Cerebras Raises $1.1bn to Delay IPO, Doubling Valuation as AI Chip Race Intensifies

Cerebras is buying itself more time before an IPO by locking in $1.1 billion in new private funding, a move that doubles its valuation to $8.1 billion and positions the startup as one of the few challengers to Nvidia in the red-hot market for AI chips.

The company filed to go public exactly a year ago, but ran into headwinds after U.S. regulators raised concerns about its reliance on a single Middle Eastern customer, G42. The Treasury Department’s Committee on Foreign Investment in the United States (CFIUS) reviewed Cerebras’ plan to give G42 a bigger stake, slowing down its path to market. That dependence had also drawn scrutiny from analysts who flagged it as a potential vulnerability.

Despite those setbacks, investor enthusiasm has only grown. Co-founder and CEO Andrew Feldman confirmed that Cerebras still intends to go public, but that raising fresh funds was necessary to seize opportunities in the fast-changing AI landscape.

“I don’t think this is an indication of a preference for one or the other,” Feldman said in an interview. “I think we have tremendous opportunities in front of us, and I think it’s good practice, when you have enormous opportunities, not to let them fall by the wayside for lack of capital.”

Feldman has previously said the company aspires to list in 2025.

The funding round included heavyweight investors such as 1789 Capital, Alpha Wave, Altimeter Capital, Atreides Management, Benchmark, Fidelity, Tiger Global, and Valor Equity Partners. Feldman described the lineup as one that could easily “cornerstone your IPO,” underscoring the caliber of backing the company now enjoys.

Much of the new money will go toward expanding U.S. manufacturing capacity. Cerebras’ chips are produced as wafers by Taiwan Semiconductor Manufacturing Co. (TSMC) and then packaged in the United States.

“We increased manufacturing capacity in the last 18 months 8x, and we are going to go another 4x in the next six or eight months,” Feldman said, adding that the company will also hire aggressively to meet demand.

Financially, Cerebras is beginning to show traction. The company generated about $70 million in revenue in the second quarter of 2024, a leap from less than $6 million in the same period a year earlier. It has recently won business from Hugging Face, Meta, Notion, and Perplexity—clients that signal growing trust in its technology.

Cerebras’ rise comes as private capital continues to flow into AI. Databricks, which sells data analytics software, raised $1 billion at a valuation above $100 billion. OpenAI recently disclosed that Nvidia plans to invest up to $100 billion to support its buildout of data centers. Anthropic, another AI startup, secured $13 billion at a valuation of $183 billion. Against those eye-popping numbers, Cerebras looks modest, but its focus on AI hardware—rather than software or platforms—places it in direct competition with Nvidia, a $3 trillion company whose chips dominate training and inference workloads.

The contrast highlights the different paths emerging in the AI arms race. While OpenAI and Anthropic are soaking up billions to expand data center capacity and scale software models, Cerebras is betting on specialized chips that can serve as alternatives to Nvidia’s GPUs. That strategy is capital-intensive but also positions the startup at the very heart of AI infrastructure, where demand for compute power is exploding.

Nvidia’s dominance is believed to have wetted investors’ appetite for a challenger. Feldman, for his part, framed Cerebras’ funding not as a retreat from public markets but as a bridge to ensure it has the capital needed to compete.

“When you have enormous opportunities,” he said, “it’s good practice not to let them fall by the wayside for lack of capital.”

Although Cerebras remains private for now, with revenues surging, investors circling, and competition intensifying, its eventual IPO is likely to test just how much appetite remains for high-risk, high-reward bets in the AI hardware space.