OpenAI’s Jalapeño AI chip brings new 'threat' to Nvidia margins as custom silicon gains ground
- ID
- 18136
- Status
- summarized
- Published
- 27 Aug 2026, 3:15 AM
- Fetched
- 27 Aug 2026, 3:55 AM
- Provider
- CNBC Technology
- Category
- technology
- Original URL
- https://www.cnbc.com/2026/08/26/openai-jalapeno-ai-chip-nvidia.html
- Source URL
- https://www.cnbc.com/id/19854910/device/rss/rss.html
Summary
- Score
- 5.5
- Created
- 27 Aug 2026, 3:55 AM
- Tags
- Audience
- ai_ml_learnerssaas_foundersdevelopers
What happened
OpenAI announced its first custom AI chip, Jalapeño, designed for inference and expected to deploy in its infrastructure by end of 2026. OpenAI claims it beats Nvidia's Blackwell systems on performance per watt in nearly all tested scenarios, though analysts note Nvidia's newer platform is a more fair comparison. Google, AWS, and Meta are also building custom silicon, signaling a broader shift away from sole reliance on Nvidia GPUs for inference workloads.
Why it matters
If hyperscaler custom silicon meaningfully undercuts Nvidia on inference cost-per-watt, API pricing for inference-heavy workloads could drop over the next 12-18 months. Builders currently locking into expensive GPU-backed inference contracts or pricing models should factor in potential price compression before committing to long-term deals.
Discussion angle
Whether custom silicon from OpenAI and hyperscalers will actually translate into cheaper inference APIs for builders, or whether the savings stay internal to the chip owners.