Cerebras Systems CEO Andrew Feldman is scheduled to speak at TechCrunch Disrupt 2026, addressing the growing constraints of compute, energy, and infrastructure in the AI sector. The session, titled “Can AI Keep Scaling?,” will explore how Cerebras’ wafer-scale computing approach challenges conventional chip architectures and examines the physical limitations facing today’s AI hardware.
What Happened
Andrew Feldman, who co-founded Cerebras in 2015, will present on the Disrupt Stage at Moscone West in San Francisco. Cerebras distinguishes itself by building processors directly on silicon wafers rather than cutting them into individual chips, a method designed specifically for demanding AI workloads. The company recently introduced its CS-4 wafer-scale AI infrastructure in August.
Since its May IPO, which raised $5.5 billion, Cerebras has expanded its operational footprint significantly. The company signed a multiyear agreement with OpenAI to deploy 750 megawatts of Cerebras systems from 2026 through 2028. In August, Cerebras reported having more than 600 megawatts of data center capacity either live or under contract for delivery by the end of 2027. Additionally, the company stated it is increasing its manufacturing capacity more than tenfold during 2026 and plans to bring its first European data center capacity online this year, aiming for 200 megawatts in Europe by the end of 2027.
Why It Matters
The discussion highlights that scaling AI capabilities requires more than just faster processors; it demands substantial physical infrastructure, including electricity, cooling, and manufacturing capacity. Feldman’s session aims to provide context for founders, investors, and technology leaders regarding where compute demand is heading and what is required to support it. As AI models become increasingly capable, the bottleneck often shifts from algorithm design to the availability of physical hardware and energy resources.
Cerebras’ experience with wafer-scale computing offers a case study in alternative hardware approaches. By challenging the assumption that powerful AI must rely on conventional chip architectures, the company provides insight into potential solutions for the industry’s infrastructure constraints. The event, taking place October 13-15, expects more than 10,000 attendees and features over 200 sessions across six stages, serving as a key venue for discussing the next phase of AI scale.
The Bottom Line
Andrew Feldman’s appearance at TechCrunch Disrupt 2026 underscores the critical link between AI model development and physical infrastructure. With Cerebras reporting significant data center contracts and manufacturing expansions, the company’s trajectory illustrates the scale of investment and engineering required to meet rising compute demands. The session will focus on whether current hardware approaches can sustain AI growth or if fundamental shifts in infrastructure are necessary.