Cerebras CEO Urged Balanced AI Risk Perspective
Andrew Feldman called for weighing AI potential alongside safety concerns to avoid stalled progress.
Updated on Oct. 13, 2026 in Artificial Intelligence
Live Poll
Do you believe the development of artificial intelligence is likely to result in more benefits than risks?
At TechCrunch Disrupt 2026, Cerebras Systems CEO Andrew Feldman argued that current industry discourse fixates too heavily on existential risks. He contended that while misuse remains a legitimate concern, neglecting the upside of powerful technology threatens to overshadow meaningful progress.
Why it matters
Feldman suggested that the industry's struggle with public perception—particularly regarding the energy-intensive nature of data centers—is a direct consequence of failing to engage local communities. Addressing this friction is essential for scaling infrastructure as global demand for specialized compute grows.
Cerebras Systems expects its total operational and contracted data-center capacity to exceed 600 megawatts by the end of 2027. Additionally, the company projects its manufacturing capacity will increase more than tenfold throughout 2026.
The players
Andrew Feldman
CEO of Cerebras Systems, a chip manufacturer specializing in wafer-scale processors for large-scale AI training.
Cerebras Systems
A developer of AI hardware, including the WSE-3 processor, focused on high-performance computing for training large models.
The details
Feldman emphasized that future efficiency gains must occur simultaneously across both hardware and software layers. By decoupling hardware vendor optimizations from model-level software improvements, the industry can reduce the footprint of massive compute clusters. He pointed to nations such as Finland, Iceland, and Norway as successful models for data center integration, noting that successful expansion requires better alignment with the needs of the local community.
Timeline
October 13, 2026: Andrew Feldman spoke at TechCrunch Disrupt 2026.
2026: Cerebras projects manufacturing capacity will grow tenfold.
End of 2027: Cerebras expects data-center capacity to exceed 600 megawatts.
The Tech Race
Feldman’s call for community integration marks a shift in how infrastructure providers are beginning to address the friction currently hindering the international effort to build large-scale data centers for artificial intelligence. The approach contrasts with current deployment struggles in the United States and highlights successful collaborative models in countries like Norway and Iceland.
Users should look for a potential increase in AI training capability as Cerebras scales its manufacturing output tenfold this year. The company's expansion plans suggest that larger, more efficient models may reach production environments as data center capacity hits the 600-megawatt mark.
The takeaway
The industry is moving toward a strategy of local collaboration to overcome the infrastructure bottlenecks currently slowing data center deployments. Watch the end of 2027 as a key milestone for Cerebras to confirm its 600-megawatt operational capacity target.
Further reading
For more on the hardware driving model development, see Artificial Intelligence.
Source note: This article includes information reported by TokenPost.
Live Poll
Do you believe the development of artificial intelligence is likely to result in more benefits than risks?
