Cerebras funding, valuation and investors

Latest tracked event: IPO (May 2026). It is shown separately because it is not counted as disclosed equity funding.

Overview

Cerebras Systems designs and sells AI compute built around the Wafer-Scale Engine (WSE-3) — a single chip the size of a silicon wafer with 4 trillion transistors. Its CS-3 systems cluster for training and supercomputing; Cerebras Inference Cloud offers an OpenAI-compatible API for production LLM serving with claimed up to 15× faster inference than GPU clouds on supported models. Founded in 2015 in Sunnyvale, the company serves hyperscalers, foundation-model labs, enterprises, and sovereign AI programs. After raising $1B at ~$23B in February 2026 led by Tiger Global, Cerebras completed its Nasdaq IPO (CBRS) in May 2026, raising roughly $6.4B gross at $185/share per SEC filings.

Why Cerebras is interesting

Cerebras went public at $185/share in May 2026 after two consecutive $1B+ rounds, betting that wafer-scale silicon (WSE-3) can beat GPU fleets on inference latency and tokens-per-dollar — a separate investable layer from Nvidia supply chains.

Product & use cases

Cerebras sells CS-3 wafer-scale AI supercomputers for training and an inference cloud/API that runs open and proprietary LLMs on WSE-3 hardware — targeting teams that need lower latency and higher throughput than commodity GPU fleets.

• Production LLM inference for agents, coding assistants, and real-time chat at scale
• Training and fine-tuning large models on clustered CS-3 systems
• Sovereign and enterprise AI deployments requiring on-prem or regional inference

Key facts

• IPO (May 2026): ~$6.4B gross at $185/share on Nasdaq (CBRS) per SEC 8-K
• Series H (Feb 2026): $1B at ~$23B valuation led by Tiger Global (Reuters)
• Series G (Sep 2025): $1.1B at $8.1B led by Fidelity and Atreides (company press release)
• WSE-3: 4T transistors, 125 petaflops; inference cloud with OpenAI-compatible API
• Partnerships include OpenAI open-model hosting, AWS Bedrock, and AMD Helios integration

Funding history (newest first)

Investors in our directory

Funds linked from Cerebras's profile — open a fund page for stage focus and related deal articles.

Competitive landscape

Edge: Single-wafer WSE-3 eliminates GPU memory-bandwidth bottlenecks for certain workload shapes; Cerebras has held daily inference speed benchmarks versus Nvidia GPUs since launching its inference service in late 2024, per company claims.

Specialty AI accelerators compete on inference economics as production AI shifts from training to 24/7 serving. Cerebras leads on wafer-scale differentiation and went public to fund scale, but faces Nvidia's ecosystem depth, hyperscaler silicon, and customer concentration risk disclosed in its S-1.

• Nvidia incumbent

  Volume GPU leader (Blackwell and successors); broader software ecosystem but higher latency on some inference benchmarks Cerebras cites.

• Groq direct

  Inference-specialized LPU architecture; similar tokens-per-dollar thesis but different silicon approach.

• SambaNova Systems direct

  Full-stack AI hardware and software for enterprise training and inference.

• AWS Inferentia / Trainium incumbent

  Hyperscaler captive silicon bundled with cloud contracts; Cerebras also partners with AWS for hybrid deployments.

Notable stories

• Cerebras filed for IPO in 2024, withdrew in October 2025 after raising Series G, then returned to public markets in May 2026 at $185/share — one of the largest AI hardware listings of the cycle (SEC 8-K, May 2026).
• The company partnered with OpenAI to host gpt-oss-120B at claimed world-record inference speeds on wafer-scale infrastructure (Business Wire, August 2025).

Industries

Market / IPO context

Related funding articles

Venture Capital Tracker pieces that cover Cerebras's financing or category context.

• Cerebras' $1B Round: The Inference Economy Needs Its Own Silicon