Benchmark Raises $225 Million Special Fund, Doubles Down on Cerebras to Challenge Nvidia

Venture Capital Giant Benchmark Doubles Down on AI Chip Track

According to TechCrunch, on February 7, 2026, Benchmark Capital announced the successful raising of a $225 million special fund specifically to deepen its investment in AI chip company Cerebras Systems. This move marks a further escalation of Benchmark's ongoing bet on the company since 2016. At a time when Nvidia firmly dominates the AI GPU market, Benchmark's decision is undoubtedly a strong endorsement of Cerebras' technology roadmap and market prospects.

Benchmark Capital has been an investor in the Nvidia rival since 2016.

As a top-tier Silicon Valley venture capital firm, Benchmark is known for its precise investment vision, having invested in unicorns such as Uber and Snapchat. The establishment of this special fund demonstrates its deep deployment in the AI hardware sector. Cerebras was founded in 2016, headquartered in Los Altos, California, by former AMD executive Andrew Feldman. The company's core product is the world's largest monolithic AI chip—the Wafer Scale Engine (WSE), whose area is equivalent to thousands of traditional GPUs combined, significantly accelerating AI model training.

Cerebras' Technological Revolution: Wafer-Scale Chips Disrupt Tradition

Unlike Nvidia's GPU cluster approach, Cerebras adopts a wafer-scale integrated circuit design, transforming an entire silicon wafer into a single processor. The second-generation WSE-2 features 850,000 AI cores, 40GB of HBM2e memory, and a total computing power of up to 4 Petaflops. This design greatly reduces data transmission latency between chips, solving bottlenecks in traditional AI training. For example, when training large language models, the Cerebras system can shorten the time from weeks to hours.

Industry data shows that the global AI chip market is expected to exceed $200 billion by 2028. Nvidia, with its H100 and Blackwell series GPUs, holds over 90% market share, but high costs and energy consumption issues are becoming increasingly prominent. Cerebras' CS-3 system (based on WSE-3) has further increased computing power to 125 Petaflops FP8, integrating liquid cooling technology for higher power efficiency. The company has partnered with multiple research institutions and enterprises, such as Los Alamos National Laboratory, for high-performance computing tasks like weather simulation and drug discovery.

Benchmark's Investment Timeline and Strategic Considerations

Benchmark led a $30 million investment in Cerebras' Series A round as early as, and has since participated in multiple rounds, totaling over $700 million. This $225 million special fund is not an ordinary follow-on investment but is earmarked specifically to support Cerebras' next-generation product development and capacity expansion. Benchmark partner Peter Fenton stated: "Cerebras is not following Nvidia; it is redefining the boundaries of AI computing."

This investment background stems from the generative AI boom. Since ChatGPT exploded in popularity, AI training demand has grown exponentially, leading to severe GPU shortages in data centers. Competitors such as Graphcore and SambaNova are also making efforts, but Cerebras' wafer-scale architecture is the most disruptive. Benchmark's increased bet also reflects the VC circle's expectations of a "post-Nvidia era": a diversified chip ecosystem will rise, reducing dependence on a single supplier.

AI Chip Market Volatility: Opportunities and Challenges Coexist

Looking back at the history of AI chip development, 2016 was a key turning point. That year, the concepts of TPUs and dedicated AI accelerators emerged, and Cerebras was a pioneer of this wave. Benchmark's investment is not blind but based on Cerebras' milestones: the launch of the CS-2 system in 2021, the introduction of the WSE-3 in 2024, and investment from Middle Eastern consortiums such as G42.

However, challenges remain. High wafer manufacturing costs (relying on TSMC) and software ecosystem development are pain points. Nvidia's CUDA platform has deep moats, and while Cerebras' software stack is compatible with PyTorch/TensorFlow, developer migration takes time. In addition, geopolitical risks such as the US-China chip war also affect the supply chain.

Nevertheless, market optimists believe Cerebras has unique advantages. Its Condor Galaxy supercomputer cluster has reached 4 Exaflops, surpassing many national-level supercomputers. In the future, as edge AI and real-time inference demand grows, Cerebras may break through in the enterprise market.

Editor's Note: Benchmark's Foresight Signals a New Era in AI Hardware

As an AI tech news editor, I believe Benchmark's $225 million special fund is not just a capital injection but also a bellwether for the industry. It signals top VCs' renewed confidence in pure hardware innovation. In the software-dominated AI era, underlying computing power remains king. If Cerebras can overcome the ecosystem barrier, it will form a five-way competition with Nvidia, AMD, and Intel, promoting AI democratization. However, investors need to be wary of valuation bubbles—Cerebras is valued at over $4 billion, and its path to profitability remains unclear. This move may trigger more capital inflows, accelerating the AI chip arms race.

This article is approximately 1,050 words, compiled from TechCrunch, original author: Marina Temkin, date: 2026-02-07 13:23:04.