Recently, AI semiconductor company Cerebras Systems disclosed a successful capital raise of $1 billion, pushing its valuation to $23 billion. This figure represents an almost threefold leap from the $8.1 billion valuation its competitor to Nvidia achieved merely half a year prior.
Although Tiger Global spearheaded this funding round, a significant portion of the fresh investment originated from Benchmark Capital, an early supporter of the company. A source close to the transaction revealed that the notable Silicon Valley venture firm contributed a minimum of $225 million to Cerebras’s most recent financing.
Benchmark initially backed the decade-old Cerebras by leading its $27 million Series A round back in 2016. As Benchmark intentionally manages funds below $450 million, regulatory documents indicate the firm established two distinct investment vehicles, both named ‘Benchmark Infrastructure.’ The individual familiar with the deal confirmed that these vehicles were exclusively formed to support the Cerebras investment.
Benchmark chose not to provide a statement.
The remarkable physical size of its processors distinguishes Cerebras. Its flagship chip, the Wafer Scale Engine, unveiled in 2024, measures roughly 8.5 inches per side and integrates 4 trillion transistors onto a single silicon component. For context, this chip is produced from almost a complete 300-millimeter silicon wafer—the foundational circular substrates for all semiconductor manufacturing. While conventional chips are tiny pieces diced from these wafers, Cerebras uniquely utilizes nearly the entire circular disc.
This design incorporates 900,000 specialized cores that operate simultaneously, enabling the system to execute AI computations without the need to transfer data between numerous discrete chips, which is a significant constraint in typical GPU clusters. The company asserts that this architecture facilitates AI inference operations at speeds over 20 times greater than rival solutions.
This financing arrives as Sunnyvale, California-based Cerebras accelerates its progress in the competitive AI infrastructure sector. Just last month, Cerebras secured a multi-year contract exceeding $10 billion to supply 750 megawatts of computational capacity to OpenAI. This collaboration, set to continue until 2028, is intended to enhance OpenAI’s ability to provide quicker responses for intricate AI requests. (Notably, OpenAI CEO Sam Altman also holds an investment in Cerebras.)
Cerebras asserts that its systems, constructed around its specialized chips engineered for AI applications, outperform Nvidia’s offerings in terms of speed.
The company’s journey toward an initial public offering has been intricate due to its association with G42, an AI company based in the UAE, which constituted 87% of Cerebras’s revenue during the first half of 2024. G42’s previous connections with Chinese technology firms prompted a national security assessment by the Committee on Foreign Investment in the United States (CFIUS), which delayed Cerebras’s original IPO timeline and even led the company to retract a prior filing in early 2025. By the close of last year, G42 was no longer listed among Cerebras’s investors, thereby paving the path for a renewed IPO endeavor.
Cerebras is currently targeting a public market launch during the second quarter of 2026, as reported by Reuters.