Baron Asset Fund on Cerebras Systems
Thesis
Baron initiated Cerebras following its IPO because it believes the company has developed one of the most differentiated architectures for fast AI inference. Demand for tokens per second is rising rapidly as AI applications and agents combine increasingly complex combinations of underlying models, creating greater need for low-latency inference infrastructure. Cerebras' specialized semiconductor architecture is viewed as having a genuine performance advantage for these workloads, with validation from a $20 billion OpenAI data-centre contract and engagements with hyperscalers including Amazon Web Services. Baron expects AI infrastructure investment and fast-inference demand to remain strong and believes Cerebras can continue innovating to maintain its lead.
“we believe that Cerebras holds a genuine performance advantage for a specific set of workloads that require fast inference.”
Key risks
- Cerebras is currently pursuing both cloud services and semiconductor systems business models
- The thesis depends on fast inference becoming a sizable portion of the broader AI inference market