Andrew Feldman, co-founder and CEO of Cerebras Systems, will speak at TechCrunch Disrupt 2026 on a session titled "Can AI Keep Scaling?" He will examine how continued advances in AI capabilities are driving rising requirements for compute, energy and infrastructure, and whether alternate hardware approaches can push those limits further.
Cerebras’ technical approach
For more than a decade Cerebras has challenged the conventional assumption that ever-more-powerful AI must rely on standard chip architectures. Since its 2015 founding, the company has pursued wafer-scale computing: rather than slicing a silicon wafer into many separate chips, Cerebras designs processors that occupy the wafer itself, an architecture tailored for demanding AI workloads.
Andrew Feldman came to Cerebras after building multiple companies in computing infrastructure. He co-founded SeaMicro (acquired by AMD in 2012) and previously held leadership roles at Force10 Networks and Riverstone Networks.
Recent milestones and agreements
Cerebras completed a $5.5 billion IPO in May, and signed a multiyear agreement with OpenAI to deploy 750 megawatts of Cerebras systems between 2026 and 2028. In August the company unveiled CS-4, the latest generation of its wafer-scale AI infrastructure.
Infrastructure is the broader constraint
More powerful processors alone do not solve the scaling challenge. Sufficient datacenter capacity, electrical supply, cooling and manufacturing are required to put compute to work. Cerebras reported in August that it has over 600 megawatts of datacenter capacity either live or under contract for delivery by the end of 2027, and said it will increase manufacturing capacity more than tenfold during 2026. The company also plans to bring its first European datacenter capacity online this year and expand to 200 megawatts there by the end of 2027.
These figures illustrate that scaling AI is as much about physical infrastructure as it is about chip design.
Why this matters to founders, investors and operators
Feldman’s session aims to help founders, investors and technology leaders understand where compute demand is heading, what it takes to support that demand, and where today’s hardware might reach its limits. Hearing from a founder who has long pursued a different hardware path provides direct insight into practical constraints — from energy and cooling to manufacturing lead times — that will shape the next phase of AI deployment.
Event details
TechCrunch Disrupt 2026 runs October 13–15 at Moscone West in San Francisco. The conference features more than 200 sessions across six industry stages, roundtables and breakouts; organizers expect over 10,000 founders, investors, operators and tech leaders, alongside 250+ speakers and 300+ exhibiting startups.
Feldman’s talk is one opportunity at Disrupt to learn whether wafer-scale and other nontraditional approaches can extend AI’s growth path when conventional hardware and supporting infrastructure start to constrain progress.
(Source: TechCrunch event materials and provided announcements.)
