GoKawiil
Tech News
OpenAI has fully unveiled Jalapeño, its first AI accelerator chip, boasting up to 13.4 petaflops of 4-bit compute and 232GB of high-speed memory. The company says benchmarks show up to 3.6x lower latency than Nvidia's GB300 at lower power, and that the chip went from concept to silicon in under 20 months using its own large language models to accelerate design work with a team of fewer than 100 engineers.
Dutch startup Euclyd raised a 200-million-euro ($231 million) Series A round, with Samsung, Somerset Capital Partners, EQT's Scaleup Europe Fund and Innovation Industries co-leading the investment. Founded in 2024, Euclyd is building an AI chip system with a different architecture from Nvidia's GPUs, aimed specifically at inference workloads, covering both processor and memory design.
OpenAI Korea's general manager Harrison Kim said the company has made significant progress working with Samsung Electronics on next-generation chip development and production. This suggests OpenAI could source future AI accelerators from both Samsung Foundry and TSMC, rather than relying solely on TSMC as it does for its current Jalapeño inference chip built with Broadcom.
Broadcom topped Wall Street expectations for its latest quarter, with adjusted earnings of $3.32 per share and revenue of $29.59 billion, both above estimates, while net income more than tripled year-over-year to $13.09 billion. However, the company's fourth-quarter revenue forecast of $34.8 billion came in slightly below the $35.03 billion analysts anticipated, sending shares down 5% in after-hours trading.
OpenAI revealed its first custom AI chip, called Jalapeño, claiming it delivers industry-leading speed and efficiency for inference workloads. The chip, co-developed with Broadcom, is set to be deployed in OpenAI's own infrastructure by year-end, with two more generations already in development.
At the Hot Chips conference, OpenAI released the first benchmark results for Jalapeño, its custom AI inference chip built with Broadcom. On SemiAnalysis' InferenceX test, the chip beat current Nvidia Blackwell systems in tokens served per user and throughput per kilowatt of power. OpenAI plans small-scale deployment by late 2026, with broader rollout in 2027.
OpenAI has revealed details of Jalapeño, a custom inference chip developed with Broadcom over roughly 16 months, at the Hot Chips conference. SemiAnalysis tested the chip using its InferenceX benchmark suite and found it outperforms Nvidia, AMD, and Google chips across multiple open-source models, despite being OpenAI's first hardware effort.
OpenAI detailed benchmark results for its Jalapeño chip, an inference-focused ASIC built with Broadcom, claiming it outperforms Nvidia's GB200 and GB300 chips on efficiency and response speed. Using its InferenceX benchmark across models like GPT-OSS 120B, DeepSeek R1, and Kimi K2.5, OpenAI reported 1.5-1.9x more work per watt and up to 3.6x lower latency. Small-scale deployment is planned by year-end, with volume scaling into 2027.