Baron Opportunity Fund on Cerebras Systems Inc.
Thesis
Cerebras has developed a differentiated wafer-scale architecture optimized for fast AI inference, an increasingly important workload as applications and agents chain together multiple models. The manager believes demand for tokens per second is expanding rapidly and Cerebras has a genuine performance advantage in this niche. A multi-year OpenAI agreement for 750 megawatts of inference capacity valued above $20 billion provides significant commercial validation, while an AWS partnership broadens distribution. The company can monetize its technology through both direct systems sales and cloud services. If fast inference becomes a larger portion of AI compute, Cerebras is positioned to benefit as a category leader.
“On balance, we find the setup compelling and initiated a position during the quarter.”
Key risks
- Execution of the OpenAI deployments
- Workloads may evolve toward architectures requiring significantly more memory
- Customers may become unwilling to pay a premium for faster inference