The conference with hot chips and even hotter companies
128 chips, 1.7 exaFLOPS, and 27 TB of HBM give Altman and crew a leg up over Blackwell, and maybe even Rubin
OpenAI still isn’t giving up Nvidia chips, though.
OpenAI's first AI accelerator fails to beat Nvidia's Blackwell in terms of raw performance, but it can offer very good performance-per-watt and low latency, which is exactly what the doctor ordered for inference…
OpenAI showed off "Jalapeño," its first in-house inference chip, with benchmarks at the Hot Chips conference. According to SemiAnalysis tests, the chip beats Nvidia's Blackwell and even Rubin in throughput and energy…
AI is speeding up chip design. That was a key theme in conversations I had this week with engineers, researchers and a handful of startups pitching AI-driven design help at the annual Hot Chips semiconductor conference…
OpenAI introduces its Jalapeño chip, promising lower latency and higher throughput, enabling faster and more efficient AI responses compared to competing systems.
The chip is designed for serving up AI models, rather than training new ones.
Tested on Semianalysis’s InferenceX benchmark, Jalapeño registered both more tokens per user and more throughput per kilowatt than the currently available state-of-the art.
OpenAI's custom Jalapeño inference chip outperformed Nvidia's GB300 processors by up to 3.6x in latency tests, with a planned rollout beginning
OpenAI showed off "Jalapeño," its first in-house inference chip, with benchmarks at the Hot Chips conference. According to SemiAnalysis tests, the chip beats Nvidia's Blackwell…
Nvidia's Vera Rubin wasn't in the comparison.
OpenAI just published benchmark results for Jalapeño — its first custom AI inference chip — and the...
The custom inference chip, built with Broadcom, led on tokens per user and throughput per kilowatt in third-party tests
OpenAI’s Jalapeño chip beat Nvidia Blackwell systems on key inference-efficiency tests as custom AI silicon gains ground among major tech companies.
OpenAI's Jalapeño chip tops Nvidia's GB200 and GB300 in AI work per watt, not per chip. The real story: with power scarce, efficiency is what buys AI capacity.
OpenAI's first AI accelerator fails to beat Nvidia's Blackwell in terms of raw performance, but it can offer very good performance-per-watt and low latency, which is exactly what…
AI is speeding up chip design. That was a key theme in conversations I had this week with engineers, researchers and a handful of startups pitching AI-driven design help at the…