The Real Bottleneck in AI Isn't Code — It's Power
Artificial intelligence doesn't run on clever prompts and open-source weights alone. It runs on electricity. Massive, continuous, always-on electricity. And if the agenda at TechCrunch Disrupt 2026 is any indication, the energy crunch is about to become the single most talked-about constraint on AI's next wave.
October 13–15, right here at San Francisco's Moscone Center, the Smart Systems Stage is going to pull back the curtain on exactly how startups, utilities, and fusion companies plan to keep AI from hitting a hard ceiling. We're talking commercial fusion, grid modernization, fuel cells, and data center energy expansion. None of these are abstract concepts anymore. They're happening now.
How Fusion Companies Plan to Plug Into the Grid
Fusion has spent decades as a "coming soon" story. But two of the best-funded players on the planet are making concrete claims about delivering electricity within this decade, and they're bringing those claims to the stage at Disrupt.
David Kirtley, Helion's CEO, will sit down with Brandon Sorbom, Commonwealth Fusion Systems' Chief Science Officer, to walk through what it actually takes to get fusion onto the grid at scale. Their companies have taken radically different technical approaches — Helion with its pulsed plasma-magnetic confinement strategy, CFS with tokamak-based magnetic confinement — but they share one thing: the belief that fusion is no longer a future idea, but a path to clean, reliable, always-on electricity.
Helion's case has gotten harder to ignore. In June 2026, the company closed a $465 million Series G round led by Thrive Capital, pushing its post-money valuation to $15.5 billion. Total investment now stands at $1.5 billion. That kind of capital isn't just about credibility; it's about manufacturing capacity. Helion is building Orion, its first actual fusion power plant, in Malaga, Washington. The company's 7th-generation prototype, Polaris, already became the first privately funded fusion machine to operate with deuterium-tritium fuel, hitting plasma temperatures exceeding 150 million degrees Celsius.
Then there's Inertia, the fusion startup founded by former Twilio CEO Jeff Lawson. Lawson's pitch at Disrupt will be a candid fireside chat on scaling a fusion company — and how his experience scaling Twilio's infrastructure informs his approach to talent, timelines, and the brutal engineering questions that come with building a power plant from scratch. It's a fascinating crossover: the guy who taught a generation of developers how to scale communication APIs is now trying to scale physics.
The takeaway? Fusion is moving from lab experiments to construction sites, and the people building these plants are betting their careers — and billions in investor capital — on delivering electricity within this decade.
Rewiring the Grid for What Comes Next
Electricity demand is growing faster than the infrastructure built to support it. That's the blunt reality facing grid operators, utilities, and the startups trying to modernize an aging system — a challenge that's already pushing American power grids to their limits, as the 200-gigawatt surge driven by AI expansion makes clear.
Drew Baglino, founder and CEO of Heron Power, and Apoorv Bhargava, CEO and co-founder of WeaveGrid, will lead a panel on exactly what it takes to rewire the grid for the electric age. WeaveGrid, in particular, has been working on demand response and grid flexibility tools that could make a real difference during peak load events. Heron Power is focused on grid-edge resilience and microgrid solutions. Between them, they represent two sides of the modernization coin: flexibility on the demand side, resilience on the supply side.
The panel will also cover where investment is flowing. It's not just about building new transmission lines — though we need plenty of those. It's about software-defined grids, real-time load balancing, and making infrastructure flexible enough to handle the kind of demand spikes that AI data centers create. If you've ever watched a data center come online and pull hundreds of megawatts from the grid, you know why this matters.
Feeding the Data Center Beast
As compute demand skyrockets, data center operators and energy companies are in a race to secure power before it becomes the bottleneck that slows AI's next wave. This isn't hypothetical. Several major cloud providers have publicly acknowledged that power availability, not compute capacity, is now their primary constraint — a trend underscored by Google's 37% electricity spike driven by its AI buildout.
Sara Spangelo, president and co-founder of Ambrosia Energy, and Bill Thayer, SVP of Head of Datacenter Solutions at Bloom Energy, will tackle this problem head-on. Ambrosia Energy focuses on energy-as-a-service models for data centers, bundling power procurement, on-site generation, and grid connectivity into single contracts. Bloom Energy brings its fuel cell technology to the table, offering distributed, on-site power generation that can supplement grid supply.
The dynamic here is interesting. Data centers can't wait ten years for grid upgrades. They need power now. So companies like Ambrosia and Bloom are essentially acting as intermediaries — bridging the gap between immediate compute demand and long-term infrastructure buildout. It's a stopgap, sure, but a necessary one.
Why This Stage Matters
Whether you're building the next energy breakthrough, rethinking grid infrastructure, or just trying to understand what's really constraining AI's growth, the Smart Systems Stage is built for founders and operators who need the full picture. Not just a headline and an LLM summary. The actual engineering, the actual capital, the actual timelines.
TechCrunch Disrupt 2026 runs October 13–15 at the Moscone Center in San Francisco. The Smart Systems Stage is one of several tracks, but it's arguably the most consequential for anyone building AI infrastructure. Because at the end of the day, no amount of model optimization can compensate for a power shortage — and as analysts warn about the grid stability risks of concentrated AI workloads, the stakes extend far beyond simple capacity planning.
Register for Disrupt if you want to see these conversations play out in real time. The people on this stage aren't theorizing about the future of energy. They're building it.
