Everyone and their mother wants to know where margins will accrue and how to find the diamond in the rough that will capture a majority; as someone in the venture seat I sort of have to want that too. What I’m really interested in though is the physical underneath. Partly because that’s where the bottlenecks are and partly because of the choices we can make at this layer - it’s no longer just “one thing”. For the better part of a decade, “AI Compute” meant the usual suspects: an NVDA GPU, a CPU, the NVLink and InfiniBand fabric, and the memory stack. The GPU was the organizing function of the industry and it served very, very well. I feel like that is less true with every passing month. Over this period, we’ve paid a lot of attention to the dollars per token figures and ignored the joules per token. This is largely because the storyline was quite predictable: chips improved and so did models, serving stacks squeezed more throughput, and prices fell. But it’s not as predictable anymore; there is a decoupling in the two units and the gap is widening.

I’ll state the obvious up front: inference is becoming harder to manage. The queries are getting more varied; between a 100k token prefill, a reasoning trace, and an agentic query, you get three vastly different energy signatures. At the same time, the general-purpose GPU chip we have grown used to has run out of levers to improve at its previous pace. This new reality is characterized by scarce watts, increasingly complex workloads, and silicon that might be falling behind. It’s a confusing landscape but it presents a set of opportunities scattered across the stack.

How the bill got strange

In 2023, we all thought that intelligence would be free. After GPT-4 came out at $30/M input tokens, by the time you had 4o mini, Claude Haiku, and a whole host of open-weight models in less than a year or two, you could basically get the same performance for your average use case. And because we assumed it would keep falling, we made everything “AI-enabled”, “AI-native”, “AI-powered” - you’d integrate OpenAI and Anthropic APIs wherever possible and make that an integral part of your business, whether it was customer service, recruiting, sales, etc. Granted, prices have come down, but only at the unit level. During that time, the floor went lower and the ceiling of costs went higher. GPT‑5.6 Luna is very competitive with Sol, and it’s a lot cheaper per task. You can set the “thinking power” for these frontier models, ranging from low to high to max (using more tokens and more computation). Naturally, that changes the spend. For instance, Claude Opus 5 uses about 12 million output tokens on the Intelligence Index at the lowest setting, and about 100 million at max effort.

We don’t really think about where the costs of this show up because the labs abstract away all of the infrastructural complexity. But if you’re running an open source model like DeepSeek‑V‑whatever, the expenses are easier to spot because they directly impact your electricity bill, especially since they recently increased their prices dramatically. I get heated whenever PG&E charges me more than $100, so I’m lucky I’m not running any of these models extensively.

The early AI-pilled pilots mostly lived near the floor, so the bills were always trivial. But as features scaled into production, tokenmaxxing quickly went out of style. We saw this widely in industry: Uber reportedly burned through its entire 2026 AI‑coding budget by April and ended up capping individual engineer spend at $1500; Coinbase went the other way and cut its AI bill roughly in half by routing engineers through an internal LLM gateway that defaults to cheaper open‑weight models like Kimi K2 and GLM; Databricks experimented with OSS and smart routing for models as well as harnesses, leading to savings across the board. There are many efforts to find solutions to balance the growing need as well as growing costs of compute.

The traffic is new

This is a much older problem: the real energy cost has shifted from computation to the movement of data. Picture a very fast worker at the end of a very slow conveyor belt. If the belt is the constraint, hiring a faster worker changes nothing; inference is limited by the belt and not the worker. (Yet another inference analogy for you. Moving on.)

Shifting data costs significantly more energy than the math you do. Here are a few numbers to help with framing: in the TPUv4i paper, Google’s team found that a single FP32 multiple equates to roughly 1.31 picojoules but fetching one 64-bit value from DDR costs about 1,300 picojoules. That’s a 1000x difference. A 70B parameter model at 16-bit weights is roughly 140 GB and most of that has to be dragged back and forth across the memory bus for every single token generated.

None of this is new, since the memory wall has been a central pillar of computer architecture for many decades. However, the likes of interactive decoding at batch sizes near one, reasoning traces, agentic loops, and post-training runs, are new patterns of traffic we have not seen before. For a concrete example, take agentic RL. One simple task can become many long tool-using trajectories, which are then analysed, rescored and reused as training signals instead of being consumed just once. Tokens, thus, grow across the number of tasks, rollouts, interaction turns, context length, etc. These tokens, though, are valuable because they turn all of this exploration and failure into learned skills.

These are new patterns that have changed the shape of both computation and memory access, making the memory wall feel even taller and the GPU even more inefficient.

Quick math

Doing some back of the napkin math, here’s one way to think about the attribution of joules per token: how much of the model “wakes up”, how far the weights need to travel, number of trips needed, the precision, the arithmetic itself, and how busy the machine is throughout (switched on vs doing work). That last term is increasingly important in an agentic setting. This is a crude way to think about what actually affects the bill; each of these terms is nuanced. For instance, MoE model wakes up only a handful of parameters while a dense model is fully awake. Weights might come from DRAM, HBM, SRAM or bypassing that entirely if they’ve been etched straight into the metal.

This was largely intuition, so I ran it against the ML.ENERGY numbers in Chung et al.’s “Where Do the Joules Go?”, to see whether it made any sense. It did but I wouldn’t over generalize here because of the setup. Batching resulted in a 3-5x lower median energy per token between a small and saturated batch, while the type of task drove the biggest swing at 25x per query. Everything else was messier and more hardware-dependent. I’ll link it for anyone who wants to play around. It makes sense that efficiency gets dramatically better when enough work shows up together to spread this weight haul around, but real products are not so cooperative. This is simply about the numerator vs the denominator. A voice agent can’t wait after saying “Hi”, now can it?

The choices are also new

While the demand story is dominated by volume, I find the shape more interesting. Between early 2024 and late 2025, we saw average sequence length rise 3x on OpenRouter. There’s a lot that changed under the hood, with queries having vastly different physical characterizations. Kim et al.’s 2026 paper, “The Cost of Dynamic Reasoning,” measured this energy consumption per query across agent frameworks. My prior was that they would diverge but not this much - their baseline single-turn query consumed 2.55 watt-hours, a LATS agent jumped to 158 wh, while with a Reflexion agent, it jumped even higher to 348 wh (perhaps easier to think of this in hairdryer-terms). While these agents ran, the GPU sat idle for up to 54.5% of execution time waiting on tools, but drawing static power throughout. This static power draw, usually around 10% of TDP, comes largely from power supply maintenance, HBM idle refreshes, and comms link maintenance. You’d expect the power trace to look more sine wave-y (a drawdown while it thinks and back to nothing while it waits, think, wait, think, wait) but IRL it’s closer to a much flatter line. So, it looks like the software knows that the agent is waiting but no one told the power system? This idle stretch is actively being worked on. For example, the serving system can multiplex other agents’ work onto the same resident weights to serve more than one trajectory and background agents could be ideal to slot into that gap. You could also push the paused context off the expensive silicon to reduce energy intensity. All of this comes with tradeoffs though, between HBM capacity and utilization, among others. There is also a conversation here around latency, which I’ll save for now.

I briefly mentioned this earlier but these energy signature results map cleanly to reasoning as well. Using the same dataset, Qwen 3 8B goes from 0.010 watt‑hours per query on conversation to 0.388 on a graduate‑level problem solving exercise (a 40× jump) and DeepSeek V3.1 energy usage jumps 23.5x.

That breaks the napkin math a little bit. While most terms are “fixed” by the hardware, model or even serving system, test-time compute (and similar paradigms) is not. The same model can be told to think a little or to think a lot and more often than not, we do this without thinking about the impact on the infrastructure below. I’m going to call this the depth dial for now. This is not the only dial - you can turn the breadth one (sampling and selecting) or an iteration one (agent loops, self correction) and spend tokens that way, too. These stack multiplicatively. So who really controls these dials? It could be the user (only for depth, though less and less so), the provider using some reasoning effort / thinking effort default, or increasingly so, the agent framework. I find the concept of agents self-administering token budgets pretty interesting but that’s a separate post.

Betting on next Tuesday?

A chip is a permanent decision that is made years in advance (although these cycles are compressing now). Since compute and memory bandwidth fight over the same die area, the same power budget, even the same reticle limit and package, every chip designer is effectively designing the hardware for specific tasks before they are well understood or even formed. The workload it has to serve could change weekly. A chip is a three-to-four-year commitment: taped out on one process node, fabbed and then sold, and then depreciated over time. So really, what we’re doing is asking a fixed architectural point to serve a distribution that could change next Tuesday. (I don’t know how better to visualise this than a Druddigon using double-team on Dragonite.)

Joules limit adoption

We spend a lot of time talking about power at a macro level, focused on the grid and interconnection. I agree that that is very important, but it’s important to zoom in, too. While our computational demand on one side of the envelope is becoming more and more elastic, the power supply on the other remains fixed. This makes a wasted joule all the more important, even determining how much intelligence we can fit into a given system. A drone, for instance, can house only as much compute as its battery will allow. Sure, I can tack on a more powerful model onboard but then its flight time would be shortened. What’s the point if my drone knows it has to somersault next when its flight time is 45 seconds? (That’s supposed to be a joke.) Even a small business can either have an AI workstation or not depending on if it can be powered by the 15-amp circuit on their wall. This would be true even if inference was static and the energy signatures didn’t vary as wildly, but it’s even more important now when we keep creating new ways to spend the joules.

The crux of this is how do we make every joule do more? People are betting on very different answers to lower the joules per useful token, and I wish they were as simple as the question. In part II, we’ll go over some of those bets and where pockets of opportunity might lie.

Inside the data center, the challenges are just as severe: a single pod now approaches the power consumption of a small city, and large facilities contain enough optical fiber to wrap around the world five times. As a result, energy and operational complexity have become key design constraints, instead of afterthoughts. The explosive growth of LLMs and their infrastructure has made physics the bottleneck. At Congruent, we think the solutions will come from a stack that’s energy-aware from the ground up. Historically, power was treated as a utility - you’d design the compute, then provision enough electricity. The new stack inverts this structure. Compute systems are starting to understand power constraints, and power systems are starting to understand compute workloads. That bidirectional awareness is the thread tying together what might otherwise look like scattered opportunities. That’s where we’re focused: orchestration software that coordinates across power, cooling, and workloads; tools that compress the years-long path to grid interconnection; and step-function hardware bets in memory, photonics, and power electronics. If you’re building in any of these areas, we’d love to talk.

What makes this moment unique is that none of these problems are waiting their turn. Physical limits in power distribution, memory bandwidth, and grid stability are all hitting at once. Moore’s Law economics are, to some extent, breaking down precisely when AI demand is accelerating exponentially, which means there’s never been more urgency, or frankly more opportunity, for architectural innovation. Breakthrough innovations at any single layer - particularly in memory architecture - would have sweeping effects on their own. However, because these constraints are interdependent, solutions that work across layers can compound in ways that point solutions can’t. The entire stack is evolving together, and this interdependence creates opportunities for coordinated innovation beyond what point solutions can achieve.

In this post, we’ll cover the three primary data center constraints, what the energy-aware stack looks like, and where new venture opportunities may emerge.

Three Fundamental Problems & Emerging Solutions

Three fundamental constraints are colliding simultaneously, driving innovation across the entire compute stack.

1. Model size outpacing hardware (The Memory Wall)

Model parameters are growing exponentially while memory capacity and bandwidth improve linearly. The widening gap forces tight coupling of GPUs - models require low-latency memory access, which means compute must be densely packed. Training workloads are often memory-bound, leaving expensive GPUs underutilized. The result is monolithic racks that are a nightmare to cool and operate.

New memory and compute architectures represent the largest opportunity for breakthrough innovation, though the path ahead remains uncertain.

A range of approaches are being explored - vertical 3D integration, compute-in-memory, compute-in-network architectures - all of which place memory closer to compute. Memory specialization by workload type, analogous to today’s specialized accelerators, might dramatically improve efficiency for specific tasks, though it’s early. Groq’s LPU architecture, which uses on-chip SRAM to sidestep HBM bandwidth bottlenecks for inference, is one example of this approach attracting significant capital. If these approaches work, they wouldn’t just improve performance, they’d simplify compute architecture and substantially reduce energy consumption. We’re watching this space closely, as is everyone else in the industry, because step-function improvements here would change everything downstream.

Photonic interconnects offer a clearer path forward. As bandwidth scales, they deliver the highest performance while addressing several key challenges. On the packaging front, they move more bits through fewer, more manageable cables and connectors. They also promise lower energy per bit. And because optical signals can travel longer distances with minimal loss and latency, they enable more flexible rack layouts within data centers and more flexible data center architectures across buildings or regions.

Here’s where the interdependence becomes visible. Photonic disaggregation enables smaller distributed data centers mainly for inference workloads, which are individually less demanding on local power grids and can be deployed more quickly, with greater flexibility in site selection and power procurement.

2. Surging power needs at the rack level and unsustainable current densities

Memory bandwidth requirements force dense coupling of GPUs, which has power implications that are hard to overstate. GPU power draw has increased from around 300W to over 1000W in just a few years, and NVIDIA’s upcoming Rubin platform will push entire racks to 600kW - the power consumption of hundreds of homes, concentrated in a single cabinet. At these densities, the physics of low-voltage distribution starts to break down. Resistive losses scale with the square of current, copper busbars become prohibitively massive, and you quickly find that floor space is being consumed by power infrastructure rather than compute.

The industry is responding by moving to high-voltage DC distribution (±400V, 800V, with 1500V under consideration) which dramatically changes the local power distribution architecture. A 600 kW rack at 48V requires over 12,000 amps; at 800V, just 750 amps. That difference shows up in smaller copper busbars, less space consumed by power infrastructure, and more room for actual compute. Superconducting transmission, which companies like Congruent’s portfolio company VEIR are commercializing, could push efficiency even further. That said, HVDC creates its own demands. Efficient power conversion from medium voltage down to chip-level voltages may require semiconductor technologies and circuit topologies that are still being developed, and safety standards for HVDC protection haven’t fully caught up yet. Liquid cooling is a similar story - it enables higher densities, but demands tight integration with power delivery and rack design, which is why companies like portfolio company Alloy Enterprises are focused specifically on the thermal management piece. Beyond the rack, microgrids and solid-state transformers (the likes of HeronPower, Amperesand, Blixt, and unannounced incumbent vendor efforts, among others) provide smart grid connection and bidirectional power management.

What ties all this together is that memory growth drives the need for architectural density, and power delivery is what makes that density physically possible. Neither problem is being solved in isolation, which is another reason why we find this space so interesting.

3. Grid compliance and load volatility

The challenge here operates at two levels: securing interconnection, then operating as a responsible grid participant once connected.

As we’ve discussed in previous posts, securing grid interconnection takes 5-8+ years and has become a primary bottleneck for data center development. These timeframes can be dramatically shortened to [< 1 yr] by using flexible interconnections that maximize existing grid capacity through real-time management and data center load flexibility. Software platforms like portfolio company Camus Energy are using this approach to reduce time-to-power by working with grid constraints rather than against them.

Once connected, AI workloads create unprecedented volatility. Training and inference alternate between compute-intensive phases (high power) and communication-intensive phases (lower power), with fluctuations of hundreds of megawatts on millisecond timescales. For grid operators accustomed to slow, predictable load changes, this is unfamiliar territory and not contemplated in the controls architecture in legacy grid systems. Massive, rapid load swings threaten voltage and frequency stability across entire regional grids.

Massive power fluctuations can lead to grid instability, as nodes simultaneously transition between compute-intensive (high power) and communication-intensive (low power) phases.

The operational challenges that follow are significant. First, the data center must smooth load profiles to prevent grid instability. Smart energy management addresses this: workload schedulers that understand which jobs can pause run in the background, energy storage that buffers fluctuations, and on-site generation for baseload support. Solutions like Emerald AI and Phaidra actively manage consumption patterns while maximizing compute utilization. Second, data centers must respond appropriately when grid events occur. Grid operators increasingly expect large loads to provide services (or at the least stability), not just consume power passively. At the application layer, schedulers must distinguish which training runs can pause during grid stress from which inference services require guaranteed power for latency SLAs. These signals propagate down through compute hardware to facility-level power management. Solid-state transformers function as intelligent power routers, dynamically balancing between grid, solar, and storage inputs to optimize cost and availability. Virtual power plant platforms like portfolio company Leap can rapidly deliver capacity from existing grid assets while also enabling data centers to participate - effectively turning them into distributed grid resources, which is a significant shift in how we think about what a data center represents as a grid asset.

This kind of coordination doesn’t exist today. It requires integration across what are currently vendor-balkanized layers.

The Energy Aware Tech Stack

The sheer scale and complexity of AI data centers requires both best-in-class performance at each layer and energy-aware coordination across the stack to deliver:

• Faster deployment time via accelerated power access (more flexible siting of smaller data-centers regionally clustered and flexible interconnect)

• Stable grid operation through full-stack energy management to stabilize data center loads and comply with stringent grid requirements

• Higher compute utilization via smart workload orchestration that minimizes curtailment

• Lower operating costs and lower carbon footprint through coordinated energy management

Individual elements provide value independently but amplify when operating as a system. HVDC distribution enables higher rack densities and provides the foundation for microgrids, energy storage, and solid-state transformers. When you combine those with workload scheduling and compute controllers, you can smooth data center load for grid stability, provide increased capacity when energy is available, and shed loads intelligently when curtailed.

Photonic interconnects provide the lowest power and fastest connections, and they simplify deployment by breaking up mega racks for easier siting. Absent algorithmic breakthroughs, training still favors dense, co-located clusters, but inference workloads are different - they can leverage optical connections to create logical mega data centers across distributed physical sites, which is where the deployment flexibility of photonics really pays off.

New Venture Opportunities

With this foundation in place, new optimization surfaces open up. Some opportunities are structural: they emerge because power, thermals, and grid constraints are now first-class design variables rather than their historical status as afterthoughts. Others are paradigmatic: step-function breakthroughs that can rewrite the constraints entirely. Our core thesis is that, regardless of where breakthroughs land, the stack is becoming energy-aware. Workloads will increasingly express flexibility and requirements - whether a job can pause, what latency it requires - and infrastructure will increasingly translate those intents into decisions about compute, power, thermals, and grid participation.

We see opportunities falling into two broad categories, though we’re certainly not capturing everything here.

Platform and infrastructure plays

These are companies building on the new constraints, not trying to eliminate them, but turning them into product advantages. They tend to have clearer paths to revenue and more predictable technical risk:

• Platform orchestration and energy management is where we’re spending a lot of time. The facility-level platform that coordinates power distribution, energy storage, cooling, workload scheduling, and grid interaction. Someone is going to own this layer, and it’s unlikely that it is a single hardware vendor. Power components may commoditize, but the orchestration layer creates platform value with strong network effects, particularly systems providing bidirectional information exchange across applications, schedulers, hardware, facilities, and grid. We’re looking for platforms that can actuate, not just monitor, while working with existing hardware systems.

• Workload-aware inference routing is the application-layer counterpart to facility-level orchestration - dynamic routing of inference requests across distributed infrastructure based on latency SLAs, energy costs, and real-time capacity. This feels inevitable to us; the question is how it integrates with the layers below and whether there is room for an independent company to own this functionality, or whether it will be insourced.

• Accelerating deployment is perhaps the most immediately actionable category. Tools that turn data center load flexibility into faster grid interconnection, or that enable operators to site facilities based on real grid capacity rather than nameplate ratings. The unlock here is regulatory as much as it is technical, and is key to bringing data centers online in a reasonable timeframe.

Step-function bets

These are higher-variance opportunities where success could rewrite the constraints entirely. They’re harder to underwrite as a venture investor, but the market scale justifies taking thoughtful swings.

• Novel memory architectures represent the biggest prize. High-throughput memory, application-specific memory architectures, compute-in-memory designs that eliminate the bandwidth bottleneck rather than engineering around it. A true breakthrough here simplifies everything downstream, which is why everyone is chasing it and why we remain cautious about picking winners too early. The scale, capital needs, and expertise to address the memory challenge are not for the faint of heart.

• Alternative computational substrates: Thermodynamic computing, stochastic architectures, photonic compute - approaches that could achieve order-of-magnitude improvements in energy per operation for specific workloads. These are longer shots but the payoff would be enormous if any of them pan out at scale.

• Photonic integration: Copackaged optics offers near-term power and latency gains by moving optical components onto the package. The material platform question - silicon photonics vs. InP vs. hybrid approaches - is still being contested;our sense is that the answer will vary by application, which may mean room for multiple winners.

• Advanced cooling and power electronics: Immersion cooling, two-phase systems, and novel power conversion topologies. Current approaches have hard limits on rack density and efficiency. Innovations here could break through those limits in ways that unlock value across the rest of the stack.

We are energized by the challenge here: to maximize intelligence and outcomes delivered per watt. If you’re working on any of this, we’d love to hear from you.

As we head into 2026, we’re excited to share some of the themes we’re discussing around the Congruent office. We’re actively investing Fund III, are halfway deployed in our first growth fund, and are spending time supporting our 54 companies and founders. From fusion and geothermal, to climate risk software for real estate and finance, to the systems that run factories and move goods faster and cheaper—the portfolio is executing (check out some highlights at bottom!)

While the list below is by no means an exhaustive one, if you’re building in or around these themes, we want to hear from you!

Data Centers (for a change…)

Observation: In 2025, data centers dominated the energy conversation. In 2026, the narrative will shift from headline gigawatts to execution: construction, commissioning, interconnection, and the hard realities of cost, timelines, and reliability. It won’t be pretty.

Opportunity: We crossed the point where inference energy >> training energy last year. What’s next? Specialized inference silicon, new approaches to managing dynamic loads, and AI-native compute architectures that break from today’s memory and topology constraints? What are the next technologies that drive step-change improvements, not just incremental efficiency?

Resilience

Observation: Climate-change-exacerbated wildfire, storms, flooding, and heat are attacking critical infrastructure. Incremental fixes won’t deliver the cost or reliability the future grid needs.

Opportunity: Beyond inspection and monitoring – with more limited spend on hardening – how do we bridge to the missing piece: resilience by design? What technologies can prevent overhead line ignition, harden water and sewage systems against megastorms, and manage saline intrusion into aquifers?

Power Costs

Observation: Electrification is cutting GHGs and local pollution, but power prices are rising (much) faster than inflation. Electrification needs to be the cheaper option or there will be no adoption.

Opportunity: The marginal cost of generation isn’t the bottleneck—how do we focus more innovation into transmission, distribution, and new line capacity with technology innovation? What approaches can increase grid utilization without sacrificing reliability? How can grid operators improve load coordination, smarter siting and development while integrating modern grid ops? Will the increasing need to manage costs drive utility adoption of existing and new approaches?

Insurance

Observation: Climate risk is making large swaths of residential and commercial property effectively uninsurable. This is spilling into state (and taxpayer) budgets, insurer balance sheets, and property values. Regulations that force insurers to price off of historical data rather than forecast risk is amplifying the problem.

Opportunity: We see two primary approaches to address this challenge. First, new insurance products: what parametric, risk-shaping, or other tools let carriers re-enter constrained markets sustainably? Second, what new technologies can reduce loss severity through property hardening or infrastructure improvements, lower total claims and make coverage viable (and affordable) again?

A few 2026 lightning round predictions:

• The first startup fusion reactor will demonstrate net gain by analysis (Q>1 if tritium is used in place of Deuterium).

• Geothermal becomes a credible fourth zero carbon option alongside solar, nuclear, and wind.

• AI robotics, humanoid and specialized, deploy at scale with immediate cost and energy benefits.

• Data center power density keeps climbing; 1MW racks enter the conversation. That’s like a 10-story office building in a 72” rack.

• Voluntary carbon markets professionalize further, with at least a quarter of durable CRUs trading on exchanges.

• The EU and IMO hold firm on air and sea decarbonization but introduce new subsidies to ease implementation.

ICYMI

Some highlights from last year:

• Both Pano and Muon Space were featured in TIME’s Best Inventions of 2025 and both Fervo Energy and Pano were named to TIME’s 100 Most Influential Companies of 2025

• Parallel’s autonomous freight train was featured in the LA Times

• Fervo Energy was the spotlight of this WSJ piece on the One Renewable Energy Source Loved by Both Parties

• AMP Sortation was recognized for its >10 megaton contract to invest in SPSA’s next-generation single stream waste sortation facility.

• Bonsai’s robots got featured in this Forbes piece on AI-First Autonomy For Agriculture

• The Economist featured Avalanche in its How to build table-top fusion reactors piece, as well as its 2025 Fusion Fest. Avalanche also released their mini-documentary: The Sun in Our Hands.

Thanks for staying close. Now back to work!

—The Congruent Ventures Team

As the AI arms race escalates across the geopolitical and technology spheres, headlines gush over OpenAI, Microsoft, Google, Meta, Anthropic, Perplexity, xAI, and Chinese disruptors, and they are right to do so. But the existential threat to American AI dominance is not the latest AI model weights, but access to electrons. In an environment where new infrastructure is characterized by decades of planning, power needs for compute needs are immediate and growing.

To unpack this challenge Congruent recently hosted a panel as part of San Francisco Climate Week featuring Karan Mistry from Boston Consulting Group, Bobby Hollis from Microsoft, and David Ulrey from Fervo Energy. There was consensus that AI compute is driving unprecedented demand for electricity and macroeconomic risks may affect the pace of the data center buildout. We believe that AI compute and the necessary data center infrastructure are the single largest energy investment opportunity in recent years. If approached correctly, the AI data center buildout has the potential to enhance the overall power grid and accelerate our transition to a cleaner energy future while serving the needs of the tech community.

As climate investors, we are seeing the impact of AI firsthand across our portfolio, with multiple companies capitalizing on AI to make the world more efficient and sustainable. These companies include First Street using AI to assess physical climate risk, Fervo leveraging AI compute to unlock vast new resources of clean geothermal power (including for use in Google’s data centers), Pano.AI for AI enabled early detection of wildfires, Camus Energy with AI-based software to accelerate interconnection for data centers and other large loads, and Halcyon AI, which uses large language AI models for better energy-related decision making from vast amounts of published regulatory data.

As we scope the urgency of powering data centers, we have distilled our questions to the following:

1) What is the expected trajectory and timeline of AI data center power demand growth?

2) How is this growth creating new opportunities in energy and climate?

3) Can we scale technologies that deliver strong financial returns and climate benefits – while keeping pace with AI compute demand?

Innovation is in full force with novel business models, software optimizers, and new hardware solutions. If you’re innovating in these areas with a venture-appropriate model, we would love to hear from you.

Framing the AI Data Center Opportunity – and Challenges

According to JPMorgan, the major AI hyperscalers (Alphabet, Amazon, Meta, and Microsoft) spent $199B in capex in 2024, which is more than 4x growth over the past 5 years and estimated to further grow by 34% to $268B this year.1 Remarkably, these four tech companies, which have historically been lauded by investors for their high margins and free cash flow, will spend more on capex than the four largest global mining companies and the four largest global oil and gas companies combined. Market intelligence firm International Data Corporation (IDC) projects that AI-enabled solutions will yield a global cumulative economic impact of $22.3 trillion by 2030, representing approximately 3.7% of the global GDP2. Given the traditionally fierce competition between the hyperscalers and an increasingly multi-polar world where nations are seeking self-sufficiency and resilience, a compute-driven arms race where parties vie to match each other’s capabilities is well underway.

Source: McKinsey and Company

Mobilizing such massive spending on technology infrastructure has precedent. The private sector has invested approximately $2.2T in US broadband infrastructure since 1996.3 In spite of initial overbuilding between 1999 and 2001, investment in additional capacity has been fairly consistent at $70B/yr over the past 20 years with growth accelerating recently to $100B/yr. Incredibly, steady technology advancements have increased throughput of a given fiber bundle by more than 645x since the early 2000s!4 This is because applications driven by data and compute have followed three powerful and mutually reinforcing trends:

· Moore’s Law: The number of transistors that can be packed into an integrated circuit doubles approximately every two years (from Gordon Moore, co-founder of Fairchild and Intel).

· Jevons Paradox: The total consumption of some resources continues to increase as they become more affordable and widespread (from the 1865 book The Coal Question, in which William Jevons observed that the introduction of more efficient steam engines led to more coal demand, not less).

· Metcalfe’s Law: The value of a communications network increases with the square of the number of connected users (from Bob Metcalfe, founder of 3Com and inventor of the Ethernet).

Is AI compute the modern equivalent of the steam engine? As compute power augments and substitutes for cognitive labor, the demand for energy may be virtually limitless. As climate-focused investors, this is the opportunity for us to explore.

Implications for Energy Demand and Supply

Training and running AI models requires specialized graphics processing units (GPUs) crunching billions to trillions of parameters in ways that are far more energy-intensive than the traditional data retrieval, streaming, and communications applications that drove data center growth over the past two decades. At 2.9 watt-hours per ChatGPT request, AI queries are estimated to require 10x the electricity of traditional Google queries.5

Running AI models is faster and more efficient in large data centers, and the larger the data center, the faster and more efficient it is. This has led to dramatic growth in the size of hyperscale data centers, which are 300-1,000MW in capacity vs traditional cloud and edge data centers which were typically below 100MW. A modern hyperscale data center consumes the same amount of power that would be required for 240,000-810,000 average US homes6. Based on publicly available information and expert forecasts, the Electric Power Research Institute (EPRI) recently projected US data center load requirements and growth through 2030:

Source: EPRI

In power terms, EPRI expects data center load growth from approximately 17GW in 2023 to a high case of 57GW in 2030. For context, the US consumed 4 TWh in 2023 (approximately 460GW of average hourly demand) and has seen little to no aggregate load growth over the past two decades. Wood Mackenzie analyzed the combined effects of data centers, domestic manufacturing, electrification of transport, and heating, and projects US electricity demand to increase between 4% and 15% through 2029, with likely additions able to accommodate growth of just 2% per year.7

Why the 2-13% gap in expected demand vs. supply? Although projected growth may seem modest in percentage terms, it presents enormous challenges for the power sector. First, 80% of today’s data center demand is concentrated in just 15 states, with particularly high concentrations in Texas and the Mid-Atlantic. Data centers are sited based on favorable characteristics across a multitude of factors such as existing data connectivity, land and power costs, availability of skilled labor, and reliability and availability of primary and backup power. Future data center growth will likely continue to be concentrated in favorable geographies. Second, AI data center loads are individually massive, requiring costly and time consuming high-voltage interconnection if supplied from the grid or the build out of substantial co-located generation. Third, data centers are critical engines of our modern economy and must function uninterrupted, 24/7 every day of the year, particularly if they are customer facing. In our discussions with hyperscalers and data center developers, four needs overlap and compete as shown in the chart above with “X” marking the ideal power source.

Speed: In the arms race to keep up with the competition in AI computing, speed is on par with reliability for power sourcing criteria. Given the scale, cost and complexity of developing and constructing AI data centers, our understanding from speaking with industry stakeholders is that hyperscalers are currently talking with project developers today for projects that will come online in the 2029 timeframe. This schedule is constrained by the timelines for a 500MW AI compute facility, an example of which is shown below.

As outlined in Congruent’s recent thought piece Uncorking the Interconnection Bottleneck, the median time from requesting new generation grid interconnection to operations was 5 years for projects built in 2023, and this continues to lengthen. In 2022, eastern grid operator PJM froze all new requests until at least 2026. Integrated resource and capacity planning is a slow, back and forth exercise between utilities and regulatory commissioners that seeks to ensure all utility customers (“rate payers”) are treated fairly and that system costs are both manageable and spread equitably across the rate base. The job of utilities and their regulators is further complicated by the uncertainty in AI market growth and the speed at which customers are currently requesting interconnection, making interconnection the key bottleneck for grid interconnected data centers.

Reliability: Data centers are the foundation of the modern economy, from retail commerce to financial transactions to critical communications. The need for high reliability does not appear to be changing for AI compute8 and hyperscalers are seeking 99.99-99.999% uptime from their power source; “five nines” equates to just 5.26 minutes of unplanned down time per year. The reputational and consequential damages of downtime are significant. Incidents with Amazon and Meta in 2021 resulted in revenue losses of over $500k and $200K per minute respectively for those organizations.9^,10 EPRI reports that 83% of responding utilities had requests from data center customers for higher reliability than the utility’s typical large industrial customers.11

Cost: While speed to power and reliability are paramount, cost is still important. Stakeholders report willingness to pay up to $125/MWh for firm, clean power that can be brought online in the next several years in desirable geographies. This is almost double the cost of wholesale gas-generated power and several times the cost of intermittent wind/solar power that would be available to these customers in the same geographies and timeframe. JPMorgan estimates that AI data center provider CoreWeave generates $809M of revenue annually per 100MW of compute power at ~70% EBITDA margins. This equates to EBITDA of ~$650 per MWh of energy consumed. With industrial electricity costing $80/MWh on average across the US12, energy is a meaningful operating cost, but not a primary driver of CoreWeave’s overall profitability.

Carbon Intensity: While hyperscalers continue to voice their commitments to reducing and eventually eliminating their carbon emissions, industry stakeholders sheepishly report that carbon intensity is taking a back seat to the three other criteria identified above. Hyperscalers recognize that data centers are long-lived assets and unmitigated carbon emissions introduce financial and customer risks for the massive investments they are making into data centers. Companies are pursuing a variety of strategies such as Microsoft’s plans to work with Constellation Energy in restarting the nuclear facility formerly known as Three Mile Island in Pennsylvania, or Meta’s plans to site with Entergy Louisiana using new gas power paired with renewable power generation. EPRI’s utility survey reported that 57% of utility respondents had requests for 24/7 carbon free energy from data center customers.13

Solutions

Hyperscalers are among the largest and best capitalized companies in the world, with the ability to invest and attract large financing partners to their projects. With concentrated loads of 300-1,000MW, requirements for high reliability and a desire to minimize carbon emissions, the opportunity is compelling across many fronts for solution providers. Although large-scale transmission could theoretically enable data center load growth by better leveraging existing power generation capacity, only 55 miles of high-capacity transmission were built in 202314. The speed of anticipated AI data center deployment calls for new solutions and different ways of thinking. As we consider solutions, we focus on the timeline to technology maturity, and the design and build lifecycle of a data center. Hyperscale data centers costing $10/W to build equate to massive projects that are $5B-10B in scale, so the cycle of integrating new solutions must be started years in advance of construction. We view solutions to quickly and effectively power data centers across a broad spectrum, from initial site selection to operational solutions, and a diverse opportunity set, from innovative business models aligning project stakeholders to new technologies for power generation

Exciting solutions such as small modular nuclear reactors can be game-changers but are likely 5+ years out from first commercial units even for the most advanced startups. These will compete with scaled carbon capture solutions and conventional nuclear generation that may be lower cost. Local power generation with modular gas turbines and engines are much less efficient than conventional generation, but are filling the gap because they can be deployed quickly. These deployments have attracted significant local and environmental opposition for allegedly violating environmental and permitting regulations15. Retrofitting such generation with carbon capture solutions could be a future opportunity across many sites.

xAI’s data center in South Memphis Tennessee with 35 engines generating 422MW:

Photograph: Steve Jones/Flight by Southwings for Southern Environmental Law Center

Increasing the efficiency of the data center itself could mitigate the need for lengthy and costly power grid enhancements. While compute is the single largest load in the data center, cooling is responsible for 30-40% of total data center load. Power use efficiency (PUE) is an industry-wide metric that divides the total energy used by a data center by the energy consumed by the IT equipment alone. While PUE has plateaued at approximately 1.6 over the past decade, leading hyperscalers report a 1.1 PUE16 – indicating potential for industry-wide advances in efficiency

Source: Uptime Institute Global Data Center Survey 2024

Source: ERPI

Flexible interconnect combined with software approaches (like Camus Energy) that can identify and better leverage existing and underutilized infrastructure such as substations, transmission capacity, gas and renewable power assets are appealing since they can compress time and cost. Duke University recently published a study showing that 76 GW of new load—equivalent to 10% of the nation’s current aggregate peak demand— could be integrated with an average annual load curtailment rate of 0.25%17 .We are intrigued by business models that bring together hyperscalers, utilities, and traditional power developers aiming to create win-win partnerships integrating existing infrastructure.

Congruent’s Bull and Bear Perspectives

There are many reasons to believe that AI compute is a generational opportunity and that approaches to powering AI data centers could make for a compelling investment. But there are also uncertainties, complex regulatory landscapes and diverse stakeholder needs to balance. We summarize our Bear and Bull views on this market:

Enablers (Bull Case)

1. Size of opportunity. It’s rare to see so much capital mobilized so quickly on what is fundamentally energy infrastructure, and the hyperscalers are among the largest and best capitalized customers in the world. If approached correctly, the data center buildout has the potential to enhance the power grid and accelerate the deployment of cleaner, advanced energy technologies.

2. Breadth of the opportunity. Related to size is the breadth of approaches and solutions that can be brought to bear on the solution. Innovative business models, software optimizations, and hardware can attack different areas of the problem individually and in concert.

3. Opportunities for partnership. AI compute has attracted the best and brightest from across the globe. Entrepreneurs, large companies, utilities, government and regulators are keen to catalyze solutions and create opportunities. No previous sector has valued speed to power this highly.

Obstacles (Bear Case)

1. Unclear sector growth. Some argue that AI infrastructure is being overbuilt for AI's near-term, commercial applications. If the hyperscalers are unable to monetize their compute investments, growth will pause.

2. Compute efficiency. In January 2025, a Chinese company gained attention for releasing a dramatically more compute-efficient Large Language Model (LLM) called DeepSeek. While the reported results were not as remarkable as initially perceived by the financial markets, improved compute algorithms and efficiency may short circuit the ravenous demand for unbridled power into AI compute.

3. Macro Uncertainty and Tariffs. AI data centers are very costly and have inputs ranging from steel to semiconductors, and concrete to electrical connectors. Construction relies upon broad global supply chains and tariffs, inflation, or fluctuating trade policies create uncertainty that will hinder capital investment into largescale data center infrastructure.

Final Thoughts: What We’re Looking For

We love bold ideas attacking billion-dollar problems and this area seems ripe with opportunity. We have been energized by the conversations we’ve had with entrepreneurs, startups and other stakeholders seeking innovative solutions to quickly and cleanly power AI data centers. Here are a few ideas to get the conversation started:

· Flexible compute and rapid grid response solutions to power management

· Advanced microgrid solutions coordinating diverse behind the meter resources

· More efficient power conversion from high voltage AC to low voltage DC

· Cost effective, modular point source carbon capture

If you are a tech innovator, power developer, entrepreneur, customer or potential corporate partner we would love to collaborate!

Our TL;DR

Grid interconnection is not a new challenge, but the recent surge in demand – initially driven by renewable generation growth but increasingly by disproportionately large loads like AI data centers – has brought the issue to the fore. Interconnection’s central role in transmission development makes addressing its bottlenecks paramount for renewable energy deployment and grid expansion.

We’ve spoke with entrepreneurs, investors, regulators, project developers, policy experts, and technical subject matter experts to evaluate the opportunity for software to streamline the interconnection process. Based on these conversations and our research, here’s what we’re looking for:

1. Platform Functionality: Interconnection tools alone are unlikely to produce a venture outcome, but there is a much larger market opportunity for startups that can deliver value across operations. For example, integrating the interconnection process with longer-term transmission planning and process management may provide sufficient platform functionality for scale.

2. Ease of Integration & Customization: Seamless integration with existing utility tools and workflows is paramount, as are a high degree of flexibility and customization for developer-facing applications. Sales to this sector are notoriously slow and subject to startup-challenging piloting cycles. Approaches that can leapfrog this difficult pattern will be more attractive.

3. Informed Go-to-Market Strategy: Understanding of developer idiosyncrasies along with a command of utility regulatory frameworks, planning and budgetary cycles, as well as data security and privacy concerns are key to teams focused on addressing interconnection challenges.

Interconnection 101

Interconnection is the process by which new generation (supply) and load (demand) resources gain access to the grid in order to draw down or feed in power. In the case of new generation, a developer submits an interconnection request to the grid operator and undergoes a series of studies that simulate each proposed project’s impact on the grid and evaluate the necessity of infrastructure upgrades to ensure grid reliability. These studies are highly technical analyses requiring significant computational resources and engineering expertise. The graphic below illustrates the key steps in connecting a generator to the bulk transmission network.

Overview of the Interconnection Process

Source: Lawrence Berkeley National Lab

Interconnection is plagued by several systemic challenges. Data is nonstandard across grid regions, quickly outdated, fragmented, and potentially subject to onerous approval processes if classified under Critical Energy/Electric Infrastructure (CEII). Information asymmetry drives developers to submit speculative projects years in advance, as queue entry becomes the primary means of obtaining reliable data. Even when accessible, data often lives in formats not conducive to discoverability or analysis; Lawrence Berkeley National Lab (LBNL) spent ~1,700 person-hours scraping PDFs to glean project-level interconnection cost estimates for their recent interconnection cost study (new tools like interconnection.fyi are directly addressing this challenge).

Technological limitations and complex cost allocation mechanisms further convolute the process. Developers bear unpredictable costs for network upgrades that could benefit the broader system, with post-construction Affected Systems Studies that evaluate impacts on neighboring grids potentially adding millions in unexpected costs. The process of identifying mitigations flagged in System Impact Studies remains largely manual, while industry standard modeling tools like PSS®E and Cyme for Power Flow, Contingency, and Stability Analyses are decades old. Analysis of electromagnetic transients (EMT) is becoming increasingly necessary as inverter-based resources (IBRs) proliferate, but existing tools (e.g. PSCAD, PowerFactory) are computationally intensive and impractical for large-scale analyses. These technical challenges are exacerbated by a shrinking workforce of power system engineers, as retirement-driven ‘brain drain’ leaves utilities resource-constrained.

Regulators and policymakers are not blind to this problem. DOE’s i2X (Interconnection Innovation e-Exchange) program – a multi-stakeholders collaboration including utilities, regulators, developers, and researchers – published the Transmission Interconnection Roadmap in April (snapshot below) with recommendations for streamlining and accelerating the interconnection process. FERC hosted a two-day workshop on Generator Interconnection in September, convening representatives from utilities, RTOs and ISOs, academia, developers, startups, and think tanks to address needed reforms in the interconnection process and highlight potentially innovative approaches such as SPP’s proposed ‘Entry Fee’ and ERCOT’s ‘Connect-and-Manage’ approach.

Source: Department of Energy i2X

The Interconnection Problem

Rapid renewable development has driven significant supply-side interconnection bottlenecks over the past decade. LBNL’s latest ‘Queued Up’ report estimates there are 2.6 terawatts of generation waiting in queues – over 2x the installed capacity of the entire US generating fleet. The vast majority of that capacity is low carbon, with solar and battery storage accounting for ~80% of the 900+ GW that entered the queue in 2023.

Source: Lawrence Berkeley National Lab

Unfortunately, most of these projects will never get built. Per LBNL, just 19% of queued projects (14% of capacity) between 2000-2018 reached Commercial Operation (COD) as of EoY 2023. Median time from interconnection request to COD has more than doubled to 5 years for projects built in 2023. Queue backlogs are so severe that CAISO (the grid operator for California) froze an entire cluster of projects from 2022-2023, while PJM (the grid operator for most of the East Coast states) froze all new requests in 2022 until at least 2026.

Source: Lawrence Berkeley National Lab

The aforementioned cost study highlights how interconnection costs have increased in all regions, driven by broader network upgrades flagged to maintain network reliability or stability. Costs for projects that have completed all studies often doubled, while active projects see even greater increases and projects withdrawn from the queue face the highest costs.

Source: Lawrence Berkeley National Lab

Longer queue times and increasing interconnection costs jeopardize project economics at a time when renewable development needs to accelerate. Despite continued growth in US clean energy deployment (~40 GW in 2024 across solar, wind, storage, and nuclear), meeting the nation’s goal of 100% carbon-free electricity by 2035 would require a near doubling of annual capacity deployments. Interconnection bottlenecks are hamstringing renewable deployment at a time when we need to put our foot on the (proverbial) gas.

A Load Growth Problem

Load growth of unprecedented speed and scale complicates the picture. Last year NERC, (the North American Electric Reliability Corporation, the entity tasked with ensuring power system reliability) nearly doubled its 5-year load growth projections to 4.7%—a sea change for a slow-moving utility industry that has seen essentially zero net load growth for two decades:

Source: Grid Strategies

More recent estimates from ICF nearly double that again, projecting a 9% increase in national load and a 5% increase in summer peak by 2028:

Source: ICF

Massive investments in domestic manufacturing, electrification, and large-scale compute infrastructure have sent utilities scrambling. WoodMac puts it succinctly: “[the] last time the US electricity industry saw unexpected new demand growth like this was during World War II.” The disproportionate compute demands of AI data centers in particular are driving acute pockets of load growth. The GPUs required to train and run these AI models are very power hungry; AI data centers can require 3 to 5 times the power of conventional facilities of equivalent scale, with ‘hyperscale’ sites requiring hundreds of MWs if not GWs of power. Securing interconnection is the key bottleneck to data center development, and while electricity costs account for roughly 46-60% of site OpEx, hyperscalers are willing to pay a premium to secure reliable, clean power because AI data centers are such significant profit drivers. Geographic constraints and intense competition leave little flexibility in development timelines and will likely lead many groups to bypass the interconnection logjam by pursuing behind-the-meter solutions. Amazon and Talen Energy are revising a planned expansion of power sales from the Susquehanna nuclear plant in Pennsylvania to a co-located AWS data center after FERC recently rejected their proposal due to concerns of undue impact on ratepayers. Google plans to contract for 115MW from Congruent portfolio company Fervo Energy via NV Energy, and Microsoft signed a 20-year PPA with Constellation Energy to restart Unit 1 at 3 Mile Island and purchase 100% of that reactor’s output. All three have invested in – if not already contracted power – from early-stage nuclear fusion or small modular fission reactor companies.

Transmission planning teams at 9 of the country’s 10 largest utilities cited data centers as a main driver of load growth in earnings calls earlier this year – up from just 2 during the same period last year. Dominion Energy (the utility for VA and NC) is regularly getting asked to interconnect loads of several gigawatts against a generating base of 34 GW. Our grid is physically unprepared to accommodate this influx of demand due to lagging transmission development, and grid operators are logistically unprepared to manage it given their long planning cycles and resource constraints. Most don’t even have formal process for interconnecting large loads given historically tepid demand growth.

A Failure of Transmission Planning
These challenges compound to constrain transmission development at a time of needed acceleration. DOE’s National Transmission Needs Study estimates a nationwide buildout of 54,500 GW-mi is required to meet “future scenarios with moderate load but high clean energy assumptions” (i.e. as enabled by the current policies like the IIJA and IRA). This translates to a ~64% expansion of the transmission system as of 2023. The annual growth rate of transmission over the last decade has been around 1% per year, with average deployment of 2,000 miles from 2012-2016 decreasing to just 700 miles from 2017- 2021. Just 55 miles of new high-voltage transmission were built in 2023. Despite the lack of new capacity, reliability upgrades and replacement of aging infrastructure drove to annual transmission expenditures to a record $25+ billion last year.

Lack of proactive, coordinated transmission planning has resulted in interconnection requests being a primary driver of transmission development and led to the patchwork grid we have today. Regulators have recognized these challenges but continue to promulgate incremental, piecemeal reforms. Last November’s FERC Order 2023 mandated cluster-based queue processing along with stricter penalties on both developers and grid operators to improve efficiency and disincentivize speculation but fell short of addressing cost allocation or granting FERC backstop authority for siting critical projects. Order 1920 from this May enhanced transmission planning by requiring improved coordination and the adoption of advanced modeling techniques to better address emerging grid challenges. It updated the long-term planning time frame to 20 years, encouraging utilities to incorporate renewable resources and future grid needs more proactively, but stopped short of mandating regional coordination or addressing cost allocation for large-scale transmission projects. Pressure is mounting on FERC to overhaul both the interconnection and transmission planning processes, but policy experts do not expect another major ruling on interconnection anytime soon.

This regulatory and technical morass leaves climate investors with a unique conundrum: many infrastructure projects are now bottlenecked by power needs, and the path to popping the proverbial cork is not clear. So how does an entrepreneur or investor address this problem? Congruent does not have the answer, but we are interested in innovative solutions that can address a meaningful slice of this system-wide challenge.

The Interconnection Innovation Landscape

Congruent invests in bits, atoms, and everything in between. Digital solutions can accelerate and optimize the deployment of physical solutions such as Advanced Conductors (shoutout to Congruent portfolio company VEIR), Dynamic Line Rating, Topology Optimization, and other Grid-Enhancing Technologies. Many startups are addressing interconnection, but few have made it their entire business. There is significant functional overlap across grid software applications, making distinct characterization challenging. The list below is not intended to be exhaustive – please reach out if we missed your company!

Congruent’s Bull & Bear Perspectives

Enablers (Bull View):

1. Transmission Constraints: Significant growth in transmission capacity is required to meet the needs of an increasingly renewable and distributed grid and ensure system reliability. Rapid and acute load growth coupled with the increasing penetration of non-dispatchable resources on a transmission-constrained grid is making Resource Adequacy a real concern during utility scenario planning. Software solutions that streamline the interconnection process can meaningfully accelerate grid expansion because interconnection is such a significant driver of transmission.

2. Technology & Process Gaps: Interconnection is dominated by legacy tools from incumbent software providers. Persistent analog and manual processes limit coordination as well as data availability and quality. Modern software tools leveraging the latest advances in computational techniques, data management, cloud computing, and AI/ML can drive significant process efficiencies across applications.

3. Regulatory Pressure: FERC, DOE, and state-level actors are applying tangible pressure on grid operators to streamline the interconnection process and integrate emerging technologies. We believe there is an opportunity for startups to fill this technology gap.

Obstacles (Bear View):

1. Adoption Cycles: There are inherent challenges to scaling startups in this space. Utilities’ and grid operators’ reliability mandate makes them risk-averse and slow adopters of new technologies, while developers are more commercial but constitute a smaller market opportunity and tend to be idiosyncratic.

2. Regulatory Risk: Interconnection challenges are symptomatic of broader dysfunction in grid planning. Recent FERC orders and ISO/RTO-level reforms, while slow-moving and incremental, signal willingness to address these challenges head-on. A regulatory overhaul could diminish the value proposition of interconnection startups by obviating the process inefficiencies they are addressing.

3. Limited Opportunity Size: This space is highly intermediated by consultants and entrenched legacy software providers. With few public or sizeable M&A comps to benchmark against, it is unclear there is potential for a large outcome for a startup building in interconnection.

Focus areas

A quick policy detour: the outgoing Biden administration recently published a memorandum calling for interagency coordination to “streamline permitting, approvals, and incentives for the construction of AI-enabling infrastructure,” including transmission. Just yesterday, DOE announced $30M in funding from the Bipartisan Infrastructure Law for solutions that leverage artificial intelligence to accelerate the interconnection process. Although the incoming Trump administration’s stance on transmission is unclear, grid expansion and power system reliability are bipartisan issues. Congruent is actively tracking the ‘GRID Power Act’ (H.R. 9801), Congressman Troy Balderson’s (R-OH) bill directing FERC to initiate a rulemaking requiring that transmission providers prioritize new dispatchable generation in interconnection queues. While the fate of Sens. Joe Manchin’s (I-WV) and John Barrasso’s (R-WY) ‘Energy Permitting Reform Act of 2024’ (S.4753), is uncertain, a unified Republican government will almost certainly pursue permitting reform by weakening regulations such as NEPA along with EPA rules and other perceived forms of red tape. It is too early to gauge whether the administration will pursue a technology-agnostic stance or seek to penalize renewables in favor of fossil energy development. As for FERC, president-elect Trump can nominate a new chairperson on day one, but the five-member Commission still proceeds by vote and could maintain its existing Democratic majority for another 18 months given its rolling terms and the spate of recent appointments.

While we are concerned about the net environmental impacts of rolling back environmental regulations, we are cautiously optimistic that a second Trump term may result in a more favorable permitting regime for transmission and other critical infrastructure, and will continue to update our thinking as the new administration takes shape.

There is no clear roadmap for selling software to utilities, developers, or grid operators, and few examples exist of startups that have successfully scaled in those segments. Congruent invests in companies we believe can grow rapidly to capture a large market opportunity and generate outsized returns on both capital and impact. We have identified at least four discrete critical technology gaps that could be filled by a software startup, but ultimately believe value will accrue to platforms that can integrate several functions and sit across multiple utility/grid-operator business units:

1. Data Layers: Utility or grid operator-maintained platforms that provide transparent, real-time grid information to developers such as queue status and location-based availability of injection/withdrawal capacity. AEMO’s Connection Simulations Tool is a prime example and allows registered participants to run their own analyses using the grid operator’s full power system model.

2. Advanced Grid Modeling Tools: Automation of complex grid studies like contingency and dynamic stability analyses offers a significant opportunity to reduce interconnection timelines and improve reliability. EMT tools capable of larger, system-wide analysis in a computationally efficient manner are needed as IBRs make grid behavior increasingly complex.

3. Workflow Automation: Streamlining the highly manual and cumbersome queue management processes will be essential to reducing project delays and speeding up interconnection timelines. Everything from application management to communication of timelines and costs with developers.

4. Data Center Performance Improvement: given the extreme power needs of AI data centers, we expect to see innovation focused on behind the meter power usage effectiveness (PUE) in many forms to maximize compute output within the fixed load footprint.

Please reach out if you are building in any of these areas. With the exception of Camus Energy, we have not yet made an investment directly in this category, and we look forward to meeting entrepreneurs tackling these challenges head on!

Sources

1. Lawrence Berkeley National Lab (1)

2. Lawrence Berkeley National Lab (2)

3. DOE Interconnection Innovation E-Exchange

4. DOE I2X Transmission Interconnection Roadmap

5. FERC Generator Interconnection Workshop

6. Lawrence Berkeley National Lab (1)

7. Lawrence Berkeley National Lab (2)

8. World Resources Institute

9. Grid Strategies

10. ICF

11. WoodMac

12. IDC

13. Utility Dive (1)

14. Bloomberg

15. MIT Technology Review

16. New York Times

17. Reuters

18. DOE Grid Deployment Office

19. Utility Dive (2)

20. DOE Office of Policy

21. Clean Energy Grid

22. Ibid

23. FERC

24. FERC(2)

25. DOE Office of Electricity

26. Whitehouse.gov

27. DOE Grid Deployment Office

28. Congress.gov

29. Senate Committee on Energy and Natural Resources

30. The Hill

31. Utility Dive (3)

Our TLDR

The transition to a low-carbon economy needs enormous quantities of raw materials – particularly metals – to deploy and scale. Congruent spent time with innovators, mining companies, academics, and investors to understand how biological approaches to mineral extraction – “biomining” – might reduce the cost, carbon footprint, and environmental impact of extraction and refining processes. This part of the mining space faces particular challenges regarding adoption and scalability. We are intrigued, but cautious, about investing in biomining.

Here's what we’re looking for at Congruent:

1. Industry Lean-In: Demonstrated willingness from mining companies to work with startups on early-stage technology development and integration.

2. Technology Readiness: Commercially relevant proofs-of-concept are necessary with at least high-level cost estimates for implementation in real-world environments.

3. Credible Path to Scale: A clear go-to-market strategy and roadmap for integration with existing sites as well as a credible path to meeting the massive scale of large mining operations.

The Energy Transition: A Resource Question

A wide range of metals are essential to electrification, as well as the broader energy transition (see below for a breakdown by application from McKinsey).

Today, sourcing those metals has a considerable environmental impact: mining accounts for 4-7% of global GHG emissions and creates significant local water, air, and land quality concerns. Despite indications that the low-carbon economy will be significantly less extractive than the status quo, innovation is essential to unlock the growth in global metal supply necessary to support the energy transition while reducing environmental impact.

Our research led us to focus on copper due to its criticality across key technologies, lack of substitutes, acute supply concerns, and large market. Widespread — and growing — demand for copper across energy transition technologies is projected to create a mid- to long-term supply deficit. Meanwhile, decreasing copper ore quality coupled with the depletion of existing resources is driving extraction costs upwards; BNEF expects a ~30% reduction in the average copper ore grade this decade, with a decrease from 0.7% in 2020 to 0.5% in 2030.

The graphic below illustrates the main steps in the copper mining process. Copper is found in two main types of minerals: oxides (20% of global production) and sulfides (80% of global production). Pyrometallurgy, the high-heat (~1200 °C) and multi-step process through which most copper is processed today, is an energy-intensive process but is faster and higher yielding for sulfides and, therefore generally more economical than existing alternatives. In hydrometallurgy, copper is extracted from the orebody by applying aqueous solutions (often sulfuric acid) to dissolve and separate the valuable metal from the unwanted minerals. Developing efficient hydrometallurgical extraction methods for sulfides – particularly chalcopyrite, the most common copper sulfide, and thus the most common copper mineral – is the holy grail for copper extraction but impractical today because the sulfur creates a passivation layer that makes the metal unreactive and limits extraction. Biological approaches – or “bioleaching” – that bypass or prevent this effect could help sustainably unlock this valuable resource.

Overview of the Copper Mining Process

Biomining 101

Biomining leverages biological mechanisms to extract and recover metals from ores and waste materials, and broadly encompasses bioextraction (mostly heap leaching, where biological agents are applied to prepared piles of ore and then allowed to slowly perform their extraction processes without additional outside intervention), bioremediation (e.g., tailings remediation), and phytomining (engineering organisms, usually plants, to hyperaccumulate metals from the soil). From a climate perspective, biomining generally has a lower environmental impact than conventional approaches due to lower energy requirements (avoids or reduces energy-intensive steps like milling and smelting) and reduced chemical use. The microorganisms in question are also usually CO2-fixing autotrophs. Our research has led us to focus on biomining as applied to the heap leaching step in copper hydrometallurgy.

The use of biological processes in hydrometallurgy has been documented for centuries, but bioleaching did not gain significant attention in modern mining processes until recently due to limited understanding of microbial processes, technological limitations, and the dominance of conventional mining methods. Despite its long history, biomining remains niche, often limited to treating low-grade ores or complex sulfide deposits. Biological extraction methods accounted for just 1.2% of copper production in 2019-2020. Bioleaching tends to be slower than chemical methods, requiring larger tankage for equivalent throughput. In our view, the most notable impediments to biomining adoption are 1) scale up risk, 2) site specificity, 3) adoption risk, and 4) economics. Recent advancements such as the proliferation of low-cost sequencing and genetic engineering tools, as well as the commodification of bioreactors, have turned attention towards biomining as a viable approach to mineral extraction and generated a flurry of innovation.

The biomining ecosystem is nascent but quite broad. Many companies are developing processes applicable across multiple metals, ore bodies, and resources (e.g. applicable to both heap leaches and mine waste). The list below is not exhaustive but highlights the breadth of companies leveraging biology to unlock new metal supply (not just copper).

Biomining Companies At-A-Glance

Congruent’s Bull & Bear Perspectives

There are many reasons novel biological processes for copper extraction could be an exciting area for early-stage venture investment. Conversely, several technology and market dynamics could inhibit venture-scale outcomes for emerging companies. We’ve summarized our Bullish and Bearish views on this market below:

Enablers (Bull View)

1. Supply Pressure: Ore grades are declining globally, and mining companies face increasing pressure on unit economics as the marginal cost of metal recovery rises. Mechanical and chemical approaches to improving recovery are reaching their limits and create environmental challenges. Biomining could offer a breakthrough in extracting copper from chalcopyrite, overcoming passivation layers that hinder current methods.

2. Growing Tailings Liabilities: Tailings, the waste produced by mining operations, are a significant cost and liability for mining companies. The lack of additional tailings storage capacity is constraining future development, while regulatory and investor scrutiny are mounting. Biomining offers a solution by potentially reducing the volume and toxicity of tailings produced, managing environmental liabilities for mining companies while enabling supply expansion.

3. Massive Market Opportunity: Copper is a highly traded global commodity with broad applications in industrial equipment, manufacturing, and electronics. With sulfide ores accounting for 80% of global copper production from sulfide ores, the target ore type for bioleaching, biomining presents a significant opportunity. If successfully commercialized and scaled across the industry, sulfide leaching could be applicable to ~2.4 million metric tons of McKinsey’s estimated 6.6 million metric ton copper supply gap by 2031 – a ~$45 billion opportunity.

Obstacles (Bear View)

1. Scaling Biology is Challenging: Biological processes are intrinsically difficult to scale, particularly in industrial applications like mining where achieving consistent performance at a scale that matches the output of large mining operations has historically been challenging. Reaction environments must be tightly controlled to prevent contamination and maintain activity, adding complexity in scaling efforts.

2. Highly Heterogeneous Operating Environments: Ore composition, microbiome, and environmental conditions vary across and within mine sites, requiring site-specific engineering of both microbes and process integration. This variability makes it difficult to engineer a repeatable, scalable process. Biomining technologies will likely require significant restudy and reconfiguration for each deployment.

3. Customer Adoption Challenges: Mining companies are extremely risk-averse for very good reasons. They need to see significant improvements in throughput, purity, or yield with no risk to unexpected downtime for their operations. Long sales cycles and site-specific technology procurement make widespread adoption challenging. Additionally, many mining majors have in-house biohydrometallurgy expertise or hire specialized engineering firms, making it tough for startups to gain a foothold.

4. Economics: The marginal additional yield for lower-value metals like copper may not justify the CapEx and OpEx investment necessary to implement a biomining approach. While this may make more sense for higher value metals like gold, those markets are much smaller and less important to the energy transition.

Focus Areas

While biology has many potential applications in mining, biomining is still very niche. The long sales cycles associated with mining companies and the site-by-site specificity of mining operations are nontrivial impediments to scale. Congruent looks for companies that can quickly scale to produce outsized returns and impact. While we haven’t yet made an investment in this category, we have identified three potential areas where innovation could provide a significant unlock and may fit the venture model:

1. Improving copper extraction from chalcopyrite (70% of world’s copper supply) – likely a breakthrough around overcoming passivation.

2. A platform approach that leverages proprietary data and insights to help mines better engineer their biohydrometallurgy processes.

3. Applying biology to “comminution” (the first process step in mineral extraction, which is grinding or milling the ore) to pre-condition the orebodies to reduce the energy required to crush the rocks, which also reduces energy downstream throughout the entire extraction process.

Please reach out if you are working in any of these areas, and we look forward to meeting entrepreneurs tackling these challenges head on!

Sources

1. Sustainability By Numbers

2. McKinsey [1, 2]

3. GlobalData

4. The International Energy Agency (IEA)

5. Bloomberg New Energy Finance (BNEF)

6. Springer [1, 2]

7. University of Arizona

Our TLDR

A lot more investment is needed to decarbonize the heat used to power the industrial sectors we rely on for food, infrastructure, materials, clothing, medicine, and more. We spoke with dozens of industry incumbents, innovators, and policymakers to form a unique view on what an industrial heat startup needs to have to fit the venture model. We cover why startups tackling industrial heat decarbonization are underfunded relative to emissions abatement potential, and our view on whether these emerging companies are good venture capital investments.

Here’s what we screen for at Congruent:

1. Process Integration: Solutions that are modular and can replace existing systems with minimal friction are the most likely to overcome customer adoption hurdles.

2. Business Continuity: Technologies that maintain reliable heat and do not threaten to disrupt plant operations are table stakes.

3. Financial Viability: Zero upfront CapEx is best, but if CapEx is required, we believe customers will look for payback periods between 3-5 years (or less) to consider an investment.

4. Technology Risk: We look for technologies in the sector that are already de-risked or will not require a substantial amount ($20+ million) of additional equity to reach commercial pilot.

If a company checks all four of these boxes, it may be a fit for our portfolio.

The hotter the better?

Not necessarily. While higher heat is typically linked to higher abatement potential, we believe there are compelling companies across the entire heat spectrum (from food to steel). Each industrial sector has unique challenges to deploy clean heat technologies, but the common denominator among them is a large TAM that may have significant revenue potential.

Selected Industrial Heat Innovation Areas

For companies targeting low-medium heat sectors (typically food, paper, chemicals, textiles, pharma), we prioritize proven technologies and innovative go to market strategies that address key adoption obstacles (e.g., industrial grade heat pumps). Specifically, we’ve seen structured financing and equipment leasing structures that reduce or eliminate upfront costs while enabling energy savings and carbon reduction. Since these technologies already exist and are commercially deployable today, they may have lower customer adoption hurdles and faster growth trajectories. Companies operating here need to be able to prove a defensible, sustainable margin profile as penetration and competition scale.

For companies targeting high heat sectors (typically steel, cement, glass), there is inevitably more technology risk that exacerbates the challenge of commercializing on a venture timeline (e.g., thermal energy storage, hydrogen). However, Congruent doesn’t shy away from investments with technology risk, particularly when we have high conviction in a team’s ability to hit early technical milestones such as industrial pilot projects. When looking at high tech risk opportunities, we like to see incremental investment dollars materially reduce risk at every step along the development path.  Companies that need to build a full-scale plant before economics are clear may not be a fit for us. The notable tailwinds behind industrial decarbonization from policy to cost volatility reduction are further catalysts that can make investment in first-of-a-kind technology uniquely timely today (more on this below).

Industrial Heat Background

There’s a reason why the Paris Agreement benchmarks its global temperature increase limit of 1.5°C to “pre-industrial” levels. The Second Industrial Revolution (~1870-1914) followed by World War I (1914-1918), was an extended period of mass-production for cement, steel, and consumer goods. This marked the start of the industrial sector’s significant and consistent contribution to fossil fuel emissions and climate change over the past century.

Global Temperature Mean Since the Second Industrial Revolution¹

Globally, industrial processes consume roughly one-third of all energy and heat represents ~75% of the sector’s energy consumption. Today, 90% of this energy is still generated by fossil fuels. Industrial processes require heat at a wide range of temperatures, roughly 50% “low-heat” and 50% “high-heat.” Sectors like food, paper, chemicals, textiles, and pharmaceuticals require low-temperature heat below 200°C to power processes such as washing, drying, and sterilization. Sectors such as steel, cement, and glass require high-temperature heat well above 1,000°C to enable material transformation such as melting iron ore to form steel.

Industry Leads Global Energy Consumption ²

As venture capital investors, we recognize a large opportunity to decarbonize industrial process heat by replacing fossil-based incumbent technologies with low carbon alternatives over the next decade. However, industrial heat is one of the most difficult areas to decarbonize, particularly when high temperatures are required, for a variety of reasons. In our view, the most notable impediments to clean technology deployment are 1) customer adoption, 2) cost, and 3) time to deploy. This is perhaps why there is a significant shortfall between the investment dollars earned by industrial heat startups relative to their emissions abatement potential.

Abatement Potential Funding Gap ³

Heavy industry received only 8% of overall venture capital investment between Q1 2013 and Q3 2022 despite contributing over one-third of global emissions. While this number ticked up to 14% between Q4 2022 and Q3 2023, it is still substantially below where it should be relative to total emissions. For comparison, mobility has received roughly half of all investment despite only generating 15% of global emissions. 

Congruent’s Bull & Bear Perspectives

There are many reasons the decarbonization of industrial process heat could be an exciting area for early-stage venture investment. At the same time, there are certain dynamics underlying industrial heat that could inhibit venture-scale outcomes for emerging companies. We’ve summarized our Bull and Bear views on this market below:

Enablers (Bull View)

1. Large TAM: We estimate that $500 billion or more of total CapEx will be needed to decarbonize industrial process heat. Our internal analysis suggests total CapEx of $300 billion just to replace existing boiler capacity with industrial grade heat pumps ^4,5. We estimate the opportunity for other low-medium heat technologies and emerging high heat solutions to be similar based on per plant CapEx of $1-50 million and the US plant footprint across industrial sectors. 

2. Government Funding: While we don’t overly credit access to government funding in our investment decisions, it can be a critical tailwind at the earliest stages where we invest. $6 billion is earmarked under the IRA for early-stage commercial-scale industrial demonstration projects and $10 billion is available from investment tax credits for eligible projects under 48C, both encompassing industrial heat.

3. Regulatory Tailwinds: There are several notable examples of favorable cross-state and cross-border regulations underpinning industrial demand for clean heat. CA SB 253 will require over 5,000 businesses operating in California to report Scope 1-3 emissions beginning next year. CBAM in Europe will incentivize industrial importers to decarbonize faster to avoid tariffs beginning in 2026. For industrial companies with foresight, this type of policy should stimulate attention and investment of resources toward decarbonizing heat.

4. Technology Availability: Several technologies that can be used to decarbonize low- and medium-heat sectors, such as industrial heat pumps, are largely de-risked and ready to commercially deploy today. These technologies can also be more efficient than their natural gas counterparts. For example, industrial heat pumps have a coefficient of performance of 3x compared to <1x for natural gas, meaning that heat pumps can more efficiently turn power into heat ⁶. This often overcomes the spark spread where methane is less expensive per BTU than electricity, but less efficient in converting that energy to usable heat.

5. Offtake Commitments: Coalitions such as the First Movers Coalition can create early market demand to bring emerging industrial heat decarbonization technology to commercial scale.

Obstacles (Bear View)

1. Customer Adoption: Industrial customers will simply not disrupt existing processes, jeopardize business continuity, take unnecessary technology risk, and spend additional CapEx/OpEx to decarbonize their heat. Industrial processes require constant heat, operate with razor thin margins, and thus have a high cost of failure. There is also an acute sense of inertia, lack of awareness, and resistance to change more generally that will take time to shift. 

2. High CapEx: As mentioned previously, we estimate CapEx per plant to be $1-50 million. Manufacturing facilities often have very low operating margins and low CapEx budgets creating a funding mismatch. Decision makers are either unwilling or unable to commit large dollars to these types of projects. 

3. Unit Economics: Natural gas is still the cheapest way to generate industrial heat in many manufacturing-heavy regions in the US, creating a further disincentive for decision-makers to pursue alternative forms of energy for heat generation. The lower the natural gas price, the lower the value of cost savings and longer the payback relative to the capital cost of a clean alternative.

4. Technology Risk: The high-heat segment specifically requires advanced innovation beyond mainstream engineering that is talent-constrained and may take longer than 10 years to successfully commercialize.

5. Infrastructure Risk: Electrifying processes currently powered by fossil fuel will take a significant amount of electricity, perhaps in excess of what the current grid can support.  Building out a higher capacity grid will take time and likely significant regulatory and permitting support and reform.

Final Thoughts

We’re excited to see so many startups tackling the problem of industrial heat from all angles.  As venture investors, we look for companies that can scale quickly to billion-dollar outcomes. While we have not found a company that fits our investment criteria for this category, we do believe status quo changing companies will emerge. We are fueled to continue tracking investment opportunities in the space based on large addressable market opportunities and urgency for decarbonization. If you are an innovator working on a solution to industrial heat, we’d love to hear from you! 

Sources

1. NASA

2. Engie Impact 

3. IPCC, Pitchbook, PwC Analysis

4. Department of Energy

5. NREL. Assumes heat pump cost of $50-$200k per MMBtu of heat delivered.

6. Renewable Thermal Collective: Industrial Thermal Decarbonization Package

When we started Congruent in 2016, raising an early stage climate fund was an exercise familiar to the founders we back: constant rejection. We were raising an institutional seed fund focused on the most “risk on” stage of a sector that had been returns-challenged in the prior years. With the close of Fund III, we have crested $1B in AUM. We are proud of what we’ve built (and what our portfolio has accomplished) but most of all we are optimistic that the LP investor community is eager to deploy capital to address our climate challenges. With great climate challenges comes great opportunity.

Much to our excitement, we were 3x oversubscribed with $600M of LP interest on a $200M target size, exceeding this target in less than five months and hitting $275M of total commitments in a first and only close. While institutional investors are not homogeneous, it is universally the case that it is easier to back seasoned investor teams, and as Congruent has grown, the portfolio has matured, and our team has grown to include 12 of the most talented and diverse in the space. We are grateful for the success in the face of a challenged venture environment, aptly shown by a Pitchbook analysis.

Why not take more?

One of our favorite inside-venture quotes is attributed to Mike Maples: “Your fund size is your strategy.” If we allowed our fund to bloat much beyond the $275M, we believe there would be pressure to move to large Series A’s and Series B’s as entry points given portfolio construction. We love supporting the earliest stage entrepreneurs from inception, and the size of Fund III allows us to continue to do so with even more depth. We are fortunate to be able to continue to support to our growing early stage portfolio through our Continuity Fund, announced in April of this year which allows us to invest in later stage rounds in our breakout companies.

With 53 companies in the portfolio, 12 team members, and $1B in AUM, we are delighted to redouble our efforts around our core mission: backing early stage climate entrepreneurs and delivering best in class returns to our investors.

The full press release can be found here, and we are grateful for the support of all of our LPs including the California State Teachers’ Retirement System, the Grantham Foundation, the University of California, Sobrato Capital, Strategic Investment Group, clients advised by Cambridge Associates and new investors including Northwestern University and Vintage Investment Partners in addition to multiple institutions representing some of the largest endowments, pensions, sovereign wealth funds, and foundations in the world.

If you’re looking for your first (or occasionally second) institutional round, come find us!

-Josh & Abe

Subscribe now

Share

The state of venture was a key theme this year as the rising interest rate environment, banking crisis, and tech sector headwinds have brought the soaring venture market plummeting back to earth. U.S. venture deals and dollars deployed have dropped precipitously quarter over quarter since their 2021 peak.

Climate, however, has proven more resilient than the broader venture market. Congruent invests early – and at our stages (Pre-Seed/Seed/Series A), climate venture continues to move along at a healthy clip.

At later stages, an IPO-starved environment and the hasty retreat of crossover funds spurred by the declining value of their public books has caused a dramatic drop in venture activity. Still, late-stage climate had a banner year in 2022 compared to every year other than 2021.

What’s going on here? While macro conditions have hammered general tech over the past year, stronger public company performance in industrials, energy, and other physical world sectors that many climate startups sell into creates a more robust customer base. Between state and federal policy, declining renewable energy costs, significant investments in electrification and decarbonization from industry, and corporate sustainability commitments, tailwinds abound for climate tech.

While our data sets differ, our friends at Carta tell a similar tale for early-stage climate venture in Q1 2023:

The upswing of climate capital fund formation in the past couple years implies the presence of dry powder to deploy into climate companies, and while transaction volume notably declined recently (likely due to the regional banking crisis and the collapse of SVB in particular), Seed-stage climate tech companies raised larger rounds and at higher pre-money valuations compared to broader venture.

Venture will likely continue to be in flux for many quarters to come and we continue to closely monitor market conditions. We’ve included the market analysis slides that accompanied our AGM with this post. While our snapshot of climate venture activity is more conservative than most, we’re optimistic that climate venture is well-positioned to weather the storm and deliver positive both impact and returns.

Cv 2023 Annual Meeting Overall Climate 2023

12.1MB ∙ PDF file

Download

Download

Today, we’re announcing our $300M Continuity Funds alongside the release of our 2022 Impact Report.  We’re delighted to see that our efforts have catalyzed over $1B of capital across 51 companies employing over 2,000 people.  We’re also gratified that our capital was the first institutional capital invested in four out of five of our portfolio companies, with more than one in four of those women-led. There is always more work to do on our impact, and we will continue to invest in these efforts.

The Continuity Funds are designed to let us continue to back the breakout companies from our first two funds as they scale. We raised this fund largely to retain the agility we value throughout our firm and in response to our growing portfolio’s needs.  We were joined by new and existing LPs who represent more than $2 trillion in assets including CalSTRS, investors advised by Cambridge Associates, multiple U.S. and Canadian pension funds, the Grantham Foundation, Builders Vision, the Three Cairns Group, Sobrato Capital, and multiple other institutions representing some of the largest endowments, pensions, and foundations in the world. This brings Congruent’s AUM to more than $700M with 51 investments across our early-stage pure play climate tech portfolio, making us one of the most active firms in the sector.

Since launching the Continuity Fund, we have made four investments in Congruent companies including Meati Foods, AMP Robotics, Fervo Energy, and Span.  These companies are all scaling in their respective markets and have raised substantial capital from world-class syndicates.

We have also been building the firm beyond the dollars. Since closing Fund II, we have added two Partners, Jackie Kossmann and Tanuj Dutta, promoted Christina O’Connor to Partner and Nick Adeyi to Investor. We also brought on Suzie Parmar as our first VP Finance and most recently, Abby Steele as our first Executive Assistant. The team stands at 11 and each member brings a wide and valuable range of experience (and humor) to the firm.

We are excited to continue to look for and back amazing founders at the earliest stages of the journey and work side by side to address the biggest challenges in climate change.  Join us!

The Climate Growth Fund Targeting Its Early-Stage Portfolio Makes Congruent One of the Largest Early-Stage Climate Investors with More Than $700M in Assets under Management

SAN FRANCISCO, CA, April 19, 2023 — Congruent Ventures, an early-stage climate and sustainability venture firm, today announced the close of its first Continuity Funds focused on quickly scaling its early-stage portfolio with more than $300 million in commitments, highlighting the interest of institutional investors seeking exposure to climate venture despite a challenging fundraising environment.

This brings total AUM for Congruent to more than $700 million, making it one of the largest climate portfolios in early-stage venture and nearly doubling the firm’s AUM. The fund’s commitments represent Limited Partners with more than $2 trillion in assets, bucking the recent trend of LPs trimming their exposure to venture capital and private equity.

The Continuity Fund navigates the ever-evolving climate environment to support Congruent’s early-stage portfolio companies as they scale to meet the demands of the changing climate innovation landscape. Congruent partners with companies at the earliest stages, and the Continuity Fund will be deployed to support them throughout their life cycles.

“Congruent invests in early-stage transformational companies in climate tech, where we leverage our capital and deep in-sector experience to fund the best companies in the space,” said Joshua Posamentier, co-founder and managing partner of Congruent Ventures.

Congruent’s portfolio includes category-leading climate companies from food to fusion, including Meati Foods, AMP Robotics, Fervo Energy, Span.IO, Hippo Harvest, Pano.AI, Avalanche Energy, Parallel Systems, and Lightship, among many others.

Congruent Ventures, launched in 2017, is co-led by Joshua Posamentier and Abe Yokell, who have decades of climate tech investing expertise between them. Congruent’s team has grown to 11 members who are among the most experienced and diverse in early-stage climate, having invested through the first Cleantech cycle. Investors supporting Congruent include CalSTRS, investors advised by Cambridge Associates, multiple U.S. and Canadian pension funds, the Grantham Foundation, Three Cairns Group, Sobrato Capital, and multiple other institutions representing some of the largest endowments, pensions, and foundations in the world.

“The fast-paced innovation and change now powered by funding from the 2022 Inflation Reduction Act must be met with sufficient venture funding,” said Yokell, Congruent’s co-founder and managing partner. “We saw a strategic need to help our existing portfolio companies, and we’re delighted to have the support of our Limited Partners despite the broader pullback in venture investing.”

Congruent’s previous two fund families (Fund I family in 2017 at $92M and Fund II in 2021 at $175M) have been focused on supporting entrepreneurs who are building solutions to address sustainability challenges, investing early across hardware, software, enterprise, consumer, deep technology, fin-tech, and business model innovation. Our climate sector focuses on mobility and urbanization, energy transition, food and agriculture, and sustainable production and consumption.

Congruent has supported 51 companies over the past six years, making it one of the most active and experienced U.S. investors in climate. Its portfolio represents companies that will help to decarbonize every sector of the economy—energy, fleet electrification, farming, new food products, sustainable aviation fuels, manufacturing, and more.

Today, we’re 10 people strong having added three Partners (Christina, Jackie, and Tanuj), a Venture Partner (Marianne), and a veteran bench of operations, finance, and legal team members (Tina, Suzie, and Nadia). Last, but not least, Nick Adeyi joined us two years ago as an Analyst from Guidehouse Associates’ Energy, Sustainability & Infrastructure consulting practice.

Our goal has been to build a world-class team of knowledgeable early-stage climate venture investors with whom companies love to work. As part of that effort, we work to support our team in their personal and professional development while continuing to build our portfolio (now 49 companies deep).

It gives us great pleasure to announce that Nick has been promoted from Analyst to Investor. We hired Nick in a pre-MBA role with the expectation that he would follow the traditional path of moving on to an operating role or grad school after a few years. His presence, attitude, curiosity, and maturity when evaluating sectors, companies, and teams are well beyond his years and we’re delighted to have him graduate into our Investor position.

As the capital markets rotate towards decarbonization across every sector of our economy, we will need the next generation of talent to focus on the challenges and opportunities ahead of us.  While we’ve gone from two people and a modest inaugural early-stage fund of $92M to ten people and over $700M of assets under management, we are just at the beginning of our collective work.

And as always, we welcome the climate-curious to come join us. There are over 275 job openings across our portfolio despite the economic environment – please contact our companies; they are hiring!

-Josh and Abe

Several high-profile projects are leveraging the Web3 tech stack to address inefficiencies in the carbon markets, but the integrity of these markets is only as robust as their supply. Facilitating transactions and building financial products before ensuring quality in the underlying asset does not solve the markets’ entrenched quality concerns. We believe improving supply quality is the most pressing need if carbon markets are to serve as a meaningful lever for decarbonization.

In March, the SEC proposed long-awaited climate disclosure regulation that would require companies to publish their material exposure to climate risks and any associated impacts to their business. The rules contain provisions for gross Scope 1 & 2 emissions disclosures as well as Scope 3 if deemed material or included in the company’s stated sustainability commitments. Importantly, the proposal also stipulates detailed REC and carbon offset disclosure requirements including quantity of CO2 claimed, sourcing, underlying project details, registration, 3rd-party verification, and cost. We welcome this move from the SEC as a signal that US regulators are beginning to consider climate risk as part and parcel of the financial system, and recognize the role offsets will play in corporate climate action for the foreseeable future.

Climate risk is financial risk: regulatory emphasis on corporate disclosure is crucial because we believe that public companies not directly addressing climate risks will underperform in the long term. The C-suite is voting with their balance sheets, as corporate commitments have driven demand for offsets in the voluntary markets. Visibility into offsetting activity empowers shareholders and employees to pressure public companies to improve the quality of their procurements. Regulated disclosure requirements will grow nascent carbon accounting companies like Planet FWD — one of our newest portfolio companies — whose Scope 3 emissions platform helps brands measure and manage their emissions at the product level. Scope 3 emissions typically account for the majority of a company’s footprint and yet are the most difficult to measure and mitigate over time; emissions reductions efforts at this level create ripple effects throughout supply chains. As regulations tighten and pressure from stakeholders mounts, we expect improved accounting capabilities will facilitate both reductions in operational emissions and more targeted offset procurements.

Corporate demand also drives and shapes new supply. In a welcome approach, some of the largest, most sophisticated buyers in the market recently penned Advance Market Commitments to the tune of $925m from Frontier (led by Stripe, Alphabet, Shopify, Meta, and McKinsey) and $10B by 2030 from First Movers Coalition, indicating a consensus that large capital outlays from early adopters are required to both scale up and bring down the cost of carbon removal and other crucial but nascent decarbonization technologies. Yet these commitments are orders of magnitude lower on a CO2e basis than what is needed to avoid an undesirable overshoot of 1.5°C. We need far more supply, and that supply — of both removals and avoided emissions — must be verified, additional, and durable.

As climate investors, we take the optimistic view of technology’s role in addressing the climate challenge. However, overemphasizing the importance of transactive infrastructure before supply has meaningfully evolved seems like putting the cart before the horse. The influx of startup and venture activity into this last category is cause for concern given the state of the voluntary markets. To be sure, transparency and liquidity — rather the lack thereof — along with double-counting and impermanence are legitimate problems that must be addressed for the healthy functioning of carbon markets in the long term. However, as recent analyses demonstrate, high-profile projects designed to rectify these issues have fallen notably short of the task; instead of increasing the quality of supply, they have actively induced the addition of lower-quality credits to the market. The philosophical promise of Web3 projects is not without merit, but financialization does not directly address the most pressing barrier to success for the carbon markets: high quality credits are scarce, and existing modes of evaluating quality leave something to be desired.

In our ongoing effort to focus on supply, we made our first direct foray into carbon markets when we invested in Evergrow in Q4 of last year on the thesis that carbon markets were supply-constrained due to a lack of access to traditional project financing. While our thesis continues to evolve, we believe supply-side efforts will yield the most impactful results where aligning markets with climate impact is concerned. We are encouraged to be tracking an ever-increasing pace of company formation focused on these challenges. Project origination and financing, low cost technology approaches to carbon removal, credit quality assessment, and MR&V are just a few areas where concerted investment could move the needle, and we look forward to meeting more founders directly tackling these challenges.

Nicholas Adeyi is an Investment Analyst at Congruent Ventures, an early stage venture firm founded in 2017 that partners with companies in the climate and sustainable technology ecosystem.

Market-based mechanisms have long been touted as the silver bullet climate policy for driving economy-wide emissions reductions. When thoughtfully designed and properly implemented, these mechanisms create economic incentives that push polluters towards emissions-reducing behaviors. Carbon offsets are a particular carbon pricing mechanism and asset class whereby buyers certify emissions reductions through credit purchases in lieu of reducing their own emissions. In theory, carbon offsets buy time to decarbonize hard-to-abate activities such as aviation and heavy industry while bringing emerging carbon removal technologies (such as Direct Air Capture) down the cost curve by financing initial deployments and scaleup.

The ugly truth of today’s carbon market

Or so the story goes. In practice, the development of a healthy carbon offset ecosystem has been hampered by questions of additionality and permanence, and markets lack robust, standardized protocols for the measurement, reporting, and verification (MR&V) of carbon removal. Despite continued market growth, pricing across the asset class is inefficient, failing to account for permanence, reversal risk, and co-benefits.[i] Carbon markets suffer from an oversupply of cheap, low-quality credits; in 2020, estimated trade volumes in the voluntary markets increased 80% YoY to 188.2 Mt CO2e while credits transacted at a weighted average price of just $3.13 per metric ton.[ii] Compare this price to the $51 per metric ton estimated social cost of carbon[iii] and you’d be right to be skeptical of this 16x+ disparity.

Quantity at the expense of quality makes it far too easy for polluters to claim emissions reductions and hampers progress towards the IPCC’s estimated 2–10 Gt CO2e of annual removal required to achieve global net-zero emissions by 2050.[iv] Of the total 154 Mt CO2e and 16 Mt CO2e Microsoft and Stripe reviewed in recent carbon removal procurements, just 3% and 0.15% respectively both met their criteria for high-quality carbon removal and were immediately available.[v] What we need, as Ed Smith of Carbonware puts it, is “a market rally for quality,”[vi] starting with massive investment into durable, verifiable, offsets.

Where do we turn to fix our supply issues? (hint: it’s developers and price security)

This market environment has made financing for carbon project developers particularly scarce. Few credit-worthy offtakers will sign long-term carbon purchase agreements, without a regulatory or market-based imperative. And absent predictable cash flows, developers struggle to access development capital and commercial debt. Similar hurdles compound upwards through the capital stack, effectively hamstringing developers’ access to project financing. While a growing number of startups address MR&V challenges in offset markets by ensuring accurate measurement of carbon flows (Pachama, Sylvera), and others facilitate offsetting operational emissions (Cloverly, Patch) or provide a marketplace for offset transactions (Nori, puro.earth, NCX), few if any focus on solving the challenges developers must overcome in order to grow the supply of high-quality carbon removal projects.

Enter, Evergrow. As the world’s first company purpose-built to finance carbon developers, Evergrow will both contract offtake of credits from projects that avoid, reduce, or remove carbon, and provide project capital to finance development. We believe Evergrow is poised to drastically reduce both the cost and time to deploy offset projects while improving the overall health of carbon markets by kickstarting a more robust financing environment, providing pricing clarity and stability. While we are in the early days, we are excited to partner with James and Luke alongside the XYZ and First Round teams to raise the bar in the carbon markets.

Nicholas Adeyi is an Investment Analyst at Congruent Ventures, an early stage venture firm founded in 2017 that partners with companies in the sustainable technology ecosystem.

[i] https://www.nature.com/articles/d41586-021-02606-3

[ii] https://www.ecosystemmarketplace.com/publications/state-of-the-voluntary-carbon-markets-2021/

[iii] https://www.scientificamerican.com/article/cost-of-carbon-pollution-pegged-at-51-a-ton/

[iv] https://apps.ipcc.ch/outreach/documents/451/1551801374.pdf

[v] https://www.nature.com/articles/d41586-021-02606-3

[vi] https://carbonware.substack.com/p/carbon-credit-sales-are-booming-a

We founded Congruent Ventures to reboot and reinvigorate climate-tech venture ecosystem by focusing on the first institutional venture rounds in companies often before product launch or revenue generation. As a result of our early-stage focus, we approached the quantification of our portfolio’s impact cautiously. We have resisted creating reports with false precision and grandiose claims. Instead, we have labored to capture the potential impact of our companies effects at scale. With our maturing and more seasoned portfolio, we believe now is the time to share what our portfolio has accomplished and give a sense of what is to come. In aggregate, our current portfolio, once scaled, can reduce global GHG emissions by 10s of gigatons of CO2-equivalent and sequester gigatons more — and importantly do so while bringing value, efficiency, and cost savings to their respective sectors. It is a solid start, but we need our companies (and others) to scale to meet the challenge.

We are pleased to share Congruent Ventures inaugural Climate Impact Report for 2020.

Our Portfolio at Congruent Ventures covers a variety of sectors across four themes: Mobility and Urbanization, the Energy Transition, Food and Agriculture, and Sustainable Production and Consumption. Within these themes, we invest in all business models including fintech, frontier tech, software, and services. Investing against the climate change thesis is both a broad and deep challenge, creating additional challenges to quantifying impact. A core tenet of our firm is that the greatest underutilized tool for impact is a financial tool: there exists above-market financial returns for those with the appropriate investment skillset. We have leveraged our team’s depth and our network’s breadth to develop fundamental views not just on technology and product but how companies and products get deployed, making optimal use of all tranches of the capital stack.

As of the end of 2020, we’ve catalyzed over $362M of capital into climate-related companies. Despite writing the first institutional check into four out of five of our portfolio companies with a weighted investment period of only 17 months, over two thirds of our companies have already launched or released their respective products and services. We have brought in or joined forces with over 186 different co-investors including numerous mainstream firms, which we are particularly pleased to see participate in the climate ecosystem. Despite the young age of our portfolio, 69% of our portfolio companies have raised follow-on capital since our initial investment.

Beyond our core climate focus, we value diversity. One out of four of our portfolio companies are women-led. While better than the average of one out of 50 in venture-backed startups, it is still a far cry from the 50% mark. We will continue to focus on adding diversity to our portfolio and our team.

We are proud to share our Climate Impact Report and will be updating it on an annual basis to reflect on our growing portfolio and progress across our companies. As climate-tech enters the mainstream, we continue to believe a bigger tent benefits all. Please reach out with feedback or comments.

Sustainably,

The Congruent Team

Congruent Ventures invests in next generation climate and sustainability entrepreneurs at the earliest stages of company formation: Pre-Seed, Seed, and Series A. As our friends in the venture and entrepreneurial ecosystem know, the data indicate that the probability of an investment is higher when companies are introduced by a trusted source. We acknowledge that our historical networks are predominantly white, educated at a small number of schools and generally located in a few coastal cities. We also recognize that the venture industry has consistently shown implicit structural bias. We must break this cycle deliberately and thoughtfully; words will only get us so far. One of the best digestible summaries of this problem was penned in a fantastic article by Del Johnson here.

It is tragic that it took the on-camera murder of George Floyd and the countless (and ongoing) acts of police brutality to focus attention on America’s culture of structural racism, and in turn, our attention on our own bias. Partly because of our own largely homogeneous network, we are part of the problem. We are taking two immediate steps at Congruent to begin addressing our own shortcomings related to investing in Black and underrepresented founders.

First, we are reaching out to organizations, universities, and institutions that are predominantly Black with the goal of permanently diversifying our trusted network. While this will be a firm-wide undertaking, we will be dedicating a person to focus on this effort.

Second, while we currently endeavor to review all inbound emails regardless of source, we will be actively working to meet with North America-based early stage climate and sustainability founders who are Black and people of color. Please email us at investors@congruentvc.com. Our entire investment and ops team is on this email.

This is admittedly a small step and we will continue to self-reflect on what we can and should be doing better. As always, we welcome constructive feedback.

Yours in Solidarity,

-The Congruent team

You can read some details from Greentech Media, Axios, GreenBiz, and Bloomberg Environment, and we will be sharing more about our specific investments in the coming months.

The investments we make in hardware, software, enterprise, consumer, deep technology, fin-tech, and business model innovation will have a positive environmental impact while creating venture-grade returns.

Some of our major themes are Urbanization and Mobility trends, the Clean Energy Transition, Food and Agriculture, and Industrial and Supply Chain innovation. We have invested in nine companies to date ranging from additive manufacturing to LIDAR to improving the impact of fashion on the environment and expect to make many more.

We raised significantly more capital than we originally contemplated largely due to the support of our founding Limited Partners. The University of California’s Office of the Chief Investment Officer and Prelude Ventures had the vision to reinvigorate early stage investment in the sector while acknowledging the challenges of the past. We look forward to working with UC’s broader community, Prelude, and our other Limited Partners and thank them for their support and thought leadership.

Our journey will carry us forward for many decades and we’re thrilled to be working with the best and the brightest to address global sustainability challenges.

Josh & Abe, Founders and Managing Partners, Congruent Ventures