Editorial Team
6 min read
Vice Premier Ding Xuexiang’s visits to Huawei and computing facilities connect China’s domestic AI-chip push to a national effort to schedule compute, data and power. The harder test is whether those systems can make domestic hardware efficient and dependable while Chinese developers and cloud rivals reduce their reliance on Huawei.
China’s push for domestic AI hardware is moving beyond a question of who can design an accelerator. Ding Xuexiang, the vice premier who also sits on the Communist Party’s Politburo Standing Committee, has paired an inspection of Huawei’s chip-research laboratory with a call for a national integrated computing network. That makes his intervention consequential: it joins central policy authority to the still-unsettled task of turning domestic silicon into usable computing services.
The evidence does not establish that this is a Huawei mandate, or that domestic capacity already matches the leading foreign alternative. It does show a more demanding strategy: coordinate chips, model algorithms, data supply, network scheduling and power. That can expand the market for Chinese hardware, while also raising the standard it must meet.
In Shanghai, Ding visited Huawei’s Lianqiu Lake research center and spoke with its founder, Ren Zhengfei, in a chip basic-technology laboratory, the account of the visit said. It did not disclose their conversation. The same account placed Huawei alongside visits to battery maker CATL, the Chinese Academy of Sciences and Huairou National Laboratory during a broader basic-research tour.
Ding urged leading technology companies to pursue original “from 0 to 1” breakthroughs and tackle problems at their source and underlying layers. The visit therefore signals political backing for foundational research; it is not evidence of a procurement decision or an exclusive role for Huawei.
Later in May, during inspections in Beijing, Hebei and Inner Mongolia, Ding called for top-level planning of a national integrated computing-power network. The official account says the network is meant to coordinate supply and demand, connect dispersed resources, improve access and matching, and link computing capacity with power systems. It also calls for domestic hardware and software to move from merely “usable” to “highly effective.”
That last formulation is the important qualification. A machine count alone cannot show software compatibility, workload scheduling, energy availability or service quality.
China’s manufacturing-AI action plan calls for a secure, reliable supply of key core AI technologies by 2027. It sets policy targets—rather than reported outcomes—for three to five general-purpose large models deeply applied in manufacturing, 100 high-quality industrial datasets and 500 typical application scenarios. The plan’s summary also calls for coordinated AI-chip hardware and software development and innovation in model-training and inference methods.
Those targets explain why the chip-lab visit and the compute-network push are connected. A domestic accelerator gains strategic and commercial value only if model developers can use it, operators can deploy it and customers have workloads worth running.
The computing-network agenda is not only a capacity program. Ding said computing’s rising energy demand requires closer coordination of energy allocation and infrastructure development. In Inner Mongolia, an area the official account describes as having abundant wind and solar power, he encouraged direct green-power supply and energy-saving technologies. That places the economics of electricity alongside chip supply in the policy problem.
The same account calls for a market-oriented, application-driven approach that makes computing more efficient, accessible and user-friendly. It provides no measure of cost, utilization or delivered performance, so those outcomes remain unproven.
A July project report said a cluster in Shaoguan, Guangdong, uses 11,520 Huawei Ascend 910C accelerators across 30 supernodes and claims 9,000P of total computing power. It called the installation a core “Guangdong Computing” project and said it would be southern China’s largest operator-scale domestic 10,000-card training cluster.
The report also says Shaoguan has built capacity for 170,000 standard racks, has another 150,000 under construction, and has five 10,000-card clusters in operation with 40,000P of installed intelligent-computing scale. Those are local and project-side descriptions, with no disclosed common workload, power measurement, utilization rate or comparison against a foreign system. They demonstrate physical buildout, not performance parity or commercial viability.
Bernstein estimated Nvidia and Huawei at about 40% each of China’s AI-chip market in 2025; its 2026 forecast put Nvidia at about 8% and Huawei at about 50%. Source: Associated Press.
U.S. export controls have altered the Chinese AI-chip market. A Bernstein analysis cited in a market report estimated U.S.-based chipmaker Nvidia and Huawei each held about 40% of China’s AI-chip market in 2025, and forecast Nvidia at about 8% and Huawei at about 50% in 2026. Those are analyst estimates and forecasts, not audited shares; the report does not provide their market definition or a like-for-like performance test.
The same report says industry analysts view Huawei’s Ascend 950 series as roughly comparable with Nvidia’s H200 by some measures. But a separate report based on three unidentified people says Huawei’s offerings still lag Nvidia’s most advanced chips by a wide margin. Both descriptions can be true because they use different comparison points; neither proves across-the-board parity.
That gap also helps explain why the market is not settled around Huawei. The same report says Hangzhou-based AI developer DeepSeek—known for two efficient models that went viral globally—has used Nvidia and Huawei hardware but is developing an inference chip of its own. Inference is the stage at which a trained model generates responses, not the training of new models. The project was described as early-stage, with discussions involving chip-design, foundry and memory companies; DeepSeek did not respond to a request for comment.
The reported effort is a bid for more control over the hardware behind DeepSeek’s models, but it faces the same industrial bottlenecks as other Chinese designers. The report says advanced overseas foundries are unavailable under U.S. restrictions and that separate curbs have limited access to high-bandwidth memory, which is critical for AI inference chips. It also says competitive chip design typically takes years and substantial capital. DeepSeek was reported to be planning a first outside funding round of $7 billion at a $52 billion-to-$59 billion valuation, a departure from its earlier rejection of outside investment; that financing was reported as planned, not completed.
Chinese technology groups Alibaba and Baidu are also developing their own AI chips, according to the same report. The relevant contest is therefore broader than Huawei against Nvidia: Chinese firms are simultaneously trying to control more of their own hardware supply while sharing constraints in fabrication, memory, power and deployment.
The next evidence needed is operational. The reported policy and construction figures do not show which systems deliver a given workload most reliably or cheaply, how much capacity is utilized, what it costs to port and run software, or how power use compares over the same task and period.
Comparable measurements on those points would clarify whether the national network is converting government direction into dependable domestic compute—or chiefly adding capacity whose economics and performance remain unresolved.
Sundar Pichai said Google supports an industry letter on open models, aligning the company with developers and technology firms resisting broad limits on Chinese AI systems while leaving allegations of IP theft and national-security risks unresolved.
Brazilian courts and bar associations have responded to hidden prompts in legal filings with fines, referrals, suspensions and an investigation. The known cases were detected, however, and do not show that an AI system changed a judicial decision; they instead test how courts secure AI-assisted review of untrusted case documents.
Salamanca City Central School District has put its proposed Realbotix humanoid robot and Optio AI assistant pilot on hold while it works on student-data privacy agreements. The pause leaves the district with a more basic task: establish what the digital tutor and the physical robot each collect, do and improve before either reaches students.
Samsung Electronics and Broadcom have signed an MOU covering memory, foundry technologies and advanced packaging for next-generation AI infrastructure. The companies expect more than $200 billion of collaboration over five years, but have not disclosed how that figure translates into purchases, production or capacity commitments.
OpenAI says GPT-5.6 Sol reached a 24.9% ExploitGym pass rate under a two-hour cap, up from GPT-5.5’s 15.1%. The benchmark measures a bounded step from a known vulnerability to an exploit; a separate assessment performed with OpenAI finds continued limits on hardened targets and long operations.
Chey Tae-won, chairman of SK Group, said Anthropic had sought SK Hynix component-supply cooperation for chips it hopes to make itself. But Anthropic has disclosed no chip plan, and the same summit also produced a separate letter of intent between Anthropic and SK Telecom on a large AI data center.
South Korea announced $950 billion in planned five-year semiconductor cooperation involving Samsung, SK Group and global technology companies, alongside prospective AI-data-center projects. The figure aggregates arrangements with different terms and disclosed commitments, leaving delivery, power and the overlap between the SK and NVIDIA figures unresolved.
OpenAI said it applied a fix after a July 25 disruption affected ChatGPT, its Codex coding service and parts of its API. The available reporting points to recovery in about 50 minutes, but the retained incident records do not explain the cause, the mitigation or whether the services failed through a common component.
Anthropic CEO Dario Amodei said the Claude developer has signed supply deals with Samsung Electronics and SK hynix. The available disclosures show a useful infrastructure relationship, but not an HBM purchase, a dedicated allocation or a claim on either supplier’s output.
Moody's Ratings expects six US hyperscalers to spend $785 billion in 2026 and close to $1 trillion in 2027. The forecast signals continued demand for AI capacity, but a separate Moody's analysis shows why capex alone understates the financing question: five of the companies already reported $969 billion in undiscounted future lease payments, much of it not yet on their balance sheets.
Anthropic has released Claude Opus 5 at Opus 4.8's API price and says it approaches Claude Fable 5 on coding and knowledge work. An Epoch AI listing places it at 159 ECI, below Fable 5 and GPT-5.6 Sol, leaving customers to test whether the cheaper model delivers better economics in their own workloads.
Nvidia plans to invest $1 billion in Naver as Brookfield signs a nonbinding term sheet for up to $9 billion toward a South Korean AI-factory project. The companies target a 200-megawatt expansion by 2028, but financing, definitive terms and construction remain ahead.
Nvidia and SK Group signed letters of intent for an AI-infrastructure initiative they value at more than $500 billion, centered on an up-to-2-gigawatt Korean AI factory and a long-term SK hynix memory partnership. The disclosed projects give the plan a technical shape, but not a capital breakdown, time horizon or basis for treating the headline figure as committed spending.
Debian has opened discussion of two General Resolution choices: a ban on direct LLM-assisted contributions that would amend its Social Contract, and a conditional-use policy that puts responsibility on contributors. The official text shows that neither policy has been adopted and that the permissive option has changed from its initial mailing-list draft.
Exposed attacker infrastructure contained logs showing the Hermes AI agent running unattended reconnaissance and privilege-escalation checks against systems associated with Thailand's Ministry of Finance. The artifacts show deliberate, operator-directed targeting, but they do not establish the initial access path, the full extent of compromise or data theft.
Amazon is closing its San Francisco AGI Lab amid undisclosed job cuts, while saying its frontier-model research under Pieter Abbeel and its Nova Act service will continue. The closure follows leadership departures and an AGI reorganization; Amazon has not said how many people are affected or where the lab’s work will go.
Delhi Police says its live facial-recognition deployment at Jantar Mantar is meant to find wanted people and prevent disruption at a large student protest. The public record still does not show the current match threshold, the legal basis and scope of the watchlist, how long footage is kept or the safeguards after an alert.
The Delhi High Court has denied ANI Media's request to immediately stop OpenAI from using its news material to train the models behind ChatGPT. Its prima facie fair-dealing finding gives OpenAI interim relief, but leaves the copyright suit, the evidence and the question of compensation unresolved.
A viral reel recorded at a Delhi trans-rights protest with Ray-Ban Meta glasses was initially left up by Instagram and later removed. The episode shows how a recording indicator and user rules can fall short when people filmed without consent must discover the camera, navigate moderation and contend with copies after the fact.
CXMT reportedly charged more than Samsung for comparable DDR5 server modules and signed large supply agreements with ByteDance and Tencent. The reported premium reflects tight supply and state-shaped domestic demand, while CXMT remains behind leading rivals in high-bandwidth memory and lithography capability.