Cerebras revenue jumps 92% as OpenAI compute deal exceeds $20 billion
Cerebras Systems reported a 92% increase in year-over-year revenue to $191.3 million, supported by a multi-year $20 billion inference agreement with OpenAI. The company is pivoting its business toward hardware-software integration for low-latency AI inference to compete with established chipmakers like NVIDIA and AMD.
Key Takeaways
- OpenAI agreement for 750 megawatts of inference capacity is now valued at over $20 billion through 2028.
- Cloud and services revenue surged 167% to $79.8 million, now representing 42% of total core revenue.
- New AWS partnership uses 'disaggregated inference,' pairing AWS Trainium 3 for prefill with Cerebras CS-3 for decode.
- Core gross margins reached 47% in Q1, but are projected to dip to 36-38% in Q2 due to temporary capacity rental costs.
Why It Matters
Cerebras is moving beyond the 'Nvidia-lite' narrative by establishing itself as a specialized infrastructure layer for latency-critical agentic AI. By securing a $20 billion commitment from OpenAI and a technical integration with AWS, the company has solved its historical customer-concentration risk while proving its wafer-scale architecture can scale in a production environment. For the streaming and interactive media sectors, this provides a concrete alternative to GPU-constrained clouds for real-time video processing and interactive AI agents. Success now depends on fulfilling massive deployment milestones in its new European data centers. Watch for Q2 gross margin recovery as a signal that the company is successfully transitioning from rented to owned data center capacity.
Additional Context
Following its record-breaking $6.4 billion IPO in May 2026, Cerebras has aggressively expanded its physical infrastructure to meet the 750-megawatt commitment from OpenAI. Per MarketBeat, in July 2026 the company announced its first major European expansion, planning to bring 200 megawatts of data center capacity online across France, Norway, and Finland by the end of 2027. This move specifically targets the 'sovereign AI' market and aims to reduce latency for European enterprise agents. This expansion is critical as the OpenAI agreement reportedly includes an option to increase the total deployment to 2 gigawatts by 2030, according to Zacks reporting in July 2026. Technically, Cerebras is differentiating through 'inference disaggregation,' a strategy validated by its March 2026 partnership with Amazon Web Services. As reported by Investing.com, the solution integrates Cerebras CS-3 systems directly into AWS data centers via Amazon Bedrock. This architecture recognizes a fundamental split in AI workloads: AWS Trainium 3 chips handle the compute-intensive 'prefill' stage (prompt ingestion), while the Cerebras Wafer-Scale Engine 3 manages the memory-intensive 'decode' stage (token generation). This hybrid approach claims to offer speeds 5x faster than existing generic GPU-based cloud solutions. Despite the growth, Cerebras faces significant margin pressure compared to competitors. While Nvidia maintains gross margins in the mid-70% range, Cerebras forecast full-year 2026 core gross margins between 38% and 41% due to the costs of rapid scaling. Per TradingView, CFO Bob Komin indicated that these margins are temporarily depressed as the company 'rents back' its own technology from clients to satisfy immediate demand during its global build-out. Analyst targets for the stock remain high, with a median target of $300 reported by Quiver Quantitative in late June 2026, reflecting optimism about its pole position in the high-speed inference market.
Read full article at tradingview.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source