2026-05-05

Cerebras Systems prices CBRS on Nasdaq May 13 — what the $25B AI chip IPO means

Cerebras Systems prices CBRS on Nasdaq May 13 at a $24.5–26.6B valuation. WSE-3 maker reported $510M revenue in 2025 and holds a $10B OpenAI deal.

Cerebras Systems is pricing its Nasdaq IPO on May 13, 2026. The offering range values the company at $24.5–26.6 billion — making it the largest pure-play AI chip IPO since Nvidia went public in 1999.

What Cerebras actually is

Cerebras makes the WSE-3 (Wafer Scale Engine 3), a single-chip processor that covers an entire silicon wafer — 900,000 AI cores on a chip roughly the size of a dinner plate. The architectural bet is that eliminating chip-to-chip communication latency unlocks inference performance that GPU clusters can’t match for certain workloads.

The company’s flagship product, Cerebras Inference, processes LLaMA-3.1-70B at over 2,100 tokens/second — roughly 20× faster than typical H100 cloud endpoints. For real-time use cases (voice AI, live translation, interactive coding), this latency advantage is meaningful.

The numbers behind the IPO

Cerebras reported $510 million in revenue for 2025, up sharply from $79M in 2024. The growth driver: a $10 billion multi-year compute deal with OpenAI, which buys Cerebras inference capacity and represents roughly 85% of 2025 revenue. That concentration is the key risk the S-1 flags.

The company is not yet profitable. Gross margins are positive (~60%), but R&D and sales burn keeps it in the red.

The CFIUS backstory

Cerebras filed its first S-1 in September 2024, but the IPO was blocked because G42, a UAE sovereign-linked fund, held a 16% stake — triggering a CFIUS national security review. G42 subsequently sold its Cerebras stake in early 2025, clearing the path. The May 2026 offering is the second attempt.

What to watch at pricing

Revenue diversification: Can Cerebras land enterprise customers to reduce OpenAI dependency before lock-up expiry?
NVIDIA response: H200/B100 inference performance has closed the latency gap at higher batch sizes. The WSE-3 advantage is most pronounced in low-batch, real-time scenarios.
Valuation: At $26B, CBRS prices at roughly 51× trailing revenue — a rich multiple that demands continued hypergrowth.

Practitioner note

If you deploy inference infrastructure, watch the CBRS debut. Cerebras’s claim to 20× inference speed advantage over H100 is real for specific workloads — but the benchmark conditions matter (batch size, model size, context length). Their cloud API is live and metered; running a comparison against your actual workload costs about $5 in API credits. Worth doing before the post-IPO hype cycle makes the data harder to read objectively.