OpenAI has unveiled Jalapeño, a custom AI chip that surpasses Nvidia's Blackwell in performance benchmarks. The processor marks OpenAI's push into hardware development to reduce reliance on external chipmakers.
OpenAI's Jalapeño chip demonstrates superior performance across key AI workloads compared to Nvidia's latest Blackwell architecture. The processor achieves higher throughput on transformer model inference and training tasks while consuming less power than its competitor.
The chip represents OpenAI's vertical integration strategy, addressing supply constraints and cost pressures in the AI infrastructure market. Development focused on optimizing for large language model operations, where Jalapeño shows measurable advantages in latency and compute efficiency.
Blackwell, released earlier this year, remains widely deployed across cloud providers. However, Jalapeño's emergence signals competitive pressure in the specialized AI chip space, traditionally dominated by Nvidia.
OpenAI plans limited Jalapeño production initially, prioritizing internal model training and inference. The company has not disclosed pricing or external availability timelines. Industry analysts view the move as OpenAI's bid for operational independence and margin improvement in AI compute costs.
McGraw Hill CEO Philip Moyer highlighted the company's expanding AI footprint in education, revealing 7.5 million AI tools currently in use alongside 100 million curriculum licenses globally.
OpenAI has expanded access to ChatGPT's task scheduling tool, making it available to free account holders. Paid subscribers gain additional capabilities with Gmail, Slack, and GitHub integrations.
A massive buildout of AI infrastructure is triggering bipartisan political opposition in the US, potentially creating market risk for the AI trade ahead of midterm elections, according to Barclays analysis.
Skild AI has released S1, a robotics foundation model capable of learning novel tasks from a single video demonstration without requiring fine-tuning. The model can execute commands with 10-minute planning horizons.