OpenAI Bets Billions on Cerebras as Nvidia Grip Loosens
On Christmas Eve 2025, OpenAI committed billions of dollars to purchase chips from Cerebras, a startup that manufactures large-format AI inference chips, according to securities filings. The deal marks a significant shift in the AI chip market, coming the same day Nvidia announced its acquisition of inference chip rival Groq. The move signals OpenAI's effort to diversify its chip supply chain beyond Nvidia and reduce dependence on a single supplier for inference workloads.
TL;DR
- OpenAI signed a multi-billion dollar chip purchase agreement with Cerebras on December 24, 2025, revealed through securities filings
- Cerebras manufactures large dinner-plate-sized chips optimized for running AI models rapidly, positioning it as an alternative to Nvidia for inference
- The deal came the same day Nvidia announced its acquisition of Groq, another inference chip startup, intensifying competition in the AI chip market
- The agreement represents both an opportunity and a risk for Cerebras, as it gains a major customer but becomes dependent on OpenAI's continued demand
Why It Matters
The Cerebras deal reflects a critical inflection point in AI infrastructure: major AI labs are now actively diversifying away from Nvidia's dominance in chip supply. This shift accelerates the emergence of specialized inference chips as a distinct market segment, separate from training chips, and validates the business case for alternative chip architectures. For the broader AI ecosystem, it signals that no single supplier will control the entire pipeline, which could reshape competitive dynamics and pricing power in AI infrastructure.
Business Impact
For operators and founders, this deal demonstrates that large-scale AI deployments increasingly require multiple chip suppliers to manage risk, cost, and performance tradeoffs. Companies building AI infrastructure or considering chip partnerships should recognize that inference workloads are becoming a distinct purchasing category with viable alternatives to Nvidia. Founders in the chip space now have a clearer path to enterprise adoption, though success requires deep optimization for specific inference use cases.
Key Implications
- Cerebras gains a marquee customer and validation for its inference-focused chip design, but becomes heavily dependent on OpenAI's capital allocation and product roadmap decisions
- The inference chip market is consolidating around a few players, with Groq and Cerebras emerging as primary alternatives to Nvidia, raising questions about long-term viability for smaller competitors
- OpenAI's diversification strategy suggests concerns about Nvidia's supply constraints, pricing, or willingness to prioritize inference workloads, creating leverage for alternative suppliers
- The timing of both deals on the same day indicates coordinated industry movement toward multi-vendor strategies, potentially accelerating similar arrangements across other AI labs
What to Watch
Monitor whether other major AI labs and cloud providers follow OpenAI's lead in committing to Cerebras or other non-Nvidia chip suppliers. Track Cerebras' ability to scale manufacturing and meet OpenAI's delivery timelines, as execution risk remains high for hardware startups. Watch for announcements about pricing, performance benchmarks, and total contract value, which will signal whether Cerebras can achieve competitive unit economics against Nvidia's entrenched position.
Related Video
Subscribe to the newsletter
The latest stories and analysis, delivered to your inbox.
Free. No spam. Unsubscribe any time.
