The CEO of Cerebras Systems, Andrew Feldman, is scheduled to appear on stage at TechCrunch Disrupt 2026, where he will raise a question that has become central to the entire field: whether artificial intelligence can continue growing in size and power.
Feldman’s presentation centers on the increasing need for processing power, energy, and underlying systems, and he will describe how Cerebras is handling these limitations in a distinct way. The talk, “Can AI Keep Scaling?”, is part of 200+ sessions at the three-day gathering in San Francisco.
A Founder Betting on Wafer-Scale Computing
After years spent building companies around computing infrastructure, Feldman co-founded Cerebras in 2015. Prior to that, he co-founded and led SeaMicro, an energy-efficient microserver startup that AMD acquired in 2012. He also held leadership positions at Force10 Networks and Riverstone Networks before turning his attention to founding Cerebras.
Cerebras faced a challenge that had long been dismissed as impractical: wafer-scale computing. Instead of slicing a silicon wafer into individual chips, the company built a processor directly on the wafer, an architecture created with demanding AI workloads in mind.
The wager succeeded. The firm took in $5.5 billion during its May debut on the market and sealed a long-term arrangement with OpenAI to supply substantial processing for artificial intelligence work. Then, in August, Cerebras unveiled CS-4, the newest model of its wafer-scale AI architecture.
The Infrastructure Behind the Compute
Just having stronger processors doesn’t fix the growth issue. The whole setup relies on data centers, power supplies, cooling systems, and production capacity.
Cerebras is already dealing with that challenge. The company reported in August that it had more than 600 megawatts of data center capacity live or under contract for delivery by the end of 2027. It also said it was raising manufacturing capacity more than tenfold during 2026.
This year, the firm aims to activate its first European data center space and then grow that facility to reach 200 megawatts of capacity by the close of 2027.
| Aspect | Cerebras’ Approach |
|---|---|
| Processor design | Wafer-scale architecture, built on the wafer itself |
| Data center capacity | More than 600 megawatts live or under contract by 2027 |
| Manufacturing capacity | Increasing more than tenfold in 2026 |
| European expansion | First data center online this year, 200 MW by end of 2027 |
What Happens When Hardware Hits Its Limits
The session gives founders a chance to hear from someone who has spent more than a decade backing an alternative approach to building AI hardware. For founders, investors, and technology leaders weighing choices about AI infrastructure, he puts those constraints into context: where compute demand is headed, what it takes to support it, and where today’s hardware could hit its limits.
The Disrupt event takes place October 13-15 at Moscone West in San Francisco. Over 10,000 founders, investors, operators, and tech leaders are expected to attend, along with more than 250 speakers and 300+ exhibiting startups.
Why This Session Matters
More than just building a faster processor goes into scaling AI. The physical infrastructure has to be there too, so that the compute can actually get to work.
Feldman will look at how rising demand for computing power, energy, and infrastructure is affecting AI’s future. He’, and he’ll examine the consequences should standard hardware fail to match that growth.
For anyone building, funding, or deploying AI, it’s a chance to hear from a founder testing a different approach to one of the industry’s biggest constraints.
Register for Disrupt to hear what Cerebras’ experience can tell us about the next phase of AI scale. Bring a co-founder, colleague, or peer at 50% off.
Source material: “Cerebras Systems’ Andrew Feldman on whether AI can keep scaling at TechCrunch Disrupt 2026,” TechCrunch.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.