O Que É MBS

According to Beating AI News, OpenAI has released the initial test data for its first custom inference chip, Jalapeño, claiming it outperforms NVIDIA's GB300. The chip is at the Pareto optimal frontier for three publicly available external models: GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T. Compared to the best existing commercial systems, it achieves a peak AI output per watt that is 1.5 to 1.9 times higher, with end-to-end latency reduced by 1.7 to 3.6 times; in high-interaction agent scenarios, the advantage expands to 2.1 to 4.1 times. Jalapeño has a rated power of 700 watts, with measured continuous power not exceeding 550 watts. OpenAI emphasizes that the true cost under agent load should be measured by 'the amount of AI work done per unit of power consumption,' rather than single-chip performance. Jalapeño took only 9 months from design to tape-out, with significant AI involvement in circuit optimization and validation. Utilizing Codex and GPT-Astra, the team optimized three open-weight models not originally planned to high performance in just two months, with some AI-generated modules achieving speeds 1.5 to 1.8 times faster than human-written implementations. OpenAI plans to deploy Jalapeño to its own computing infrastructure by the end of the year while continuing to utilize external accelerators like NVIDIA on a large scale. This is the first generation of a multi-generation chip roadmap, with Gen 2 already in deep development and Gen 3 taking shape.

According to Beating AI news, OpenAI has released the initial test data for its first custom inference chip, Jalapeño, claiming it outperforms Nvidia's GB300. The chip is at the Pareto optimal frontier for three publicly available models: GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T. Compared to the best existing commercial systems, it achieves a peak AI output per watt that is 1.5 to 1.9 times higher, with end-to-end latency reduced by 1.7 to 3.6 times; in high-interaction agent scenarios, the advantage expands to 2.1 to 4.1 times. Jalapeño has a rated power of 700 watts, with actual continuous power not exceeding 550 watts. OpenAI emphasizes that the true cost under agent load should be measured by the 'AI workload completed per unit power consumption,' rather than single-chip performance. The design to tape-out of Jalapeño took only nine months, with deep AI involvement in circuit optimization and verification. Utilizing Codex and GPT-Astra, the team optimized three open-weight models not included in the original plan to high performance in just two months, with some AI-generated modules achieving speeds 1.5 to 1.8 times faster than human-written implementations. OpenAI plans to deploy Jalapeño to its own computing infrastructure by the end of the year while continuing to use external accelerators like Nvidia on a large scale. This is the first generation of a multi-generation chip roadmap, with Gen 2 already in deep development and Gen 3 taking shape.

On August 26, according to documents disclosed by the U.S. Securities and Exchange Commission (SEC), Micron Technology (MU.O) President and CEO Sanjay Mehrotra sold 40,000 shares of common stock at an average price of $968.9 per share on August 21, valued at approximately $38.7562 million. According to BIT (Bit.com) market data, Micron's stock rose 2.48% this morning, closing at $932.97.