The AI hardware industry is experiencing a foundational shift. The recent landmark $20 billion agreement between NVIDIA and inference-specialist Groq is more than a deal – it’s a strategic admission that the era of GPU dominance for all AI workloads is ending. In this new landscape, where high-speed, efficient model inference is becoming the critical bottleneck and a vast market opportunity, companies built with specialized architectures are poised to thrive.
One such company, Cerebras Systems, is moving decisively to capitalize on this inflection point. With its revolutionary wafer-scale engine, the company has solidified its position as a leading disruptor, recently closing a Series H round at a staggering $23 billion valuation, nearly triple its previous valuation from just months earlier. This round was led by Tiger Global, with participation from Benchmark, Fidelity Management & Research Company, Atreides Management, Alpha Wave Global, Altimeter, AMD, Coatue, and 1789 Capital among others. This renewed momentum follows its earlier $1.1 billion Series G round at an $8.1 billion valuation and comes as the company has confidentially filed for a US IPO, targeting a public listing as early as April 2026.
Adding to its momentum, Cerebras recently secured a landmark deal with OpenAI, reported to be worth over $100 billion, to provide 750 megawatts of computing power through 2028 a clear signal that major AI players are actively seeking alternatives to Nvidia’s hardware.
I sat down with Cerebras' co-founder and CEO, Andrew Feldman, following this landmark round. In a wide-ranging conversation, he detailed the company's path from a clean-sheet idea to a global hardware disruptor, the surprising epicenters of its growth, and the pragmatic steps toward an impending IPO.
From Boredom to Building the Engine of AI
Andrew Feldman's path to founding Cerebras wasn't linear. After a successful exit of a previous data center company SeaMicro to AMD, he found himself at a crossroads. "I was bored," he stated simply. The spark came from colleagues in 2015, with a prescient insight about the future, fueled by a recognition that existing compute was fundamentally misaligned with the coming AI wave. "We saw AI on the horizon… And we knew that the graphics processing unit would probably not be the right machine," he said. The insight was to start from zero. "If we started with a clean sheet of paper and designed a solution optimized for AI, not for graphics, not for databases, not for web serving, but just for AI, we could build a better solution."
That conviction attracted immediate backing. "We went out to raise money in March 2016. We made eight presentations, we got eight term sheets, and so we ...
