Inference

OpenAI says GPT-5.6 improves intelligence-per-dollar across models, inference, and agents, signaling cost efficiency is now central to frontier AI.

OpenAI and Broadcom debut Jalapeño, OpenAI's first custom AI accelerator chip, purpose-built for inference workloads and developed using OpenAI's own models.

Anthropic is reportedly in early talks to buy Fractile inference chips as AI memory costs and supply pressure intensify.

Google is negotiating with Marvell Technology to co-develop two new AI chips—a memory processing unit and a next-gen TPU—to boost AI inference efficiency.

Nvidia is set to announce a new inference chip platform at its upcoming GTC conference in March, aiming to further solidify its dominance in the AI hardware market.

Inferact, the startup commercializing the popular open-source inference engine vLLM, has raised a $150 million seed round led by Andreessen Horowitz and Lightspeed to accelerate AI application deployment.