OpenAI on Cerebras
OpenAI has unveiled its GPT-5.3-Codex-Spark model, which achieves high processing speeds by running exclusively on Cerebras Systems' CS3 accelerators, marking the first deployment of an OpenAI model on rival hardware.
Custom AI Silicon Acceleration
named product
emerging tech
supply constraint
The pursuit of compute alternatives, evidenced by OpenAI exploring Cerebras and others shunning HBM, highlights supply chain pressure. This involves designing specialized microprocessors tailored precisely for specific AI workloads, moving beyond general-purpose GPUs to optimize power, cooling, and density within hyperscale and edge infrastructure.
bytedance samsung chips
ByteDance is reportedly in negotiations with Samsung for the mass production of 350,000 specialized artificial intelligence inferencing chips, with samples expected by the end of March.
Positron AI Chips
Artificial intelligence inferencing startup Positron AI secured $230 million in funding to develop its Atlas chips, which reportedly offer three times the compute efficiency per watt compared to Nvidia's H100 graphics processing units.
cerebras systems funding
Cerebras Systems secured one billion dollars in new funding, achieving a valuation of twenty-three billion dollars shortly after announcing a significant ten-billion-dollar agreement with OpenAI.
positron vs nvidia
The Arm-backed startup Positron claims its next-generation Asimov accelerators, utilizing lower-cost LPDDR5x memory instead of high-bandwidth memory, can effectively compete against offerings like Nvidia's Rubin GPUs.