they expect the fee pool to be. (Reuters)

Hit reply with the deal I missed, the valuation you refuse to defend, or the AI startup your CEO suddenly thinks is “strategically essential.”

Share

• Nscale is buying Anyscale to add AI orchestration software to its neocloud stack.

• Cyera is buying Oasis Security for $1B as non-human identity becomes a serious AI security category.

• Saudi PIF’s $55B EA deal cleared EU foreign subsidy review and is expected to close in early August.

• Qualcomm completed its Modular acquisition, another reminder that chip companies are buying the software layer.

Subscribe now

Deals & Rumors

• Nscale → Anyscale (~$1.65B reported; terms not disclosed)
  AI cloud provider Nscale agreed to buy Anyscale, the software startup behind tools for managing distributed AI workloads. This is a neocloud buying up the software layer, because customers do not just need GPUs. They need a cleaner way to run complex AI jobs across lots of machines without turning infrastructure management into a second job. (Reuters)

• Cyera → Oasis Security ($1B)
  Data-security company Cyera agreed to acquire Oasis Security in a $1B cash-and-stock deal. Oasis focuses on managing non-human identities, including AI agents and automated software, which is becoming a bigger problem as companies give agents access to internal systems. The simple read: AI security is moving from “protect the data” to “control who, or what, can touch it.” (WSJ)

• Saudi PIF-led consortium → Electronic Arts ($55B; regulatory approval)
  Saudi Arabia’s PIF and its investor group received EU approval under foreign subsidy rules for the $55B acquisition of Electronic Arts. EA said it had received all regulatory approvals as of July 30 and expects to close the deal on August 4. Gaming remains a strategic category because it combines IP, communities, consumer attention, and global distribution. (Reuters)

• Thales → Exail (€3.9B enterprise value)
  Thales signed the tender offer agreement for its planned acquisition of drone maker Exail, moving forward a deal first announced earlier in July. This is defense tech consolidation with a clear demand signal: drones, autonomous maritime systems, and anti-submarine warfare are becoming higher-priority categories as governments increase defense spending. (Reuters)

• Qualcomm → Modular ($3.9B; completed)
  Qualcomm completed its all-stock acquisition of Modular, an AI software development platform. The deal strengthens Qualcomm’s effort to make AI models easier to deploy across different chips and environments, which matters because developer adoption is where chip strategies either scale or stall. (Axios)

• AT&T → EchoStar wireless licenses ($23B; completed)
  AT&T completed its $23B purchase of wireless licenses from EchoStar. This is spectrum consolidation, which is not flashy, but it matters. Wireless capacity is one of the core inputs for mobile broadband, connected devices, and future AI-enabled services that depend on fast, reliable connectivity. (Axios)

Deal of the Week

• The Deal: Nscale → Anyscale. Nscale is buying Anyscale to pair GPU capacity with software that helps teams run distributed AI workloads.

• Why It Matters: Neoclouds are all trying to prove they are more than GPU landlords. The first wave was about getting access to chips. The next wave is about making that compute easier to use, easier to scale, and harder to swap out.

• Sara’s Take: I like this deal because it gets at a boring but important truth about AI infra: customers don’t want to manage the mess. They want capacity, but they also want scheduling, orchestration, utilization, monitoring, and fewer “why is this workload failing?” Slack threads. Anyscale gives Nscale more to sell than raw GPU hours, which matters in a market where everyone’s pitch starts to sound the same. Compute is still the entry ticket. The software around it is where differentiation starts to show up.

Market Trends & Takeaways

• Neoclouds are moving up the stack. Compute providers are buying software because capacity alone is becoming harder to differentiate.

• AI security is narrowing into specific pain points. Non-human identity, agent permissions, data governance, and access control are becoming real budget categories.

• Regulatory approval is still a deal event. EA’s approval shows that foreign subsidy review can be a major milestone, not just a background process.

From the Data Room

• Big Tech’s AI capex debate is now the market. The FT reported that Google, Amazon, Microsoft, and Meta have spent roughly $1.1T on capital investments since the AI boom began in 2023, with combined 2026 capex expected around $745B. The question for investors is shifting from “who is spending?” to “who is converting spend into growth?” (Financial Times)

• Microsoft showed what the market wants to see. Reuters reported Microsoft added nearly $450B in market value after strong cloud and AI results, while keeping its spending plan steady at an expected $175B of capex for calendar 2026. The lesson is not that investors hate AI spending. They hate AI spending without proof. (Reuters)

• Corporate law may give public benefit companies more deal flexibility. A Delaware judge ruled that public benefit corporation boards do not have the same duty to maximize shareholder value in a sale as traditional corporations. That matters for AI companies with mission-driven structures, because future sale processes may involve more than simply taking the highest bid. (Reuters)

Hit reply with the AI infra layer you think gets acquired next.

Share

• Uber agreed to buy Delivery Hero for $14.8B, creating the largest food delivery group outside China.

• Hyundai is taking full control of Boston Dynamics by buying SoftBank’s remaining stake.

• Thomson Reuters is selling control of its print business to KKR so it can focus more clearly on digital and AI products.

• EA’s $55B Saudi-led buyout looks likely to clear EU review, which says a lot about the current regulatory climate.

Subscribe now

Deals & Rumors

• Uber → Delivery Hero ($14.8B)
  Uber agreed to buy Germany’s Delivery Hero in a deal that would create the largest food-delivery group outside China. The deal expands Uber’s delivery footprint across Europe, Asia, the Middle East, Latin America, and Africa, but the real test will be whether Uber can turn global scale into real margin improvement after years of delivery platforms promising operating leverage. (Reuters)

• Hyundai → Boston Dynamics remaining SoftBank stake (terms not disclosed; reported ~$335M)
  Hyundai Motor Group is buying SoftBank’s remaining roughly 10% stake in Boston Dynamics, making the robotics company a wholly owned subsidiary. This gives Hyundai more control over how Boston Dynamics gets deployed across factories, logistics, vehicles, and humanoid robotics, which matters as “physical AI” moves from R&D theater into industrial operations. (Reuters)

• Thomson Reuters → Global Print business / KKR JV ($500M for 51% stake)
  Thomson Reuters agreed to sell a 51% stake in its Global Print business to KKR for about $500M, while keeping editorial control and IP rights. This is portfolio cleanup with a clear strategic direction: move capital and attention away from declining print formats and toward legal, tax, corporate, and AI-powered professional tools. (Reuters)

• Saudi PIF-led consortium → Electronic Arts ($55B; EU approval expected)
  A Saudi PIF-led consortium, alongside Silver Lake and Affinity Partners, is reportedly set to win EU approval for its $55B acquisition of Electronic Arts. The deal shows how gaming has become strategic infrastructure for countries and investors trying to build consumer platforms, entertainment ecosystems, and soft power in one move. (Reuters)

• Perpetuals.com ↔ Alt5 Sigma Canada / AI Financial (talks ended)
  Perpetuals.com ended acquisition talks with Alt5 Sigma Canada, a Trump-linked crypto firm tied to World Liberty Financial. (Reuters)

Deal of the Week

• The Deal: Uber → Delivery Hero ($14.8B). Uber is buying Delivery Hero to expand its global delivery footprint and create the largest food-delivery group outside China.

• Why It Matters: Food delivery is still a scale game, but scale alone does not fix weak unit economics. The bet is that Uber can combine logistics, demand, merchant relationships, and consumer frequency across more markets, while also cutting duplication in overlapping regions.

• Sara’s Take: This deal is about density, not just size. Delivery is a brutal business when each market is fought city by city, courier by courier, and coupon by coupon. Uber is trying to build a bigger network where rides, food, grocery, and local commerce reinforce each other. That logic makes sense, but the hard part is execution. Delivery Hero is a messy global asset with different brands, market positions, and regulatory environments. The headline synergy number will matter less than whether Uber can rationalize the portfolio without losing the local density it is paying for.

Market Trends & Takeaways

• Scale is back in consumer marketplaces. When growth gets harder, platforms start buying density, geography, and frequency.

• Robotics is becoming an operating strategy. Hyundai taking full control of Boston Dynamics suggests robotics is no longer a side bet. It is part of the manufacturing and automation roadmap.

• Portfolio pruning is accelerating. Thomson Reuters selling control of print is the same pattern across tech and media: keep the data, workflow, and AI roadmap; move slower legacy assets into more focused vehicles.

From the Data Room

• Cyber M&A is headed for a record year. WSJ reported that cybersecurity deal count is tracking toward a new high, with AI-security deals rising sharply. The takeaway is simple: if AI creates new attack surfaces, buyers will pay for the tooling that secures them. (WSJ)

• Compute contracts are becoming strategic assets. Reflection AI signed a compute deal with Nebius worth more than $1B, showing that model companies are still racing to lock up capacity before they lock up customers. In AI, infrastructure access is increasingly part of the competitive moat. (Reuters)

Hit reply with anything fun you’re reading this week or the best under the radar summer vacation spots (asking for a friend, that is most definitely me).

Share

• Qualcomm bought Modular for roughly $4B to strengthen its AI software stack.

• onsemi bought Synaptics for $7B to expand into AI-enabled devices.

• Superhuman bought GPTZero as AI-generated content makes verification more valuable.

• MoEngage bought Aampe to add AI-driven customer decisioning.

Subscribe now

Deals & Rumors

• Qualcomm → Modular (~$4B)
  Qualcomm agreed to acquire Modular in an all-stock deal valued at roughly $4B. Modular helps developers run AI models across different chips without rewriting software for every processor, giving Qualcomm a stronger platform story as it expands beyond smartphones into data centers and edge AI. (Reuters)

• onsemi → Synaptics ($7B)
  onsemi agreed to acquire Synaptics in an all-stock deal valued at about $7B. Synaptics brings connected computing, wireless connectivity, and human-machine interface technology, expanding onsemi’s position in automotive, industrial systems, robotics, and other AI-enabled devices. (Reuters)

• Superhuman → GPTZero (terms not disclosed)
  Superhuman acquired GPTZero, the AI-content detection company. As AI-generated writing spreads, detection and authenticity tools are becoming part of the productivity stack. (TechCrunch)

• MoEngage → Aampe (terms not disclosed)
  Customer-engagement platform MoEngage acquired Aampe, which uses AI agents to decide what message to send each customer, through which channel, and when. Generating marketing content is getting cheap; deciding what actually drives action is becoming a valuable layer. (PR Newswire)

Deal of the Week

• The Deal: Qualcomm → Modular (~$4B). Qualcomm is buying AI software infrastructure that helps models run across different chips and environments.

• Why It Matters: AI hardware only wins when developers can deploy on it easily. Nvidia’s advantage is not just its chips. It is the software ecosystem around them. Modular gives Qualcomm a way to make its hardware easier to use across data center and edge workloads.

• Sara’s Take: Qualcomm has spent years selling chips into phones. The bigger AI opportunity is in data centers, cars, and edge devices, but developers default to Nvidia because its software is familiar and easy to deploy. Modular gives Qualcomm a credible way to reduce that gap by helping developers run models across hardware without rebuilding their stack. This does not erase Nvidia’s advantage overnight. It does give Qualcomm a stronger position as a platform company rather than just a chip supplier.

Market Trends & Takeaways

• The AI chip race is becoming a software race. Hardware vendors are buying the developer layer to accelerate adoption.

• Deals in physical AI are accelerating. Chips, sensors, connectivity, and control systems are getting bundled for vehicles, robotics, and industrial automation.

• Trust is becoming a product layer. Detection, provenance, security, and governance are moving into core workflows as AI-generated content and agents become standard.

From the Data Room

• M&A is getting bigger, not broader. PwC expects global deal value to reach roughly $4T in 2026, with transactions above $5B representing nearly half of total value so far. Deal values are rising while volumes fall, which means large, well-capitalized buyers are driving the market. (PwC)

• PE diligence has entered its vibe-coding era. The Financial Times reports that Bain is using AI-generated software replicas to test how easily an acquisition target’s product can be rebuilt. AI risk is now part of underwriting, especially for software products with weak data, workflow, or distribution moats. (Financial Times)

• Power remains the bottleneck. Reuters reports that data-center investors are buying power developers to secure generation capacity, grid connections, and faster build timelines. U.S. data-center power demand is projected to rise from 31 GW in 2025 to 66 GW in 2027. (Reuters)

Share

• Salesforce is buying Contentful to bolt a real content layer onto Agentforce and its “headless” stack.

• Anthropic filed confidentially for an IPO

• Nvidia pushed further into AI on the device

• Most big-tech deal flow was quiet this week, but the strategic chessboard got louder.

Subscribe now

Deals & Rumors

• Salesforce → Contentful (terms not disclosed)
  Salesforce signed a definitive agreement to acquire Contentful, a headless/composable CMS used by thousands of brands. Salesforce has been selling the idea of “Agentforce everywhere.” This deal is the missing plumbing: agents need structured content that can be reused across web, app, email, support, commerce, and whatever interface shows up next. If you are competing in CRM/marketing clouds, take note. Content is becoming a first-class system asset again, not just “the thing in the CMS someone forgets to update.” (Salesforce)

• Snowflake → Natoma (terms not disclosed)
  Snowflake signed a definitive agreement to acquire Natoma, which focuses on governing how AI agents connect to enterprise systems via MCP. “Agents” are forcing a new security and identity layer, and platform companies want to own it before it turns into a tax. (Snowflake)

Deal of the Week

• The Deal: Salesforce → Contentful (signed; terms not disclosed).

• Why It Matters: Salesforce has been pushing agents and APIs, but content has been an awkward gap in the stack. Contentful gives Salesforce a structured content system that plays nicely with APIs and personalization, which makes agent-driven experiences more realistic and less duct-taped.

• Sara’s Take: This is a practical deal. Salesforce wants agents to generate and deliver customer experiences across channels, but you cannot do that cleanly if your content lives in a separate world with its own workflow, permissions, and schema. Contentful solves the “where does the approved content live?” problem and makes personalization easier because everything is structured. The second-order effect is competitive pressure on Adobe, SAP, and the marketing cloud ecosystem, because “content + data + orchestration” is how you win distribution as interfaces shift. The risk is integration and go-to-market overlap; Salesforce now has to explain, clearly, what sits where and why customers should pay for it.

Market Trends & Takeaways

• Agent stacks are converging. Platforms are grabbing the layers around agents: content, identity, governance, and connectors.

• Distribution is splitting. Some value moves to the browser and device as more AI runs locally, which changes who can control default experiences.

From the Data Room

• Anthropic filed confidentially for an IPO. That is a signal that the AI race is graduating from private valuation theater to public-market scrutiny, where revenue quality, margins, and compute spend get judged quarterly. (WIRED)

• Meta is exploring raising tens of billions to fund AI capex. If the big platforms are heading back to equity markets for infrastructure, it tells you the cost curve is not stabilizing yet. It also changes M&A behavior: fewer “nice-to-have” tuck-ins, more deals that shorten time-to-capacity or lock in distribution. (Financial Times)

• Nvidia’s push into PC “AI superchips” is a distribution bet. More inference on-device means new gatekeepers (OEMs, OS defaults, local runtimes), and it increases the strategic value of owning endpoints, not just models. (Financial Times)

There will be a 2 week hiatus until the next newsletter while I globetrot. See you soon.

Subscribe now

• Uber is circling Delivery Hero and the bidding is getting serious.

• Microsoft landed a $9.7B Pentagon umbrella contract.

• PE is in software and it is buying “hard to rip out” categories.

• Europe is leaning harder into subsidy and national-interest reviews for China-linked deals.

Subscribe now

Deals & Rumors

• Uber → Delivery Hero (offer/talks; price floated at €33+ per share)
  Delivery Hero confirmed it received a takeover offer from Uber, and Reuters also reported Uber and DoorDash sounding out investors around a bid. Uber has already built a meaningful stake, which is usually what you do when you want leverage whether or not you ultimately buy the company. This is a consolidation attempt in a tough category where scale still matters and standalone margin stories keep getting stress-tested. (Reuters)

• Microsoft ↔ U.S. Department of Defense ($9.7B contract)
  The Pentagon awarded Microsoft a five-year, $9.7B contract to consolidate software procurement across the military, intelligence community, and Coast Guard. Not M&A, but shows enterprise lock-in is very real and procurement wants fewer vendors and less license sprawl. (Reuters)

• Hg → Rightsline (~$500M investment)
  Hg agreed to a ~$500M strategic growth investment in Rightsline, which sells rights and royalties management software used by IP-heavy industries. Sponsors are leaning into categories where the value is tied to complex workflows and customer-specific data. (Business Wire)

• JD.com → Ceconomy (€2.2B) (deal scrutiny)
  The EU is preparing an in-depth foreign subsidies probe into JD.com’s bid for Ceconomy. Europe is essentially saying: if state support may have helped fund your offer, expect a longer and more political review process. (Financial Times)

• ProLogium → Translational Development Acquisition Corp ($3.8B SPAC deal)
  Taiwanese battery maker ProLogium will go public via a $3.8B SPAC merger. EV and energy storage are staying on the “public markets want a story” list even when traditional TMT IPOs are uneven. (Reuters)

• Terra Quantum → Axiom Intelligence Acquisition Corp 1 ($3.5B SPAC deal)
  Terra Quantum switched SPAC partners ahead of a planned listing, bumping its stated valuation and raising fresh proceeds. Quantum remains a market where financing structures matter. (Reuters)

Deal of the Week

• The Deal: Uber → Delivery Hero (in play), with an offer confirmed and broader bid discussions reported.

• Why It Matters: Food delivery is still over-fragmented globally, and it is hard to make the unit economics work without scale, density, and operational discipline. A credible path to consolidation changes the whole sector’s pricing power and competitive map.

• Sara’s Take: Strategic optionality started with a stake. Uber is building leverage inside Delivery Hero while the company is already under pressure to run a strategic review. If this turns into a real process, the hard parts will not be the headline synergies. It will be regulators, overlapping markets, and what you do with the Middle East and Korea assets that drive a lot of Delivery Hero’s value. Also, once one large platform makes a move, everyone else gets forced into a decision. Compete, merge, or get squeezed.

Market Trends & Takeaways

• Buyers are paying for distribution, even outside classic M&A.

• Software dealmaking is back, but the targets are shifting toward workflow-heavy and data-embedded categories.

From the Data Room

• Goldman says 2026 is tracking toward near-record M&A volumes. Translation: backlog is building, corporates are moving, and the market is acting like the window is open again. (Reuters)

• OpenAI committed $250M through its foundation to help workers and economies navigate AI disruption. Signal that AI’s labor impact is now a mainstream stakeholder topic. (Reuters)

Hit reply with your hot take of the week.

Share

• Publicis agreed to buy LiveRamp for $2.2B to bulk up its data engine.

• Meta’s Manus unwind is getting real, and expensive.

• Baker Hughes’ $13.6B Chart deal is heading into EU review mode.

Subscribe now

Deals & Rumors

• Publicis → LiveRamp ($2.2B)
  Publicis announced an all-cash deal for LiveRamp at $38.50 per share, nearly a 30% premium. This is an ad holding company paying up for durable data infrastructure. (Reuters)

• Meta → Manus AI (unwind aftermath)
  Reuters reported Manus’ founders are trying to raise about $1B to buy back the company after China ordered the Meta deal unwound. High US China tension and a show that the state can reach across borders into Singapore. (Reuters)

• Baker Hughes → Chart Industries ($13.6B) (regulatory timeline)
  EU regulators are set to decide by June 26 whether to clear Baker Hughes’ acquisition of Chart. Chart sells equipment that matters for LNG and data centers, which is why this sits at the intersection of industrial tech and the buildout everyone is quietly underwriting. (Reuters)

Deal of the Week

• The Deal: Publicis → LiveRamp ($2.2B)

• Why It Matters: Publicis is trying to own more of the data plumbing behind ad targeting and measurement. That matters when third-party signals keep disappearing and ad budgets swing with the macro.

• Sara’s Take: Publicis is buying leverage with advertisers. LiveRamp is a clean-room style connector that helps companies match data without directly handing it over, and that is exactly the kind of infrastructure buyers fund when privacy rules tighten. Data connectivity is becoming a core asset. And if you cannot prove performance, you become optional fast.

Market Trends & Takeaways

• Data assets with real distribution are still getting paid for, even when everything else feels like a budget cut.

• Cross-border AI deals with China exposure are a board-level risk topic.

From the Data Room

• UK inbound M&A is running hot. Reuters noted foreign bids have pushed UK M&A to about $192B already in 2026, and national security review has become part of the process, not an exception. (Reuters)

Don’t hit reply until after the long weekend ✌️

Share

• OpenAI bought Tomoro and launched a $4B deployment unit for enterprise rollouts.

• Microsoft is shopping for AI startups as it plans for life beyond OpenAI.

• Apple quietly picked up Color.io’s creator tooling assets.

Subscribe now

Deals & Rumors

• OpenAI → Tomoro (terms not disclosed)
  OpenAI is acquiring Tomoro, an AI consulting firm, as part of launching a new “Deployment Company” backed by $4B+ in initial investment. Most companies need a team that can integrate, govern, and ship AI into real workflows without breaking everything. (Reuters)

• Microsoft → Inception (talks / exploring a deal)
  Reuters reported Microsoft is in discussions with Inception, an AI startup focused on diffusion-based LLMs, and has been evaluating other options, including previously considering Cursor. Microsoft is trying to stockpile talent and model capability so it is not structurally dependent on OpenAI forever, and it is doing it while assuming regulators will scrutinize anything that looks like “Copilot bought the next Copilot.” (Reuters)

• Apple → Patchflyer (Color.io)
  Reporting this week pointed to Apple acquiring certain assets of Patchflyer GmbH (Color.io) and hiring the founder, based on an EU acquisitions list update. If accurate, it fits Apple’s pattern: small, targeted buys to pull creator-grade tooling into its ecosystem, without doing big public M&A. (9to5Mac)

Deal of the Week

• The Deal: OpenAI → Tomoro, alongside the launch of OpenAI’s Deployment Company with $4B+ in initial backing and structured as a multi-year partnership with PE sponsors.

• Why It Matters: AI is moving from “buy software” to “change how work happens.” That creates a services bottleneck, and whoever owns deployment capacity gets to decide which models become sticky inside the enterprise.

• Sara’s Take: This is OpenAI admitting what most buyers already know. Getting AI into production is not one integration, it’s a sequence of integrations, controls, and workflow redesign. That work is slow, political, and expensive. OpenAI buying Tomoro and building a deployment unit is a way to speed up enterprise adoption and keep customers from drifting to a rival that can execute faster. Longer term, this blurs the line between software vendor and services firm, and it will pressure the big systems integrators because the model providers now want to own the “last mile” too.

Market Trends & Takeaways

• AI M&A is shifting toward execution. Implementation capacity is becoming a competitive advantage.

• The Microsoft–OpenAI relationship is still the center of gravity, but both sides are building escape hatches.

From the Data Room

• OpenAI and Microsoft rewrote the economics. Reuters reported OpenAI and Microsoft agreed to cap total revenue sharing at $38B, which effectively increases OpenAI’s flexibility to do new partnerships and makes the cap table more IPO-friendly. (Reuters)

• Deal strategy is becoming “regulator-first.” Microsoft’s reported hesitation on Cursor due to antitrust scrutiny. (Reuters)

Hit reply with the most “obvious in hindsight” AI deal you’ve seen this year. I’m building a list of what everyone will pretend they called early.

Share

• Meta bought a humanoid robotics startup

• OpenAI and Anthropic are circling AI services firms

• Roche bought PathAI

Subscribe now

Deals & Rumors

• Meta → Assured Robot Intelligence (terms not disclosed)
  Meta acquired Assured Robot Intelligence, a humanoid robotics startup. This is a talent and capability grab. If Meta wants AI to live beyond chat windows, it needs systems that can perceive and act in messy real-world environments. (WSJ)

• OpenAI and Anthropic ventures → AI services firms (talks / rumor)
  Reuters reports OpenAI and Anthropic are pursuing acquisitions of AI services companies through PE-backed vehicles to staff up enterprise deployments. Translation: enterprise AI is hitting a bottleneck. (Reuters)

• Roche → PathAI ($750M upfront + up to $300M milestones)
  Roche agreed to acquire PathAI, an AI-driven digital pathology company, after years of partnership. This is a common playbook for regulated industries. Partner first, prove clinical value, then buy to lock in the platform and accelerate productization inside an incumbent’s distribution. (Reuters)

Deal of the Week

• The Deal: OpenAI and Anthropic ventures → AI services firms (talks / rumor).

• Why It Matters: Everyone wants enterprise AI. Few companies can actually deploy it at scale without breaking existing systems. Buying services capacity is the fastest way to solve the last-mile problem: integrations, change management, and owning outcomes.

• Sara’s Take: Models are not the hard part anymore. Implementation is. Services firms give them teams that can sit with customers, connect AI to real data, and ship production workflows. It also changes how AI companies compete. They are not just selling software, they are selling delivery. If this works, expect more roll-ups of niche consultancies and more hybrid offers that look like “platform plus staffed deployment.” It will make AI stickier inside enterprises and it will put pressure on traditional SIs that move slower and cost more.

Market Trends & Takeaways

• Robotics is back as a strategic frontier. Big platforms want embodied AI, not just chat.

• AI is professional services-heavy in the near term.

From the Data Room

• SoftBank had to resize its OpenAI margin-loan ambitions. Reuters reported SoftBank cut its target from $10B to as low as $6B, showing even obvious winners can be hard to underwrite. (Reuters)
  

Share

• China ordered Meta to unwind its Manus AI deal; cross-border AI M&A just got harder

• Adobe closed its Semrush acquisition

• The week’s themes: regulation and distribution

Subscribe now

Deals & Rumors

• Meta → Manus AI (unwind ordered; deal size previously reported ~$2B+)
  China ordered Meta to unwind its acquisition of Manus as Beijing tightens scrutiny of foreign investment in frontier tech. (Reuters)

• Adobe → Semrush (completed; deal value not in the release)
  Adobe completed its acquisition of Semrush, bringing SEO/visibility tooling into the Adobe Experience Cloud orbit. Discovery is shifting from “Google blue links” to AI answers, and brands want one system that connects content creation, measurement, and visibility. (Business Wire)

• Cohere ↔ Aleph Alpha (merge / tie-up)
  Cohere announced it would merge with Germany’s Aleph Alpha to form what it pitched as a transatlantic enterprise AI player. (TechCrunch)

• SiriusXM ↔ iHeartMedia (talks / rumor)
  Reuters flagged potential merger talks. I forgot either of these companies existed, but the logic is straightforward: combine audio scale and distribution in a choppy ad market, and the bundle makes sense. (Reuters)

• Alphabet/Google → Anthropic (reported investment plan up to $40B)
  Reuters reported Google planned to invest up to $40B in Anthropic (via Bloomberg report). This is one of those “control without acquiring” moves that still reshapes the competitive landscape. (TheStreet)

Deal of the Week

• The Deal: Meta → Manus AI (unwind ordered) — China ordered Meta to unwind its Manus AI acquisition.

• Why It Matters: Cross-border AI M&A is a regulatory minefield. Even after you complete the deal, a government can effectively veto the integration if the asset is viewed as strategic.

• Sara’s Take: This deal got caught in the middle of the US–China tech fight. China ordered Meta to unwind the Manus acquisition on national security grounds, and the message is straightforward: Beijing does not want top AI talent and frontier IP ending up inside a US company, especially one building strategic AI capabilities. Manus reportedly shifted operations to Singapore, but China still asserted jurisdiction and blocked the outcome anyway.

  
  What changes going forward:

  • More deals will die quietly before signing. If a target has meaningful China ties, buyers will assume a real chance the deal cannot be completed or integrated, even if the paperwork looks clean.

  • Expect more “structure-first” behavior. More carve-outs, licensing, minority stakes, and partnerships instead of full acquisitions, since full change of control is what triggers the hardest scrutiny.

  • AI becomes a protected asset class. This is a preview of broader restrictions on outbound AI talent and technology, and it raises the bar for any US–China adjacent transaction in advanced tech.

  
  Net effect: cross-border AI M&A between US and China-linked companies becomes slower, riskier, and more political.

Market Trends & Takeaways

• Geopolitics is now a diligence workstream. If the asset is “frontier AI + cross-border talent,” expect regulatory intervention to be a first-order risk.

From the Data Room

• Bridgewater warns AI could be existential for legacy software. This is the kind of narrative that pushes boards toward consolidation, divestitures, or “buy the missing capability” M&A. (Reuters)

• Secondaries are becoming a real price signal. WSJ noted Anthropic overtaking SpaceX in secondary trading activity. More evidence that private valuations are increasingly discovered in a shadow market, not just funding rounds. (The Wall Street Journal)

Hit reply with whatever you’re feeling this week & as always, thanks for reading.

Share

• Amazon agreed to buy Globalstar for $11.57B

• OpenAI acqui-hired personal finance startup Hiro and inked a monster compute deal with Cerebras

• Theme: AI is turning “infrastructure” into the hottest M&A category again

Subscribe now

Deals & Rumors:

• Amazon → Globalstar ($11.57B)
  Amazon signed a definitive deal to acquire satellite operator Globalstar, with Apple’s existing relationship sitting squarely in the middle of the story. The strategic value isn’t “owning satellites”—it’s Globalstar’s spectrum rights and regulatory approvals, which can accelerate Amazon’s Project Kuiper (its Starlink-style satellite broadband network) and future direct-to-device connectivity plans. (Reuters)

• OpenAI → Hiro Finance (terms not disclosed)
  OpenAI acquired Hiro Finance, a consumer financial planning startup; per reporting, it looks like an acqui-hire with the team joining OpenAI and the product winding down. The signal: OpenAI is getting serious about verticalized, high-trust use cases (finance), where “close enough” answers are a liability and product UX matters as much as model quality. (TechCrunch)

• OpenAI ↔ Cerebras (>$20B spend commitment + potential equity warrants)
  Reuters reported OpenAI committed to spend more than $20B over three years on Cerebras-powered compute and may receive warrants for a minority stake. Not an acquisition, but it’s a classic control move: when compute is scarce, long-term supply (and optionality on the supplier) is core to strategy. (Reuters)

Deal of the Week:

• The Deal: Amazon → Globalstar ($11.57B).

• Why It Matters: This is Amazon buying a shortcut to operating at scale in space communications. Spectrum rights and approvals take years to secure; acquiring them can be faster than building them—especially when regulatory clocks are ticking.

• Sara’s Take: Amazon is buying access: spectrum rights, licenses, and an operating footprint that’s hard to replicate quickly. If Amazon can secure Globalstar, it can speed up Project Kuiper by inheriting permissions that usually take years to assemble. Takeaway: in infrastructure deals, the licenses and agreements can matter more than the technology.

Market Trends & Takeaways:

• Infra is back at the top of the stack. Connectivity and compute are becoming strategic choke points, so buyers are paying for control.

• AI “verticalization” is real. The easy part was chat; the hard part is building domain products people trust (like money).

From the Data Room:

• Deals are back in fashion (for now). ~11,000 global deals totaling ~$1.2T, arguing that a friendlier policy tone plus AI enthusiasm is loosening risk appetite.(Reuters)

• Secondary markets are becoming a real price signal. Anthropic overtaking SpaceX in secondary trading activity—evidence that secondaries are maturing into a “shadow market” for private valuations. (WSJ)

Hit reply with the most strategic deal you saw this week—and the one you think is going to become a very expensive lesson in “dependencies.”

Share

• OpenAI bought TBPN (yes, a media property).

• Amazon is in talks to buy Globalstar for ~$9B.

• PE kept shopping: Hg closed its $6.4B take-private of OneStream.

• The Q1 tape: fewer deals, bigger checks—AI is pulling the average up.

Subscribe now

Deals & Rumors

• OpenAI → TBPN (terms not disclosed)
  OpenAI acquired TBPN, a daily live tech-and-business show popular with founders for product news and industry talk. (Axios) (The Verge)

• Amazon → Globalstar (~$9B) (rumor)
  Amazon is in talks to buy satellite operator Globalstar, with Apple’s stake/relationship as a key complication. The strategic value is in Globalstar’s spectrum rights and regulatory approvals, which could speed up Project Kuiper (Amazon’s Starlink-style satellite broadband network) by giving Amazon a faster path to operating at scale. (Reuters)

• Hg → OneStream ($6.4B)
  OneStream announced the completion of its acquisition by Hg, taking the enterprise finance platform private. Take-privates like this are the PE playbook right now: buy durable, mission-critical software when public multiples are shaky, then do the product and pricing work away from the quarterly microscope. (OneStream IR)

Deals of the Week

• The Deal: Amazon → Globalstar (~$9B) (rumor), per Reuters/FT reporting.

• Why It Matters: This is what “AI infrastructure” looks like once you zoom out: not just GPUs—connectivity, coverage, spectrum rights, and a credible path to global service. Owning Globalstar could let Amazon skip years of build-out and regulatory friction, while putting real pressure on Starlink’s lead.

• Sara’s Take: Don’t think of this as “Amazon buying satellites.” Think of it as Amazon buying access: spectrum rights, licenses, and an operating footprint that takes years to secure. If Amazon can acquire Globalstar, it shortcuts regulatory time and speeds up Kuiper’s path to real coverage, because in space the bottleneck is approvals and launch cadence. The Apple angle matters because Apple already uses Globalstar and has influence there, so Amazon is inheriting a partner relationship that could constrain what it can do next. Takeaway: in infrastructure deals, the cap table and contracts can be as strategic as the technology.

& a second special bonus deep dive, since I had to spend some time trying to wrap my head around the latest OpenAI buy:

• The Deal: OpenAI → TBPN (terms not disclosed)

• Sara’s Take: I don’t actually know exactly why OpenAI bought TBPN, but there are a couple plausible angles. One is data: that archive—thousands of hours of candid, long-form conversations with operators and executives—could be useful training and evaluation material for better AI-generated journalism, analysis, and summarization products. Another is distribution: owning a media channel lets OpenAI reach builders directly and shape the narrative as AI becomes more politicized and more regulated. Either way, it’s a leverage deal: information, influence, and a feedback loop on what the industry is thinking in real time.

Market Trends & Takeaways

• Bigger checks, fewer swings: deal volume is down, but mega-deals keep the dollar totals climbing.

• Distribution wars are expanding: “own the channel” now includes media, not just app stores and browsers.

• PE is leaning into software dislocation: take-privates thrive when public comps are messy.

From the Data Room

• Global M&A topped ~$1.2T in Q1—up ~26% YoY—despite fewer deals. Translation: the middle market is cautious, but the winners are still writing big checks (especially in AI-adjacent categories). (Reuters)

• “SaaSpocalypse” is accelerating enterprise software M&A, not killing it. Lower public multiples are turning into a shopping season for sponsors and strategics hunting defensible, must-have software. (Fortune)

• Hyperscaler capex is still climbing hard: $309B (2024A) → $779B (2027E), and their point is simple: this is a real buildout, not just a narrative, which is why “AI infrastructure” keeps pulling deal dollars upmarket. (Redpoint)

• Public markets are getting less forgiving: S&P 500 is trailing P/E at ~19.9x (3/31/26) vs a ~21.4x 10-year average, plus sharp sector rotation—conditions that usually nudge boards toward “sell / merge / get scale” conversations. (LionTree)

Share

• Amazon bought two robot startups for last-mile delivery + in-home presence.

• OpenAI bought Astral to deepen its grip on the Python developer workflow.

• Databricks bought Antimatter and SiftD.ai and launched Lakewatch to expand into security.

• Meta and Salesforce continued the “agent org” talent land-grab (Dreamer, Clockwise teams).

Subscribe now

Deals & Rumors

• Amazon → Rivr (terms not disclosed)
  Amazon acquired Rivr, the stair-climbing delivery robot startup. This is Amazon saying the last 30 feet of delivery is still the hardest part and it would like to automate the part where your package mysteriously becomes “left in a safe place” (aka not your building). (TechCrunch)

• Amazon → Fauna Robotics (terms not disclosed)
  Amazon bought Fauna Robotics, maker of the kid-size humanoid robot “Sprout,” keeping the company operating under its name. It’s a bet that robots aren’t just warehouse tools, they’re eventually consumer devices with software ecosystems (and Amazon wants to be the platform). (AP)

• OpenAI → Astral (terms not disclosed)
  OpenAI is acquiring Astral, the Python toolmaker behind popular dev tools (think: making Python workflows faster and cleaner). OpenAI will get to control the tools developers touch daily and the distribution that comes with that. (OpenAI)

• Databricks → Antimatter + SiftD.ai (terms not disclosed)
  Databricks announced acquisitions of Antimatter and SiftD.ai alongside its Lakewatch security push. Data platforms are expanding into security because customers want governance, detection, and controls in the same place their data and AI workloads live. (Databricks)

• Meta → Dreamer (team/asset deal, terms not disclosed)
  Meta brought in the founders/team behind agent startup Dreamer into its AI org. The pattern is getting loud: Meta is buying teams to compress time-to-capability in agents, because normal hiring for superintelligence is not fast enough for Meta. (Reuters via TradingView)

• Salesforce → Clockwise (team acquisition; product shutting down)
  Salesforce is bringing the Clockwise team into Agentforce; the Clockwise app will shut down and customer data will be deleted. It’s an acquihire that brings the parts that run your day, e.g. calendaring, prioritization, and scheduling intelligence. (TechRadar)

Deal of the Week

• The Deal: OpenAI → Astral, bringing major Python developer tools into the Codex ecosystem.

• Why It Matters: The AI coding race isn’t just models, it’s distribution inside the developer workflow. If OpenAI owns the tools that lint, package, run, and ship code, it can make “Codex” feel inevitable instead of optional.

• Sara’s Take: This is OpenAI doing the unsexy but smart thing: buying defaults. Models are impressive; tooling is sticky. The next wave of “AI coding” winners will win because they reduce friction in the boring parts: setup, dependencies, tests, CI. Also, this is how you build a moat: become the path of least resistance for developers.

Market Trends & Takeaways

• Agents are driving stack consolidation. Platforms are pulling workflow, security, and governance into core suites.

• Robotics is moving to the edge. More M&A aimed at last-mile delivery and consumer-adjacent use cases, not just fulfillment centers.

• Acqui-hires are back in force. Teams with real agent experience are getting bought faster than they can be recruited.

From the Data Room

• Regulators are reworking the paperwork. The FTC/DOJ opened a public comment process on the premerger notification form—translation: more disclosure expectations are coming, and “we’ll figure it out later” is about to get more expensive. (FTC)

• AI capex is now a debt-market storyline. After Amazon’s bond activity, analysts raised their 2026 hyperscaler debt forecast to $175B—cost of capital is becoming a competitive advantage (and a constraint). (Reuters)

• Secondaries are becoming the pressure valve. New PitchBook estimates put U.S. direct secondaries volume in 2025 somewhere between $62.5B and $120.9B—messy data, clear signal: liquidity is moving from IPOs to “someone else buys your stake.” (Fortune)

Hit reply with the deal you think is quietly genius—and the one that’s going to look like a very expensive science fair project by Q4.

Share

• Google closes $32B Wiz

• Meta buys Moltbook (AI agents need their own Reddit now)

• OpenAI buys Promptfoo (agent security)

• Nvidia takes an 8.3% stake in Nebius for $2B

Subscribe now

Deals & Rumors

• Alphabet/Google → Wiz ($32B)
  Google closed its all-cash acquisition of Wiz, folding one of the hottest cloud security platforms into Google Cloud. This is Google paying to move from “credible in cloud security” to “default short-list,” especially as AI workloads and multicloud sprawl turn misconfigurations into a hobby for attackers. (Google Cloud)

• Meta → Moltbook (terms not disclosed)
  Meta acquired Moltbook, a social platform built for AI agents, and is bringing the founders into its AI org. It’s less about the app and more about talent + a live sandbox for agent behavior, because the next frontier is AI that acts like a user, not a chatbot. (Reuters)

• OpenAI → Promptfoo (terms not disclosed)
  OpenAI is acquiring Promptfoo, an AI security/evaluation platform, and plans to integrate it into OpenAI Frontier. Translation: as agents move into enterprise workflows, “can it do the task?” is no longer enough, buyers want testing, monitoring, and guardrails. (OpenAI)

• NVIDIA → Nebius (8.3% stake, $2B investment)
  Nvidia invested $2B in AI cloud firm Nebius for an 8.3% stake as “neoclouds” race to stand up GPU capacity. Not an acquisition, but it’s the same strategic play: Nvidia is financing the ecosystem that buys Nvidia chips: vertical integration, just with spreadsheets. (Reuters)

Deal of the Week

• The Deal: Alphabet/Google → Wiz ($32B), completed this week, bringing Wiz into Google Cloud.

• Why It Matters: Cloud security is the control plane for modern IT and the control plane is where platforms win. Wiz gives Google a credible “security default” story in multicloud, right as AI expands the attack surface and raises the price of mistakes.

• Sara’s Take: This is Google admitting the fastest way to win cloud isn’t another product launch, it’s buying trust. Security buying decisions are sticky, political, and risk-averse; Wiz already had the relationships, the deployment footprint, and the “we actually work” reputation. Once you’re the security layer, you influence everything upstream (cloud spend, architecture, even which AI tools get approved). Also: $32B is expensive, but “third place in cloud” is more expensive.

Market Trends & Takeaways

• Agents are becoming a product category—and an acquisition category. Meta’s Moltbook buy and OpenAI’s Promptfoo buy are both about operationalizing agents: behavior, safety, monitoring, and iteration loops.

• “AI infra” is financing as strategy. Nvidia isn’t just selling chips, it’s underwriting the customers who will need more chips.

• Security is the fastest route to platform power. If you own the policy layer, you influence the whole stack and you get invited to every budget meeting.

From the Data Room

• Private-market valuations are getting a regulator spotlight. The SEC held/teed up a roundtable on how private assets get valued as “retail access” expands, i.e., the days of vibes-based marks and friendly 409As are numbered. That matters for M&A because your comps, your earnouts, and your exit math all start with what the market will actually believe. (SEC)

• Banks are quietly tightening the screws on software-backed credit. JPMorgan is marking down collateral values on loan portfolios tied to private credit, especially software names seen as AI-vulnerable. Translation: financing gets harder → PE gets pickier → processes get slower → multiples get less magical. (Financial Times)

• Mega-cap AI capex is turning into a bond market story. Amazon’s reported jumbo bond raise is a reminder that “AI strategy” now comes with balance-sheet choreography. When capex is this large, cost of capital becomes a competitive advantage and it changes who can afford to buy what. (Reuters)

Hit reply with the deal you think will look obvious in hindsight, and the one you think becomes a very expensive learning experience.

Share

• Netflix bought an AI movie-production startup.

• BlackRock/EQT are taking AES private for $33.4B.

• Wearables are quietly turning into “gesture-first” interfaces.

• Parking tech is consolidating.

Subscribe now

Deals & Rumors

• Netflix → InterPositive (terms not disclosed)
  Netflix acquired InterPositive, a filmmaking technology company founded by Ben Affleck that builds AI-powered tools for movie production. This is Netflix buying leverage in the messy, expensive part of content: making it faster/cheaper to produce. (Reuters)

• BlackRock (GIP) + EQT-led consortium → AES ($33.4B incl. debt)
  A BlackRock/EQT-led group agreed to acquire power company AES in a take-private deal. Call it the “AI boom toll road”: if data centers are the gold rush, power generation and grid-scale storage are the picks-and-shovels with recurring revenue. (Reuters)

• Oura → Doublepoint (terms not disclosed)
  Oura acquired Doublepoint, a startup focused on gesture-recognition tech that lets users control wearables through natural movements. The subtext: the next interface war isn’t screen size, it’s how you command devices when your hands are full and your phone is buried in a tote bag. (TechCrunch)

• Arrive → Passport (terms not disclosed)
  Arrive announced plans to acquire parking-tech provider Passport. This is classic vertical software roll-up logic: unify payments + enforcement + ops data, then sell “smart city” dashboards to municipalities who mostly want fewer angry emails. (Automotive World)

• HCLSoftware → Wobby (~$5.2M)
  HCLSoftware is set to acquire agentic-AI startup Wobby. It’s a small-ticket deal with a big pattern: incumbents are buying agentic capability as a feature, not waiting for it to become a platform tax later. (Staffing Industry Analysts)

Deal of the Week

• The Deal: Netflix → InterPositive, an AI-powered movie-production tooling company founded by Ben Affleck.

• Why It Matters: Streaming is a margin business. Owning production tech is one of the few ways to bend the cost curve without asking subscribers to fund your next prestige drama.

• Sara’s Take: Netflix has spent a decade perfecting distribution and recommendations; now it’s coming for the factory floor. AI in production can compress timelines, reduce reshoots, and make crews less dependent on heroic overtime. Expect more investment in tooling buys in this space, because the best content strategy is still a lower cost curve.

Market Trends & Takeaways

• AI is dragging M&A into the physical world: power companies and grid assets are suddenly “strategic,” because GPUs don’t run on thought leadership.

• Interface buys are creeping back: gestures/controls are the new battleground for wearables and buyers are scooping up the enabling tech early.

From the Data Room

• Oracle is reportedly preparing thousands of job cuts under the weight of AI data-center expansion economics. Translation: infra growth is starting to fight CFO gravity. (Reuters)

• Nvidia’s CEO is signaling the window may be closing on mega-check “strategic” investments into model labs as IPO paths firm up; expect partnerships to get more contractual and less romantic. (Reuters)

• Palantir scrambling to swap out Anthropic inside Pentagon AI software is the lock-in lesson: when you build on someone else’s AI, you’re also inheriting policy risk. When the rules change, you have to replatform on the customer’s timeline and your dime. (Reuters)

Gesture a reply with the most underrated deal of the week.

Share

• Equinix + CPP buy Nordic data centers for $4B.

• Brink’s grabs NCR Atleos for ~$6.6B.

• Pure Storage becomes Everpure + buys 1touch.

• Stripe–PayPal mega-rumor is back.

• Apple does a stealthy acquihire (as is tradition).

Subscribe now

Deals & Rumors

• Stripe → PayPal (rumor / early-stage interest)
  Reuters reported (via Bloomberg) that Stripe is considering acquiring all or parts of PayPal; both declined comment. If true, it’s an attempt to buy consumer reach + merchant distribution + a competitive reset in one very expensive move; if not, it still signals that payments giants feel the consolidation gravitational pull again. (Reuters)

• Apple → invrs.io (assets + acquihire surfaced this week)
  Reporting tied to EU filings indicates Apple acquired certain assets from invrs.io and hired its founder/sole employee. This is Apple’s favorite M&A genre: buy niche R&D, absorb it into the mothership. (Heise)

• CPP Investments + Equinix → atNorth ($4B)
  CPP and Equinix teamed up to buy atNorth, a Nordic high-density colocation provider, from Partners Group. This is less “real estate” and more “securing electricity and cooling before the AI bill comes due.” (Equinix)

• Brink’s → NCR Atleos (~$6.6B incl. debt)
  Brink’s agreed to acquire NCR Atleos in a cash-and-stock deal to expand across ATM and cash-management networks. Translation: owning the boring rails is still a great business when banks keep shrinking physical branches but still need cash logistics to work. (Reuters)

• Everpure (formerly Pure Storage) → 1touch (terms not disclosed)
  Everpure announced a rebrand and a deal to acquire 1touch (data intelligence/orchestration). Storage companies are trying to stop being “boxes and bits” and start being “the layer that makes your data usable for AI,” because margins like that idea. (Pure Storage)

• GameSquare → TubeBuddy (terms not disclosed)
  GameSquare announced an asset purchase of TubeBuddy from BENlabs, paying with preferred stock. Creator tooling is becoming go-to-market infrastructure. (GameSquare IR)

• Corel (owned by KKR) → Split / Vector Capital acquires Corel software brands (terms not disclosed)
  Corel will separate into two companies: Vector will acquire Corel’s creativity/productivity software portfolio, while KKR retains Parallels. It’s PE portfolio hygiene: separate the mature cash cow from the strategically sensitive infra-ish asset so each can be optimized (and sold) on different clocks. (GlobeNewswire)

• FUJIFILM Business Innovation → ETG Global (terms not disclosed)
  Fujifilm Business Innovation will acquire ETG Global (ERP implementation/services). In an era where “digital transformation” mostly means “we can’t hire enough people who know how to implement the systems we bought,” acquiring delivery capacity is the fastest path to revenue. (Fujifilm)

Deal of the Week

• The Deal: CPP Investments + Equinix → atNorth ($4B) to scale high-density data center capacity in the Nordics. (Equinix)

• Why It Matters: AI growth is now constrained by physical inputs—power, cooling, and permits. Buying capacity in favorable geographies is a strategic advantage, not a facilities decision. (Equinix)

• Sara’s Take: Everyone keeps saying AI is “just software,” and then immediately fights over megawatts like it’s 2005 oil. The Nordics are winning because cold weather is a cheat code and grid access is a moat you can’t spin up in a sprint. If you’re diligencing infra, treat power contracts the way you treat IP: defensible, scarce, and worth paying for. The next wave of winners won’t be the teams with the best demos—they’ll be the teams that can actually run them. (Equinix)

Market Trends & Takeaways

• AI infra is getting “hard-asset” serious: power/cooling/location are driving strategy and valuations.

• The stack is moving upward: storage/IT vendors want orchestration + governance layers, not just hardware margins.

• Payments consolidation vibes are back: even a rumor can move the market because the incentives are real.

From the Data Room

• FTC/DOJ reopened “collaboration guidance” for comment: relevant if your “strategic partnership” starts looking like control. (FTC)

• AI server demand is still ripping: Dell projected AI server revenue to more than double; context for why capacity/infra M&A isn’t easing up. (Reuters)

That’s all for now - hit reply if you’d like to share some signal.

Share

• Salesforce → Momentum: CRM’s next moat is the customer conversation.

• Mistral → Koyeb: model shops are buying their own infra fate.

• Palo Alto → Koi: platform push—secure identity + endpoints for the agent era.

• Booz Allen → Defy: defense + commercial cyber keeps merging in practice and M&A.

• Valsoft → DigitalEd: vertical SaaS roll-ups continue to quietly print.

Subscribe now

Deals & Rumors

• Salesforce → Momentum (terms not disclosed)
  Salesforce signed a deal to buy Momentum (conversational insights + revenue orchestration) so Agentforce/Slack can ingest calls and meetings and turn “what the customer said” into “what the system does.” (Salesforce)

• Mistral AI → Koyeb (terms not disclosed)
  Mistral is buying Koyeb (AI infra platform) because every model company eventually realizes the unglamorous truth: reliability beats vibes, and GPUs don’t manage themselves. (WSJ)

• Palo Alto Networks → Koi (terms not disclosed)
  Palo Alto recently acquired Israeli startup Koi as part of its “platform-everything” push, because identity + endpoint security gets weird fast when AI agents start acting like employees. (Reuters)

• Booz Allen Hamilton → Defy Security (terms not disclosed)
  Booz Allen agreed to acquire Defy Security, expanding its commercial footprint.(Business Wire)

• Valsoft → DigitalEd (terms not disclosed)
  Valsoft acquired DigitalEd (higher-ed assessment + credentialing tools), continuing the roll-up playbook: buy sticky vertical software, modernize quietly, repeat forever. (Valsoft)

• Paramount → Warner Bros. Discovery (bid process / regulatory step)
  Paramount said the U.S. antitrust waiting period expired for its $108.4B all-cash offer for WBD. This isn’t “approved,” but it’s a meaningful signal that the regulatory road might be less brutal than expected. (Reuters)

Deal of the Week

• The Deal: Salesforce is acquiring Momentum, a conversational insights and revenue orchestration platform.

• Why It Matters: AI “agents” are only as good as their inputs and the highest-value input in sales is still the customer conversation.

• Sara’s Take: Most CRMs are missing the actual plot: the customer doesn’t live in your fields, they live in your calls. Momentum is Salesforce buying the bridge from messy human language → structured system actions. It’s also a quiet power move for Slack: if the meeting is where decisions happen, Slack wants to be where decisions get executed. The next CRM moat isn’t a dashboard, it’s capturing intent where it happens.

Market Trends & Takeaways

• Conversation is the new interface layer: buyers are paying to turn calls/meetings into usable data, because agents need context, not just records.

• Platform consolidation is accelerating: security and workflow players keep bundling point tools into “one pane of glass” (and one renewal).

• Infra is strategy again: model companies are moving closer to owning the stack so they can ship reliably.

From the Data Room

• HSR got legally messy: a federal court decision vacated the FTC’s new HSR rules (with a short stay), which is a fancy way of saying deal teams may be bouncing back to the older filing form. (Debevoise)

• Private equity’s new KPI is “cash back, not vibes”: Allianz Research notes 2025 exit value rebounded to $905B, but ~78% came from mega-exits—meaning mid-market liquidity is still tight and sponsors are under pressure to prove distributions. (Allianz Research)

• Salesforce’s rationale, in analyst English: Forrester frames Momentum as less “another sales AI feature” and more “own the highest-value input for autonomous agents: the customer conversation.” (Forrester)

  
  Until next time..

  Share

• TI → Silicon Labs ($7.5B): TI just bought wireless IoT scale—more “things that connect,” fewer suppliers to choose from.

• SiTime → Renesas timing (up to ~$3.2B): the boring clock inside billions of chips just got more consolidated.

• Siemens → Canopus AI: chip metrology is now strategic, because yield is the real KPI.

• TransUnion → RealNetworks mobile: fraud moved to calls/texts, so TransUnion followed the money leak.

• Tyler → For The Record: govtech keeps bundling workflows—turn meetings into minutes, then sell the platform.

Subscribe now

Deals & Rumors

• Texas Instruments → Silicon Laboratories ($7.5B)
  TI is buying Silicon Labs to add embedded wireless connectivity to its analog/embedded empire, because the next wave of devices needs to sense, connect, and not die in a factory. (TI) (Financial Times)

• SiTime → Renesas timing business (up to ~$3.2B)
  SiTime is acquiring Renesas’ timing assets in a deal that could push SiTime tech into billions of chips. (SiTime)

• Siemens → Canopus AI (terms not disclosed)
  Siemens acquired Canopus AI, which does computational metrology for semiconductors. Translation: better measurement = better yield = more chips when capacity is precious. (Siemens)

• TransUnion → RealNetworks mobile division (terms not disclosed)
  TransUnion is buying RealNetworks’ mobile unit to level up voice + messaging risk tools—fraud has moved to phone channels, so the data guys are following it there. (TransUnion)

• Tyler Technologies → For The Record (~$212.5M cash purchase price; ~ $258M enterprise value cited by seller)
  Tyler is acquiring court-recording tech For The Record. Govtech is quietly doing what enterprise does: automate the annoying workflow, then sell the suite. (Tyler) (PR Newswire)

Deal of the Week

• The Deal: Texas Instruments is acquiring Silicon Laboratories for $7.5B in cash.

• Why It Matters: This is TI buying its way into the “wireless edge” layer, where factories, cars, buildings, and gadgets all need connectivity plus long product cycles (read: sticky, profitable).

• Sara’s Take: Every tech cycle eventually turns into a plumbing cycle. Silicon Labs isn’t a flashy AI logo; it’s the chip that quietly makes your stuff talk to other stuff. TI’s superpower is scale + manufacturing + being the dependable supplier—so adding wireless is basically saying, “We want to own more of the bill of materials inside everything.” Also: if you’re an IoT startup, congrats—your supplier just got bigger and your negotiating leverage just got smaller.

Market Trends & Takeaways

• Semis are consolidating around “enablers,” not just compute: timing, connectivity, and metrology are getting treated like strategic assets, not components.

• Trust/fraud tooling is a real M&A lane: identity and communications security are where money leaks out first, so buyers are shoring up there.

From the Data Room

• 2026 deal mood: optimistic, with caveats: BCG’s M&A outlook frames expectations as “high again,” with tech still a driver—but valuation discipline and financing creativity are doing more work than the press releases admit. (BCG)

• EU’s “Big Tech gatekeeper” law (DMA) check: The European Commission said Apple Maps and Apple Ads won’t be treated as “gatekeepers” under the DMA, meaning they dodge extra competition rules in Europe for now—useful signal on which product surfaces regulators think actually control distribution. (European Commission)

• Outbound investment rules are tightening: the FY2026 NDAA updates to the U.S. outbound investment security regime (COINS Act / OISP) are another reason cross-border tech deals will keep getting more paperwork-y, not less. (White & Case)

Hit reply and tell me: which of these is the smartest “boring infrastructure” buy… and which one is just a company paying real money to avoid building a product roadmap?

Share

• Capital One is buying Brex for about $5.15B, going full send on B2B spend + payments.

• Apple grabbed Q.ai (reported ~$1.6B–$2B) as a bet on audio AI and “silent speech” interfaces.

• IonQ is buying U.S. chip foundry SkyWater for $1.8B to get closer to the metal.

• NVIDIA put $2B into CoreWeave to lock in compute capacity without doing an acquisition.

• SoftBank reportedly walked away from talks to buy data center operator Switch.

Subscribe now

Deals & Rumors

• Capital One → Brex (~$5.15B)
  Capital One is buying Brex to bolt modern spend management onto a bank balance sheet—software + deposits + payment rails in one tidy bundle. (Reuters)

• Apple → Q.ai (reported ≈$1.6B–$2B)
  Apple picked up Q.ai as an audio/“silent speech” interface bet—because the assistant that wins is the one that works in real life, not just in a keynote. (Reuters)

• IonQ → SkyWater Technology ($1.8B)
  IonQ is buying a U.S. foundry to get closer to manufacturing—vertical integration as insurance against roadmap slips and supply-chain drama. (Reuters)

• NVIDIA → CoreWeave ($2B strategic investment)
  NVIDIA is effectively reserving compute capacity by writing checks—strategic control without the acquisition paperwork. (Reuters)

• G2 → Capterra / Software Advice / GetApp (terms not disclosed)
  G2 is consolidating the software discovery funnel—own the marketplace, own the buyer intent, and (conveniently) the pay-to-play shelf space. (PR Newswire)

• Palo Alto Networks → Chronosphere (announced ~$3.35B; now closed)
  Palo Alto now owns more of the telemetry layer—because logs are the receipts, and receipts drive renewals. (Palo Alto Networks)

• MicroVision → Luminar assets ($33M, via bankruptcy process)
  The LiDAR space is entering its “asset sale” era—IP survives, standalone public-company dreams… less so. (MicroVision)

• SoftBank → Switch (talks halted, rumored ~$50B)
  No deal, but loud signal: AI infrastructure ambition is running into physics (power, sites, timelines), even at SoftBank scale. (Bloomberg)

Deal of the Week

• The Deal: Apple acquired Q.ai, a reported ~$1.6B–$2B-ish bet on audio AI and “silent speech” interfaces.

• Why It Matters: Apple isn’t just collecting AI talent, it’s collecting control points. If the interface shifts from screens to always-on wearables, the company that owns the sensors, distribution, and default devices has the leverage.

• Sara’s Take: The real story here is interface strategy: the “best model” doesn’t matter if it can’t hear you on the subway or understand you in the messy reality of actual humans. And, of course, this is Apple’s signature move—make a big AI bet, keep a straight face, and still avoid saying the phrase “Siri doesn’t work”.

Market Trends & Takeaways

• Buy the interface, not the demo: the most valuable AI is the one embedded in everyday inputs (audio, wearables, ambient context).

• Payments is a platform war again: big banks want modern spend software because it’s sticky, data-rich, and drags deposits along for the ride.

• AI infra is colliding with physics: power, land, and build times are shaping strategy as much as chips—hence giant investments and giant almost-deals.

From the Data Room

• UK competition regime reform is in motion: the government launched a consultation aimed at making antitrust enforcement faster and more predictable. (Reuters)

• Merger enforcement stayed unusually quiet: Dechert’s DAMITT 2025 report notes only 11 significant U.S. merger investigations concluded in 2025, near historic lows—meaning fewer headline fights, more “design the deal to survive review.” (Dechert)

That’s all for this week. See you next week for more & while you wait, please share:

Share