Tencent warns US chip curbs may slow next AI model | Magica
Tencent warns of AI model delay as US chip curbs unsettle Southeast Asian data centres
Editorial Team
••📖6 min read
Tencent says US restrictions on remote computing will likely slow its next AI model while leaving cloud revenue largely protected. Southeast Asian operators face licensing and financing uncertainty, though US cloud access and domestic demand offer partial alternatives.
Tencent says efficient inference can protect cloud revenue even as restricted computing access slows its next model.
Regional operators face separate hurdles to obtaining advanced chips and serving Chinese customers at scale.
US cloud services offer an alternative for regional AI users; licensing uncertainty still threatens investment in locally owned capacity.
Tencent Holdings, the Chinese technology group whose businesses include gaming, financial services and cloud computing, said on November 18 that new US restrictions on remote access to advanced AI chips would likely slow its next model’s development. It expects no substantial adverse effect on cloud revenue, a reassurance resting on the difference between developing models and running existing ones.
The measures also challenge Southeast Asia’s role as a host for computing rented by Chinese companies. Operators need clarity on both chip supplies and permitted customers before committing capital. Access through US cloud providers could sustain regional AI use, but does not settle who will own the infrastructure or capture the investment.
Tencent’s company-reported capital expenditure rose from 52.8 billion yuan to 81.4 billion yuan between the June and September quarters of 2026. These are group totals, not an itemised AI bill. Source: MarketScreener / Reuters.
Tencent’s revenue defence has limits
On its post-earnings call, Tencent said its cloud business primarily handles inference—the use of an existing AI model to produce outputs. Executives said more efficient models could run on relatively few domestic chips, according to the account of the call and quarterly results. That is management’s basis for expecting limited damage to cloud sales.
The company nevertheless said reduced access to advanced computing would affect model training and likely slow its next model in the near term. Management expects greater reliance on domestic chips and supply chains to accelerate their development over time. That forecast does not establish that Chinese hardware can already replace the restricted capacity.
Tencent’s earlier experience also qualifies its reassurance. A year ago, it said chip shortages were constraining cloud growth because it prioritised internal AI use over renting computing power to outside clients. The latest controls arrive on top of an existing allocation problem.
Binance Wallet says subscriptions for its Anthropic IPO xStocks campaign reached the 500 million USDC limit. Final allocations, token delivery and applicable refunds remain pending; the cap does not establish proceeds received by the AI company.
Editorial Team
Tencent reported group capital expenditure of 81.4 billion yuan in the September quarter, compared with 52.8 billion yuan in the June quarter—about 54% higher, calculated from those figures. These are total capital expenditures, not an itemised AI bill. First-half spending of 84.7 billion yuan had already exceeded the roughly 79 billion yuan spent in all of 2025.
Revenue from fintech and business services, which includes cloud, rose 9% to 63.7 billion yuan in the third quarter, with AI-related cloud demand a key driver. The segment’s results do not disclose cloud revenue separately. Group revenue grew 8% to 207.3 billion yuan, below the 210.8 billion yuan analyst consensus and slower than the previous quarter’s 11% growth as gaming momentum cooled. Rising spending makes the durability of AI demand consequential even if cloud sales withstand the immediate restrictions.
Two licensing hurdles for regional operators
The US Commerce Department’s Bureau of Industry and Security, the agency issuing the controls, described its November 16 interim final rule as a way to prevent remote computing in third countries from circumventing existing technology restrictions. Its announcement says the rule was written by and at the direction of the Treasury Department and National Security Council. It seeks to align remote-access protections with controls on physical shipments.
The new requirements cover advanced chip exports to Singapore, Malaysia, Thailand, Indonesia and Brazil, markets Washington identifies as presenting heightened diversion risks. According to the regional account of the requirements, non-US-headquartered operators need licences to obtain new advanced chips, while US-headquartered companies are exempt provided shipments are reported.
A separate requirement concerns customers: overseas data centres supplying significant computing capacity to China-headquartered companies must obtain US export licences by mid-February 2027 to continue those services or face future export restrictions. This exposes operators both when acquiring hardware and when selling access to it.
The affected businesses extend beyond cloud rental companies. Telecommunications operators, research institutes and government-backed AI projects seeking to own chips also face the procurement hurdle, Singapore-based AI adviser Matthias Hendrichs said in the regional report. Businesses renting capacity from US cloud providers would be less affected, he said.
The restrictions follow scrutiny of Megaspeed, a Singapore-based cloud computing company and Nvidia customer facing a US investigation into possible circumvention of chip export controls. Malaysia is investigating its local unit. Megaspeed denies the allegations and says it takes its US compliance obligations seriously, according to the account of Malaysia’s response. The investigations do not establish wrongdoing.
UN Trade and Development figures cited by CNA show data-centre investment in the first three quarters of 2025. These historical amounts do not estimate investment at risk from the November 2026 controls. Source: CNA.
Financing and trust determine the regional exposure
Hendrichs warned that projects with Chinese links could face months of delays and that uncertain chip arrival dates could hold up lending. He cited Malaysian data centre deals put on hold before the new rule, a limit on attributing existing delays to this week’s measures.
Malaysia’s trade minister, Johari Abdul Ghani, said he was unclear how individual companies’ financing requirements would be affected but urged operators to comply. In separate remarks to Malaysian media described in the same account of Malaysia’s response, he said the government would seek US clarification on licensing and services to third parties.
The report, citing state news agency Bernama’s account of Johari’s remarks, said Malaysia would notify all 117 approved data centre projects once requirements were communicated. These are approved projects, not a count of operating facilities. Johari also said Malaysia had yet to receive evidence that tenants traded or re-exported chips in breach of US rules, and that transactions involving non-US customers could require more extensive US due diligence.
Regulatory credibility is another constraint. In the regional reporting, ISEAS-Yusof Ishak Institute fellow Maria Monica Wihardja, who studies regional supply chains and technology, pointed to uneven safeguards: Singapore and Malaysia have export-control laws or strategic trade frameworks; Thailand has not fully implemented its framework; Indonesia relies on dispersed laws rather than a dedicated regime. Complex ownership structures can also obscure who controls data centre operations, she said. Stronger oversight could improve investment prospects, while further evidence of diversion could deepen reputational damage.
What licence decisions still need to establish
Singapore has begun talks with the US over trusted access for legitimate companies. Malaysia is seeking clarification before communicating the requirements across its approved projects. The next commercial test is whether operators can obtain licences for the equipment and customers on which their financing depends, in time to meet the February deadline for continued services.
There are reasons to resist predicting a uniform regional slowdown. Brokerage CGS International analyst Lim Siew Khee, quoted in a separate assessment of the consequences, expects enterprise cloud demand, government-backed AI and domestic digitalisation to sustain broader data centre investment despite weaker near-term AI-related spending.
But continued access through US providers is only a partial answer. That assessment also reports analysts’ concerns that costlier or restricted computing and AI models could slow adoption. It notes that Chinese companies can still lease US data centres, while warning that American cloud providers may avoid serving Chinese competitors. Formal availability therefore does not establish willingness to supply, affordability or equivalent access.
For Tencent, subsequent cloud results and the next model’s development timetable will test management’s distinction between resilient inference sales and constrained training. For regional operators, licence outcomes and financed projects will show how much computing investment can proceed. Neither the company’s reassurance nor analysts’ forecasts yet quantify the ultimate loss of capacity, revenue or investment.
Microsoft is rolling out Agent 365 Business Analytics through its Frontier early access program, pairing AI spending with estimates of assisted work. The dashboard expands its existing analytics as agent workloads add usage charges, but its dollar-value estimates do not establish financial returns.
Adobe’s Firefly update saves recurring visual subjects and creative preferences, extending an existing push into repeatable production. Memory has narrower reach than the assistant, while campaign demonstrations leave reliability and savings unmeasured.
Microsoft’s Ignite upgrades package enterprise knowledge, skills and tools into reusable business plugins. The expansion builds on IQ’s June release and competes with existing graph retrieval services; total savings and portability remain unproven.
The European Central Bank warns that debt and interlinked financing could carry an AI investment reversal into credit markets and the European economy. Resilient earnings, limited direct bank lending and stronger bank buffers temper its comparison with past bubbles.
Reported introductions helped Alice Huang secure Alibaba as a chip-rental customer and obtain financing. The relationship raises questions about Nvidia's overseas sales network, but does not itself establish illegal exports.
KPMG says AI agents pursued some issues more deeply in a parallel UK audit, though humans still had to correct their work. Client fee savings remain unconfirmed, even as KPMG International has negotiated lower fees from its own auditor by citing AI.
Payhawk announced a $280 million all-equity Series C led by Growth Equity at Goldman Sachs Alternatives, valuing it at $2 billion. The funding backs an expansion from managing business spending into broader AI finance operations, where competitors already combine automation, payments and controls.
A 16-country survey of 15- to 22-year-olds finds a median of 33% have used AI as a companion or friend. Users have higher adjusted odds of loneliness, but the research cannot establish whether AI helps, harms or replaces human connection.
South Korean President Lee Jae Myung has ordered a review of rewards of at least 100 million won for reporting technology-assisted exam cheating and legislation requiring technology companies to supply evidence. Ahead of the November 19 college entrance exam, the proposals would add informant incentives and post-exam investigations to classroom checks, but implementation rules remain unsettled.
Microsoft is pitching Copilot Workspaces as a shared environment for people and AI agents. Team collaboration is already in early access, but personal Autopilot integration is still coming soon, with final agent pricing unannounced.
Google is rolling out file references and sharing for Gemini skills as adult personal accounts lose the ability to use Gems. Automatic conversion preserves supported material, but file limits and missing tools can interrupt existing workflows.
Muddy Waters disclosed a short position in Nvidia while questioning a possible connection involving senior adviser Ajay K. Puri. Its own account leaves the key identity question unanswered, separating its new claims from an existing server-diversion prosecution.
Nvidia reported $115.8 billion in quarterly revenue in its first earnings release since co-founder Jensen Huang resigned. Strong demand gives interim CEO Colette Kress a financial cushion, while a campaign for Huang’s return collides with unresolved investigations and the board’s reported opposition.
OpenAI launched GPT-6.2 Astra and Sol for professional work, cutting Astra’s standard input and output token prices by 20%. The models offer cheaper paths through selected tasks, but research evaluations, reporting failures and unfinished independent safeguard testing complicate the case for broader autonomy.
Microsoft’s first live speech-to-speech model offers OpenAI API compatibility and planned enterprise integrations. Its published evaluations show a low-cost option, with several rivals completing more agentic tasks.
Canberra-based Trellis Data is offering its 220-billion-parameter Agentaus 1 model to business and government. Its pitch goes beyond keeping data in Australia, but performance, deployment costs and a planned capital raising remain key tests.
Perplexity is adding direct collaboration between frontier models to Computer, letting a lead agent coordinate up to 16 supporting agents. Company-reported benchmarks suggest mixed teams can outperform repeated copies of one model, but leave customer costs and production reliability unresolved.
Marsh reported 105 client cyber incidents in the last two weeks of October, against 48 in the final two weeks of May. The broker links the increase partly to AI, while warning that insurance panel access may not guarantee an available response team.