Watch CNBC's full interview with Cerebras Systems CEO Andrew Feldman
Discussed Q2 results, the growth strategy and AI chip demand, saying the company was "pretty proud of the results."
Co-Founder, CEO, President & Chairman, Cerebras Systems Inc. (CBRS)
10 recorded appearances
Discussed Q2 results, the growth strategy and AI chip demand, saying the company was "pretty proud of the results."
Argued data-center capacity, not demand, is the binding constraint on Cerebras' growth, with power availability the primary bottleneck.
Long-form explanation of the wafer-scale bet: why a dinner-plate-sized chip wins on inference latency, and why frontier labs including OpenAI route workloads to it.
Feldman's segment covers the scale of the data-center buildout, why inference and reasoning workloads break Moore's-Law economics, and his view on open source, AI sovereignty and the road to AGI.
First earnings interview since the May IPO; walked through the gross-margin outlook and the shape of demand.
Rejected the infrastructure-bubble framing - build-out is tracking booked demand, and power, not capital, is the bottleneck. Also argued most "AI-driven" layoffs are AI-washed cost cuts.
Explained the engineering and economic logic of wafer-scale integration versus GPU clusters, and why GPUs struggle to deliver fast inference.
Walked Sarah Guo and Elad Gil through the road to the IPO - the years of doubt about wafer-scale, and how inference speed became the wedge against Nvidia.
IPO-day interview: defended going public as the right funding structure for the build-out, and discussed the OpenAI and AWS partnerships.
Explained the decision to raise $1bn privately and push the IPO out, argued Nvidia's behavior showed concern about growth, and questioned whether US energy supply can keep up with AI demand.