Key Takeaways
- Shares of Cerebras plummeted 7-9% following claims that OpenAI is utilizing Nvidia GPUs rather than Cerebras chips to power its “Ultrafast” AI service.
- According to SemiAnalysis, OpenAI’s GPT-6.1 Sol Ultrafast operates on Nvidia graphics processors at low batch configurations, not on Cerebras’ Wafer-Scale Engine technology.
- OpenAI represents Cerebras’ biggest customer in terms of revenue commitments, and both companies had previously announced in August that Cerebras would supply the infrastructure for Ultrafast mode.
- No official confirmation or denial has been issued by either Cerebras or OpenAI regarding these allegations.
- Cerebras shares have experienced significant volatility since debuting at $350 during its May IPO.
Shares of Cerebras Systems experienced a notable decline this week following a report casting doubt on the company’s continued involvement in supporting OpenAI’s highest-performance AI offering. The selloff resulted in an intraday decline reaching 8.87% during Wednesday trading.
The catalyst emerged from semiconductor analysis firm SemiAnalysis, which published claims on social media asserting that OpenAI’s latest GPT-6.1 Sol Ultrafast model operates on conventional Nvidia GPUs rather than Cerebras’ proprietary chip architecture.
This assertion carries significant weight. OpenAI rolled out its “Ultrafast” service tier with promises of performance up to eight times faster than standard offerings, delivering approximately 300 tokens per second. Cerebras had been publicly identified as the technology provider enabling those performance gains.
Cerebras has structured its entire value proposition around addressing precisely these computational challenges. The company’s Wafer-Scale Engine represents a massive semiconductor design that integrates billions of cores and extensive memory capacity onto a single silicon wafer. This architecture eliminates the latency issues associated with distributing workloads across hundreds of discrete GPU units.
Breaking Down the SemiAnalysis Claims
The critical element centers on a specific technical phrase in the analysis: “low batch size.” In AI inference operations, batching refers to the practice of consolidating multiple requests to maximize GPU utilization efficiency.
Nvidia’s chips typically require substantial batch sizes to achieve optimal throughput. A low batch configuration processes fewer concurrent requests, prioritizing response speed over computational efficiency. This represents the precise market segment Cerebras has positioned itself to dominate.
Should OpenAI have developed methods to deploy premium models on Nvidia infrastructure at low batch sizes while maintaining cost-effectiveness, it would undermine a core competitive advantage Cerebras has promoted. This explains the swift market response.
Cerebras stock settled near $180.22 on Wednesday as the information circulated. This represents a substantial decline from the company’s $350 IPO valuation established in May.
The initial public offering generated considerable market enthusiasm. Investors showed strong interest in a semiconductor company claiming superior inference performance compared to Nvidia for specific computational workloads.
That optimism has diminished considerably since launch. Post-IPO financial disclosures have failed to sustain the investor confidence evident during the company’s debut.
The Strategic Importance of the OpenAI Relationship
OpenAI represents far more than a typical customer for Cerebras. The AI research company constitutes Cerebras’ most significant client measured by contracted revenue, according to Barron’s.
Any deterioration in the OpenAI partnership, even if temporary, presents legitimate concerns for shareholders. Cerebras has relied substantially on its OpenAI connection to support its market valuation.
Neither organization has issued public commentary addressing the SemiAnalysis assertions. This information vacuum has forced market analysts to speculate rather than work from verified information.
There remains possibility this situation may not be permanent. Supply chain limitations or ongoing technical refinements could explain why Nvidia hardware currently handles these workloads.
Cerebras might ultimately power Ultrafast mode once any implementation challenges are resolved. Nothing in either company’s public stance definitively excludes this outcome.
The immediate stock movement reflects investor uncertainty rather than confirmation of negative developments. Market participants are factoring in risks that Nvidia’s established CUDA software ecosystem and continuous hardware enhancements make exclusive partnerships with leading AI organizations increasingly difficult for competitors to secure.
Cerebras has not released any official communication confirming or refuting its current position within OpenAI’s infrastructure roadmap. As of Wednesday’s close, the stock remained down approximately 8% for the session.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
