Friday Jun 26
RESEARCHOpenAI

OpenAI Builds Its Own Chip

26JUN
FIRST IN-HOUSE CHIPOWN CHIP

OpenAI now makes its own silicon. With Broadcom, it unveiled Jalapeño, a chip built only for running large language models. The lab is going after Nvidia from inside its own stack.

Jalapeño is OpenAI's first chip. It is an inference processor, tuned for serving models like ChatGPT and Codex.

The design took nine months, start to tape-out. OpenAI used its own models to help build it. Early tests show much better performance per watt than today's best. Deployment starts at gigawatt scale by late 2026, with Microsoft.

This is vertical integration. OpenAI now owns models, products, and the chips underneath. Less reliance on Nvidia.

full brief & sources

⚡ Why this matters

  • OpenAI joins Google and Amazon in building custom AI silicon.
  • Owning the chip lowers inference cost, the biggest line in serving AI.
  • A direct challenge to Nvidia's grip on AI compute.

🔍 What happened

  • Jun 24: OpenAI and Broadcom unveiled Jalapeño, OpenAI's first chip.
  • Built only for LLM inference, not a general-purpose accelerator.
  • Nine-month design-to-tape-out, claimed fastest ASIC cycle ever.
  • OpenAI used its own models to speed up the chip design.
  • Performance per watt 'substantially better' than current best, per early tests.
  • Deploys at gigawatt scale by end of 2026 with Broadcom, Celestica, and Microsoft.

💬 Smart takes

  • Greg Brockman, OpenAI President: 'The world is moving to a compute-powered economy.'
  • Hock Tan, Broadcom CEO: calls it 'a multi-generation roadmap' for gigawatt data centers.
  • Skeptic: Engineering samples are not volume production. Nvidia still owns the software moat and the supply.

🧭 Where this goes

  1. LikelyOpenAI keeps buying Nvidia for training while moving inference to its own chips.
  2. Likelymore labs announce custom inference silicon within a year.
  3. Possibleinference cost per token drops enough to reset API pricing.
  4. Wild CardOpenAI sells Jalapeño capacity to other developers and becomes a chip vendor.

🥄 The Spoon Take

The model race is becoming a stack race. Whoever controls the chip controls the cost of every answer. OpenAI is copying the Google playbook: own the silicon, own the margins. Nvidia still wins on training. But inference is where the money leaks, and OpenAI just plugged the hole.

🤔 Pushback

Designing a chip and running it at gigawatt scale are very different. Yields, software, and supply could slip the timeline by years.

·OpenAI - Jalapeño chip·CNBC - First custom chip·TechCrunch - Built by Broadcom