Releases·Hardware Release
Huawei plans Q1 2027 launch of new AI chip as it takes on Nvidia
Huawei announced at its Huawei Connect conference that its next-generation Ascend 960DT AI chip will now launch in Q1 2027, six months ahead of the previous Q3 2027 plan, as it races to challenge Nvidia. The move comes despite U.S. semiconductor restrictions and alongside Huawei's new Peerium Computing Architecture for building massive AI systems.
0 engagement1 sourceThu, Sep 17, 2026, 02:06 PM
AI Overview
- Huawei moved the Ascend 960DT launch up to Q1 2027 from Q3 2027, announced by acting chairman David Wang at Huawei Connect
- Huawei's new Peerium Computing Architecture, linked via UnifiedBus, debuts in the Atlas 950 SuperPoD and SuperCluster, which connects up to 256,000 accelerator cards
- Analyst Rui Ma noted the announced Atlas 960 SuperPoD has 4,096 chips versus the previously stated 15,488 Ascend 960 chips
- Rotating chairman Eric Xu said China must accelerate AI development to catch up with the U.S., per Financial Times
- The announcement comes a week before the Trump-Xi meeting set for September 24 in Washington, DC
Related
Releases·Hardware Release
China's Huawei says AI chip demand outstrips supply as it steps up Nvidia challenge
Reuters reports Huawei says demand for its AI chips outstrips its supply capacity as it challenges Nvidia. Tom's Hardware reports Huawei is shelving a global rollout of its AI chips, focusing output on Chinese domestic demand.
Reuters reports Huawei says demand for its AI chips outstrips its supply capacity as it challenges Nvidia. Tom's Hardware reports Huawei is shelving a global rollout of its AI chips, focusing output on Chinese domestic demand.
Updated Mon, Sep 21, 2026, 07:09 PM — Tom's Hardware reports Huawei has shelved the global rollout, prioritizing China's domestic demand.
581 engagement2 sourcesredditThu, Sep 17, 2026, 11:53 AM
News·Company News
AMD announces Helios rack-scale system and new chips to compete with Nvidia
At its Advancing AI conference in San Francisco, AMD CEO Lisa Su unveiled the Helios rack-scale system and new chips aimed at powering large AI labs, with Microsoft as a customer. The system is set to ship later this year, directly challenging Nvidia's dominance in AI data center hardware.
At its Advancing AI conference in San Francisco, AMD CEO Lisa Su unveiled the Helios rack-scale system and new chips aimed at powering large AI labs, with Microsoft as a customer. The system is set to ship later this year, directly challenging Nvidia's dominance in AI data center hardware.
Updated Thu, Jul 23, 2026, 08:33 PM — AMD announces Helios rack-scale system and new chips, with Microsoft as customer, shipping later this year.
0 engagement2 sourcesrssThu, Jul 23, 2026, 12:00 AM
Releases·Hardware Releasefeatured
OpenAI releases Jalapeño inference chip benchmarks beating Nvidia GB200/GB300
OpenAI published first benchmark results for Jalapeño, its custom inference chip built with Broadcom, claiming 1.5–1.9x higher throughput per kilowatt and 1.7–3.6x lower end-to-end latency than Nvidia's GB200/GB300 rack systems. Testing ran through SemiAnalysis's public InferenceX benchmark on GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T. Deployment in OpenAI's compute infrastructure is planned by year-end 2026, with Gen 2 and Gen 3 chips in development.
OpenAI published first benchmark results for Jalapeño, its custom inference chip built with Broadcom, claiming 1.5–1.9x higher throughput per kilowatt and 1.7–3.6x lower end-to-end latency than Nvidia's GB200/GB300 rack systems. Testing ran through SemiAnalysis's public InferenceX benchmark on GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2.5 1T. Deployment in OpenAI's compute infrastructure is planned by year-end 2026, with Gen 2 and Gen 3 chips in development.
Updated Fri, Aug 28, 2026, 05:07 PM — Jalapeño presented at Hot Chips (37th) alongside Cerebras CS-5, Groq 3 LPX, and Apple M6; OpenAI says it is not an ASIC.
014121.7k engagement16 sourcesreddit, hackernews, rss +2Tue, Aug 25, 2026, 02:00 PM
News·Company News
Moonshot-NVIDIA talks signal Chinese AI models moving into enterprise adoption
A techstrong.ai report says large businesses, not just individuals and small companies, are adopting Chinese AI models. Talks between Moonshot AI and NVIDIA are cited as evidence of this enterprise shift, driven largely by cost considerations.
A techstrong.ai report says large businesses, not just individuals and small companies, are adopting Chinese AI models. Talks between Moonshot AI and NVIDIA are cited as evidence of this enterprise shift, driven largely by cost considerations.
28 engagement1 sourceredditFri, Aug 28, 2026, 07:32 PM
News·Company News
Wistron opens Fort Worth plant to produce NVIDIA AI systems
Wistron opened its first U.S. manufacturing facility in Fort Worth, Texas, a 324,000-square-foot greenfield plant that will produce NVIDIA superchips for AI systems. NVIDIA CEO Jensen Huang joined Wistron Chairman Simon Lin at the opening ceremony, highlighting the plant's role in building AI infrastructure in the U.S.
Wistron opened its first U.S. manufacturing facility in Fort Worth, Texas, a 324,000-square-foot greenfield plant that will produce NVIDIA superchips for AI systems. NVIDIA CEO Jensen Huang joined Wistron Chairman Simon Lin at the opening ceremony, highlighting the plant's role in building AI infrastructure in the U.S.
5 engagement2 sourcesrss, redditTue, Jul 21, 2026, 10:35 PM