Alibaba reveals Zhenwu M890 AI chip but admits struggling with production scale

Alibaba just admitted it's struggling to keep up with rival chipmakers and AI shops

Chinese tech giant Alibaba has revealed a new accelerator and accompanying rack-scale server rig without offering much detail about their performance - and also admitted it's only been able to make chips in trivial quantities. The new chip is called the Zhenwu M890, and comes from Alibaba's semiconductor design business T-Head. Neither company has said much about it other than stating it includes 144GB of on-chip memory, possesses "800 GB per second of inter-chip bandwidth" and natively supports precision formats from FP32 down to FP4. The Chinese giant didn't offer any info about performance other than to say it delivers "three times the performance of its predecessor, Zhenwu 810E." Based on the specs of the old and new devices, we think the M890 might give Nvidia's 2024-vintage H200 a run for its money. That means the most interesting figure in Alibaba's announcement is 560,000 - the number of Zhenwu chips Alibaba says T-Head has made to date. By way of contrast, Nvidia says AWS alone will rack and stack one million of its GPUs this year. AWS's spending on AI infrastructure is at similar levels to Microsoft, Meta, and Google, so it's conceivable that Nvidia will make and sell three or four million GPUs to satisfy those four customers alone. Alibaba's announcement doesn't offer any information about production volumes for the M890. The company did talk up the machines the M890 will run inside - a new beast called the Panjiu AL128 Supernode Server Alibaba described as "a rack-scale system that packs 128 AI accelerators into a single unit and delivers petabyte-per second internal bandwidth ... designed specifically for the concurrency patterns that agents generate: unpredictable, high-frequency bursts of inference requests that overwhelm conventional compute clusters." It seems Alibaba intends racks packed with M890s plus Panjiu AL128s to handle agentic workloads. T-Head has also created a new networking chip called the "ICN Switch 1.0," which we're told "delivers up to 25.6 Tbps of aggregate bandwidth and enables congestion-free communication across clusters of 64 accelerators." Those are specs that Broadcom and Nvidia reached years ago. Alibaba's chips therefore deliver performance that leaves its cloud well behind its western competitors, which would be a problem if its Chinese peers were buying kit from Nvidia. But despite the US lifting export restrictions on some advanced AI hardware, Beijing has not let local buyers acquire any: Nvidia recently told investors it doesn't expect to win any revenue from China for the foreseeable future. Chinese hyperscalers haven't announced capex spending at anywhere near the levels of their American rivals. Perhaps we now know why: T-Head can't get a lot of gear made, probably because accelerators like the M890 require an advanced semiconductor manufacturing process that Chinese companies can't access from offshore fabs due to US sanctions. Chinese fabs can't yet match the prowess of TSMC, the source of most high-end GPUs. So how are China's tech giants meeting demand? We've previously covered attempts to smuggle Nvidia parts into China and Bloomberg yesterday reported Taiwanese authorities have cracked down on GPU smugglers. We've also seen suggestions that Chinese companies send storage devices across borders to move data into facilities that have Nvidia kit waiting to run training workloads. ®

Alibaba reveals more powerful Zhenwu AI chip, new LLM

CHONGQING, China -- Alibaba announced Wednesday its new artificial intelligence chip would be three times as powerful as its predecessor, as rival Nvidia struggles to get its advanced chips into China. The Zhenwu M890 delivers three times the performance of the current Zhenwu 810E, Alibaba said, adding that the new processor has 144 GB GPU memory and interchip bandwidth of 800 GB per second. Alibaba said it had already delivered 560,000 Zhenwu units to more than 400 customers across 20 industries. The e-commerce giant also revealed its next generation large language model, Qwen3.7-Max, would soon be released.

Alibaba unveils the Zhenwu M890 as China's NVIDIA alternative push hardens

T-Head's new GPU lands inside US export controls, a Trump-Xi summit on AI chips, and a Chinese domestic accelerator market, the company says, is already in scaled mass production. Alibaba's T-Head chip unit unveiled detailed specifications of the Zhenwu M890 on Wednesday, the company's latest GPU-class AI chip designed as a domestic alternative to NVIDIA's accelerators. An Alibaba executive said on the same day that T-Head's proprietary GPU chips have achieved scaled mass production. The announcement lands inside an unusually busy fortnight for the Chinese-AI-chip narrative. What the chip is, on the available specifications, is the South China Morning Post's detailed account. The Zhenwu M890 is the highest-spec product T-Head has shipped to date and is positioned against NVIDIA's H100 generation rather than the newer Blackwell. The performance gap to NVIDIA's flagship is, on independent benchmark commentary, still meaningful, but the gap to the H100, which is the part of the NVIDIA line-up Chinese customers can no longer legally buy under US export controls, is closer. The combination of inability-to-buy-H100 plus a credible domestic alternative is the part of Alibaba's announcement is calibrated against. The 'scaled mass production' claim is the unusual operational detail. EE Times' coverage describes the production-line ramp as the kind of disclosure Western analysts have been requesting from Chinese chip-design houses for two years; the willingness of an Alibaba executive to make the claim on the record signals that the company is confident enough in its supply-chain redundancy to invite the technical scrutiny that follows. The chip is, on the available reporting, manufactured at process nodes that Chinese foundries can produce without US-controlled lithography equipment, which is the binding constraint that has defined the entire Chinese domestic-chip cycle. The corporate-finance framing matters. T-Head, the Alibaba chip arm formally known as Pingtouge (Chinese for 'honey badger'), was established in 2018, shipped its first AI chip (the Hanguang 800) in 2019, and has been operating as an internal-supply unit inside Alibaba Cloud ever since. T-Head is planning an IPO to bankroll a more aggressive infrastructure-investment programme, putting the unit on a direct collision course with Cambricon and Huawei's Ascend lineage for the domestic accelerator market. The Zhenwu M890 announcement, on that read, is also the operational substance behind whatever T-Head's prospectus eventually says about competitive position. The competitive pattern is the part the wire coverage frequently underplays. Chinese hyperscalers, foundation-model labs and AI-deployment customers have been ramping homegrown AI chip purchases through Q1 and Q2 2026, even as NVIDIA's H200 has been cleared for sale to a small list of Chinese customers under the new export-licensing regime. The two trends are not contradictory; Chinese customers want optionality on both sides. What Alibaba is doing with the Zhenwu M890 is positioning T-Head as the domestic-side default, the way Huawei has positioned Ascend and Cambricon has positioned the Siyuan line. The market is, on the available evidence, big enough to support all three at scale. The geopolitical context arrived earlier this month with the Trump-Xi Beijing summit. The US-China AI-policy track is now being negotiated at the head-of-state level, with H200 export licensing and AI guardrails on the same agenda. None of the H200 chips cleared for ten Chinese buyers has yet shipped. Chinese customers are, in that procurement vacuum, accelerating the domestic-alternative path. The Zhenwu M890 announcement is a calibrated commercial response to that vacuum. The wider NVIDIA-alternative arc the announcement sits inside extends well beyond China. Tenstorrent's takeover conversations with Intel and Qualcomm on the US side, and Google's $25bn TPU-cloud joint venture with Blackstone announced this week, are both versions of the same trade. The AI-compute supply curve cannot, on current capacity, support hyperscaler demand on NVIDIA silicon alone. Alibaba's positioning of T-Head is the Chinese-version answer to that allocation problem. On the operational details the wire coverage leaves to subsequent reporting: the per-chip pricing, total Zhenwu M890 shipment volumes to date, named customers beyond Alibaba's own cloud business, the breakdown of internal-versus-third-party shipment volumes, and the T-Head IPO timeline. None of those details have been formally disclosed. What is now visible is the technical disclosure plus the production-status claim. The next named-customer announcement, particularly outside Alibaba Cloud's own infrastructure, will be the visible proof point of whether T-Head can take the chip beyond captive use.

Alibaba unveils new AI chip as Nvidia access remains stalled

Beijing (AFP) - Tech giant Alibaba released on Wednesday a new artificial intelligence chip it said performed three times as well as its predecessor, showcasing growing domestic chipmaker prowess as US titan Nvidia struggles for access to China. Semiconductors have been at the centre of a fierce US-China race for AI supremacy, with Nvidia's most advanced chips banned from sale in China by Washington over national security concerns. Beijing has in response sought to bolster its self-reliance, pouring resources into promoting its domestic industry and reportedly barring firms from buying Nvidia chips. Alibaba said its new chip, Zhenwu M890, can deliver three times the performance of its predecessor Zhenwu 810E, which is widely believed to match the capabilities of Nvidia's H20. The H20 is a less powerful version of Nvidia's AI processing units designed specifically for export to China. A more high-end option, the H200, has been licensed to sell to China from the US side, but its access to the Chinese market appears to have stalled. Zhang Guobin, founder of Chinese specialist website eetrend.com, said the timing of the Alibaba launch is "extremely precise". "It's during a window when the prospects of the H200 entering the Chinese market are highly uncertain and Nvidia's business in China has effectively dropped to zero," he told AFP. Nvidia boss Jensen Huang was part of a US business delegation that travelled to Beijing with President Donald Trump last week. He said he had not discussed the H200 directly but added "the Chinese government has to decide how much of their local market... they want to protect". The new Zhenwu M890 chip "provides a reliable option that is insulated from fluctuations in export controls, enabling domestic AI companies to formulate long-term technology roadmaps," said eetrend.com's Zhang. "At the very least... (it) proves that in the field of high-end AI computing power, China now has a 'Plan B' that does not depend on Nvidia," he added. However, when it comes to performance, Nvidia still holds its lead. Even the H200 is significantly less advanced than the firm's top-range chips -- the Blackwell series and forthcoming Rubin processors. Alibaba said to date it has shipped more than 560,000 chips in its Zhenwu series, to over 400 customers, including automakers and financial institutions. The Hangzhou-based tech giant, which runs some of the country's biggest ecommerce platforms, has accelerated its pivot to AI in recent years. Its open-source Qwen AI model family is popular among developers, surpassing one billion cumulative downloads since its initial launch in 2023.

Alibaba Takes Aim At Nvidia With New 3X Powerful AI Chip And Next-Gen LLM Model As China Tech War Heats U

The Zhenwu M890, Alibaba's latest AI chip, delivers triple the performance of its predecessor, the Zhenwu 810E. The new processor, designed to handle memory-intensive agentic AI workloads, boasts 144 GB of GPU memory and an interchip bandwidth of 800 GB per second, reported CNBC. At a conference in Hangzhou, Alibaba stated it has already shipped 560,000 Zhenwu units to over 400 customers across 20 industries. This new chip could potentially enhance Alibaba and its chip subsidiary T-Head's competitiveness in China's burgeoning domestic AI processor market, which includes rivals such as Huawei and Cambricon. Alibaba also announced on Wednesday that it will soon launch its next-generation AI model, Qwen3.7-Max. Alibaba Challenges Nvidia In China Alibaba's AI models and applications are expected to generate 30 billion yuan ($4.42 billion) in recurring revenue by year-end, contributing significantly to the company's cloud-computing revenue. The company's increased investment in AI infrastructure and cloud services is a response to rising demand. That being said, SemiAnalysis analyst Myron Xie told CNBC that Alibaba's latest chip still trails leading Western rivals in memory capacity and bandwidth, with key compute performance metrics yet to be disclosed. Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Image via Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

Alibaba Targets NVIDIA's Hopper With Zhenwu M890 AI Chip, Claiming 3x The H20 Performance, 144GB HBM3 & A Roadmap Through 2028

Alibaba has unveiled its latest AI chip, "Zhenwu M890" and AI LLM "Qwen3.7-Max", designed for Agentic AI workloads. As Agentic AI Rages On, Alibaba Rolls Out Its Own AI Chip & AI LLM: Meet Zhenwu M890 GPU & Qwen3.7-Max Model The Alibaba Zhenwu M890 is based on the company's in-house PPU (Parallel Processing Unit) architecture and features a Transformer core engine. The chip is designed for Agentic AI workloads with a focus on AI inferencing, offering 0.6 PFLOPs of FP16 (Half-Precision) compute, which is comparable to the A100 from NVIDIA, and three times faster than the Hopper H20 solution. The company also states that the M890 AI chip offers 3x the compute performance of the previous generation offerings. In terms of specifications, the Zhenwu M890 is equipped with 144 GB HBM3 memory, a 50% increase over the Zhenwu 810E, which packed 96 GB memory. The interconnect bandwidth is also boosted to 800 GB/s, up 100 GB/s from the 810E chip. In addition to that, the new chip supports FP32, FP16, FP8, & FP4 formats for AI workloads. This puts the chip on par with the capabilities of NVIDIA's Rubin and Huawei's Ascent 950 series. The company is offering a full ecosystem with the introduction of a new interconnect chip, called ICN Switch 1.0. This chip offers 25.6 Tb/s of interconnect speeds, at a P2P time delay of less than 150ns. The higher bandwidth enables support for massive agent concurrency. There's also the Yitian Arm-based host CPU and Panmai series networking cards, which will all come together within the Panjiu AL128 Supernode Server by Alibaba Cloud. This new server will tightly integrate 128 AI accelerators within a single rack, delivering PB/s scale bandwidth. T-Head reports that they have shipped approximately 560,000 Zhenwu AI chips to date, with more than 400 external customers spanning across 20 industries. Looking ahead, Alibaba Cloud is working on a series of Zhenwu chips following the M890. Next year in Q3, the company plans to introduce the V900, which will feature an updated architecture, delivering a 3x performance boost, 216 GB of memory, and 1200 GB/s of bandwidth, and the follow-up, the Zhenwu J900, will arrive in Q3 2028 with even more architectural and performance updates. The model delivers exceptional agent capabilities across diverse domains. As a frontier-level coding assistant, it supports coding tasks from rapid frontend prototyping to complex, multi-file software engineering. To enhance office work productivity, it reliably orchestrates multi-agent workflows to tackle sophisticated operations. Notably, Qwen 3.7-Max can autonomously execute long-horizon agentic tasks -- sustaining continuous operation for up to 35 hours and managing over 1,000 tool calls without performance degradation. Deeply optimized for leading agent frameworks including OpenClaw, Hermes Agent, Claude Code, Qwen Paw and Qoder, it serves as a reliable backbone for different agent systems. The model achieves top-tier results across major benchmarks in coding, general-purpose agents, general capabilities and multilingualism, making it competitive with leading frontier models. It will be soon accessible through Alibaba's model service platform Model Studio for global developers. Alibaba Cloud Besides the chips, Alibaba Cloud is also launching its latest AI LLM, Qwen3.7-Max. This model is focused on advanced agentic coding, complex reasoning, and long-horizon task execution. The new model will be available to developers and enterprises soon. Follow Wccftech on Google to get more of our news coverage in your feeds.

Alibaba Launches Zhenwu M890 AI Chip to Boost China's Domestic Market

Alibaba has reached a huge technological milestone that shocked many as it announced a new artificial intelligence chip on Wednesday and said it delivers three times the performance of its predecessor. The chip, better known as the Zhenwu M890, also adds 144 GB of GPU memory and 800 GB per second of interchip bandwidth. The launch comes as Nvidia faces growing barriers in China. At the same time, Alibaba said it has already shipped 560,000 Zhenwu units to more than 400 customers across 20 industries.

Alibaba Unveils New AI Chip, Upgrades AI Model

Alibaba has unveiled a new artificial intelligence chip as well as an update of its AI model, as the Chinese tech giant aggressively pursues its AI ambitions on multiple fronts. The Hangzhou-based company released a new chip, the Zhenwu M890, which delivers three times the performance of its predecessor Zhenwu 810E, at a conference in Hangzhou on Wednesday. These capabilities make it "exceptionally" suited for complex agentic AI workloads, which demand extensive working memory for context retention and high-speed communication, Alibaba said. Alibaba also released an update of its large language model, called Qwen3.7-Max, which is engineered for more advanced agent coding and complex reasoning. Alibaba's management said in a recent earnings call that scaling up the deployment of its in-house chips represent "the highest value for money compute power," which will improve Alibaba Cloud's margins. Alibaba's chip unit has achieved widespread industrial adoption of its AI chips, with over 560,000 Zhenwu units delivered to date, Alibaba said. Alibaba's AI investment has significantly dragged on its profitability in recent quarters but the company has expressed confidence that the investment will pay off. The company expects AI-related product revenue to count for 50% of cloud unit's external revenue in about a year, and become the primary driver of revenue growth for that unit, chief executive Eddie Wu said earlier this month.