Cerebras Systems CEO Andrew Feldman is highlighting the physical and energy constraints limiting AI growth. As the company ramps up manufacturing and expands data centers to fulfill a massive deal with OpenAI, investors are tracking whether the company’s unique chip architecture can solve the industry’s hardware bottleneck.
Cerebras Systems CEO Andrew Feldman is set to address the growing physical and energy challenges facing the AI industry at the upcoming TechCrunch Disrupt. As the company continues to push its unique approach to hardware, its growth strategy is being tested by the very limits that Feldman argues are holding back the next generation of AI development.
For the past decade, the company has operated on a different philosophy than most of the industry. While standard graphics processing units, or GPUs, are diced from silicon wafers, Cerebras uses a technique called wafer-scale computing. This involves building processors directly on the entire silicon wafer. The company argues that this design helps bypass the bottlenecks that standard hardware faces when trying to manage massive AI workloads.
The firm is currently in a phase of aggressive expansion to support its technology. In August, Cerebras disclosed that it has over 600 megawatts of data center capacity either operational or under contract to be deployed by 2027. To meet this demand, the company is aiming to scale its manufacturing speed by more than ten times throughout 2026. Part of this growth includes international expansion, with its first European data center capacity expected to go live this year and a target to reach 200 megawatts in Europe by the end of 2027.
Financial and operational support for this growth is substantial. Following its $5.5 billion IPO in May, the company secured a multiyear agreement with OpenAI. This contract involves the deployment of 750 megawatts of systems between 2026 and 2028. To service this deal, the company recently launched the CS-4, its latest infrastructure platform designed to handle the compute requirements of modern generative AI models.
However, the company’s trajectory is not without significant business challenges. Scaling manufacturing throughput tenfold is a complex operational task that carries execution risk, especially when balancing the power and cooling requirements of large-scale data centers. The industry is currently grappling with severe energy shortages and infrastructure bottlenecks, which could act as a drag on deployment timelines. Investors are monitoring whether the company can execute on these ambitious capacity targets without facing cost overruns or delays.
The focus for the market will be on how effectively the firm manages these physical constraints. Investors may track the company’s ability to turn its announced capacity into operational revenue, the pace of manufacturing expansion, and whether its hardware can maintain a performance advantage as energy costs continue to influence the AI hardware sector.