OpenAI has officially entered the hardware race with the announcement of 'Jalapeño,' its first custom-designed AI inference chip developed in partnership with Broadcom. The chip is designed to drastically reduce AI serving costs and improve latency for ChatGPT and its API ecosystem. By building its own silicon, OpenAI moves closer to becoming a vertically integrated 'AI factory,' reducing its multi-billion dollar dependency on NVIDIA's H-series and B-series GPUs.
A massive 'brain drain' from legacy tech giants to venture-backed labs has reached a fever pitch. This week, Google lost key Gemini researchers Jonas Adler and Alexander Pritzel to Anthropic, while veteran researcher Noam Shazeer and Nobel Prize-winner John Jumper joined the rival labs. Analysts suggest the allure of massive equity ahead of anticipated IPOs is fueling a talent war that is hollowing out established research divisions.
By mid-2026, the frequency of major AI model releases has quadrupled compared to 2023, with the industry now maintaining a cadence of nearly 70 significant frontier model updates per year.
Which tech giant is currently the 'lone holdout' in refusing to voluntarily submit its AI models for federal safety reviews by the Center for AI Standards and Innovation?