Cerebras Systems CEO and co-founder Andrew Feldman will headline a session at TechCrunch Disrupt 2026 titled "Can AI Keep Scaling?" the conference announced.
The talk will address a question pressing the entire AI industry: whether the compute, energy, and infrastructure required to train and run increasingly capable models can keep pace with demand.
Feldman co-founded Cerebras more than a decade ago to challenge the assumption that powerful AI must run on conventional chip architectures. The company builds wafer-scale processors — chips the size of a dinner plate — that it says deliver faster training and inference than standard GPU clusters.
Cerebras sells its hardware through on-premise systems and a cloud platform. The company has positioned itself as an alternative to Nvidia, whose GPUs dominate the AI chip market.
The scaling question has become urgent as AI companies pour billions into data centers and power grids strain under the load. TechCrunch framed the session around what happens "if today's AI hardware reaches its limits."
Feldman is expected to explain how Cerebras approaches those constraints differently and what comes next for the industry. He has previously argued that wafer-scale design removes bottlenecks that slow conventional chips.
The session fits a broader debate at this year's conference. AI models keep improving, but each leap demands more electricity, more cooling, and more capital. Whether that trajectory is sustainable is now a central question for investors and executives.
Disrupt 2026 runs at the Moscone Center in San Francisco. The event draws more than 10,000 attendees, including founders, investors, and tech leaders. Ticket discounts of up to $200, plus 50% off a second ticket, end September 25 at 11:59 p.m. PT.
October 2 is the final day for companies to book an exhibit table to demo products to attendees. Registration and exhibit details are available on the TechCrunch Disrupt website.
Cerebras has raised hundreds of millions in venture funding and counts major enterprises and research labs among its customers. The company's cloud service lets developers access its chips without buying hardware.
Feldman's appearance adds to a speaker lineup focused on AI infrastructure, the economics of model training, and the limits of current technology. His talk is scheduled for the Disrupt Stage.
