OpenAI has developed its own custom inference chip, codenamed Jalapeño, which reportedly outperforms Nvidia's Rubin chip in performance-per-watt. This move signifies OpenAI's strategic effort to reduce inference costs and its reliance on Nvidia's ecosystem. Separately, Skild AI has introduced a robotics foundation model called S1, capable of learning tasks up to 10 minutes long from a single video demonstration without requiring fine-tuning. Apple has also launched its latest M6 chip, manufactured on a 2nm process, and a quad-die M5 Ultra chip, both featuring enhanced AI capabilities. AI
IMPACT This development signals a potential shift in AI hardware dependency and cost structures for major AI labs.
RANK_REASON OpenAI's announcement of its custom inference chip Jalapeño, detailing its performance benchmarks against Nvidia's offerings. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Alfred Lin
- Broadcom
- DeepSeek R1
- GPT-OSS
- Jalapeño
- Kimi K2.5 1T
- M5 Ultra
- M6
- Nvidia
- OpenAI
- Sequoia
- Skild AI
- TSMC
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →