Malaysia: 117 Data Centre Projects Must Meet US Chip Rules Malaysia will notify all 117 approved data-centre projects to comply with new US restrictions on advanced AI chips, Trade Minister Johari Abdul Ghani told Reuters, calling the chips strategic assets. The restrictions, announced November 16, 2026, require US export licences for covered advanced chip shipments to Singapore, Thailand, Malaysia, Indonesia and Brazil, and require overseas data centres supplying a "significant volume of computing power" to China-headquartered companies to obtain licences within 90 days. Johari said he did not yet know how the rules would affect individual companies' financing requirements, leaving the compliance deadline for the 117 projects unconfirmed. Malaysia’s trade minister told Reuters that data-centre operators in the country should comply with the new United States restrictions on advanced artificial-intelligence chips. Johari Abdul Ghani described the chips as strategic assets, and he said he was unclear how the rules would change the financing needs of individual companies. The short version of his message: follow the rules first, count the cost later. The restrictions were announced on November 16, 2026. Under them, shipments of covered advanced chips to Singapore, Thailand, Malaysia, Indonesia and Brazil require a US export licence, a change we first unpacked in our piece on cloud KYC for AI chips and the five license countries https://shattered.io/cloud-kyc-ai-chips-remote-access-rule-five-license-countries/ . The same package also reaches overseas data centres that supply a “significant volume of computing power” to companies headquartered in China. Those facilities must obtain US export licences within 90 days to keep providing those services. Bernama, Malaysia’s state news agency, reported that Kuala Lumpur will notify all 117 approved data centre projects so they comply once the requirements are communicated. That single number turns an abstract US rule into a concrete administrative task for one government. This analysis walks through what is confirmed, what remains unknown, and how the story could shape the market for AI compute in Southeast Asia https://shattered.io/southeast-asia-startup-funding-october-3-68b-claim-unconfirmed-2026/ . Don't miss new tech stories on Google Add Shattered once in the Google app and our stories show up in your news suggestions. Add now https://www.google.com/preferences/source?hl=en&q=https%3A%2F%2Fshattered.io%2F What Johari Abdul Ghani Told Reuters Reuters carried the core statement. Johari Abdul Ghani, Malaysia’s Trade Minister, said operators of data centres in Malaysia should comply with the new US chip restrictions. He gave one reason, which is that the chips count as strategic assets. He did not frame compliance as a favour to Washington. He framed it as a basic condition of holding hardware that governments treat as sensitive. He also admitted a gap in his own knowledge. According to Reuters, Johari said he did not yet know how the rules would affect the financing requirements of individual companies. That admission matters. Data-centre builds run on debt and on long-term customer contracts, and lenders price in regulatory risk the moment a rule appears. A minister saying he cannot yet size that risk tells investors the government has no ready-made relief package. The tone is worth reading closely. Governments in the region have spent years courting data-centre investment, so a trade minister publicly telling operators to comply is a signal that Malaysia wants to stay on the right side of the licensing regime. It is also a signal to the companies themselves. If a project leans on US-origin chips, the minister expects its owner to treat the licence as a gating item, not an afterthought. The 117 Approved Projects Bernama Says Will Be Notified Bernama’s report supplies the operational detail. Malaysia plans to inform all 117 approved data centre projects about compliance once the relevant requirements are communicated. Two conditions sit inside that sentence. The first is that the government is waiting for the requirements to be communicated to it. The second is that notification happens after that step, not before. That sequencing suggests Malaysian officials are not drafting their own licensing scheme. They are acting as a relay between US rule-makers and local operators. A relay can move fast or slow, depending on how detailed the US guidance turns out to be. Nothing in the reporting sets a compliance deadline for the 117 projects, and we treat that deadline as unconfirmed. The count also raises a practical question. Approved does not mean operating. A project can hold an approval and still sit at the land, power or financing stage, with no chips installed. The reports do not break the 117 down by status, by owner or by the kind of customers each project expects to serve. Until those details surface, nobody outside the government can say how many of the 117 face direct licensing exposure and how many simply need to be told about it. What the US Rule Covers: Five Countries and a 90-Day Clock The US package has two moving parts, and Malaysia’s comments touch both. Reuters-sourced reporting describes a country list on the chip side and a services test on the data-centre side. Keeping them apart helps, because they hit different parts of the business. Chip shipments to five countries Shipments of covered advanced chips to Singapore, Thailand, Malaysia, Indonesia and Brazil now require a US export licence. The list puts four Southeast Asian neighbours and one South American market under the same requirement. For a buyer in Kuala Lumpur, that means an order that once moved on commercial terms now waits on a government decision in Washington. Our earlier coverage of the US chip rule, its five countries and the 90-day deadline https://shattered.io/us-ai-chip-rule-china-remote-access-five-countries-90-day-deadline/ lays out how the package first took shape. Overseas data centres serving China-headquartered firms The second part targets use, not just delivery. Overseas data centres that supply a “significant volume of computing power” to companies headquartered in China must obtain US export licences within 90 days if they want to continue those services. The wording ties the licence to the customer’s headquarters, not to the server’s location. A facility in Johor and a facility in Jakarta face the same test if the customer base matches. The reports do not define “significant volume.” They also do not say when the 90-day period starts. If it runs from the November 16 announcement, the window would close around February 14, 2027. That date is our arithmetic, not a confirmed deadline, so operators should wait for the formal text before planning around it. Confirmed Facts Versus Open Questions Early coverage of export rules tends to blur what officials said with what commentators assume. The table below keeps the two apart, using only what Reuters and Bernama reported. | Item | Status | Source | |---|---|---| | US restrictions announced November 16, 2026 | Confirmed | Reuters-sourced reporting | | Licence required for covered chips to Singapore, Thailand, Malaysia, Indonesia, Brazil | Confirmed | Reuters-sourced reporting | | 90 days for overseas data centres serving China-headquartered firms | Confirmed | Reuters-sourced reporting | | Johari Abdul Ghani says operators should comply | Confirmed | Reuters | | 117 approved projects to be notified | Confirmed | Bernama | | Specific chip models and manufacturers covered | Unconfirmed | Not named in reports | | Named data-centre companies affected | Unconfirmed | Not named in reports | | Financing impact on individual firms | Unconfirmed | Minister says it is unclear | | Licence outcomes for individual Malaysian projects | Unconfirmed | None reported | | Exact compliance deadline for the 117 projects | Unconfirmed | None reported | Five rows say unconfirmed, and they cover the details that decide who pays. Anyone claiming to know which companies lose access, or which chips trigger the licence, is working from sources the public record does not show. Why Malaysia Calls AI Chips Strategic Assets Johari’s choice of words deserves attention. Calling the chips strategic assets places them in the same bucket as energy infrastructure or defence-adjacent technology. Once a government uses that language, it gains a rationale for oversight that does not depend on any single US rule. Malaysia can say the chips need careful handling for its own reasons, and the US licence then becomes one layer on top. The framing also protects Malaysia diplomatically. A statement that operators should comply with American restrictions could read as deference. A statement that strategic assets need to be handled within the rules reads as stewardship. Both lead to the same operational result, but the second keeps the political cost lower at home. For the AI model ecosystem, the logic is simple. Frontier training and large-scale inference both depend on dense clusters of accelerators. Whoever controls where those clusters sit and who rents them controls a large share of who can build and serve advanced models. That is why chip export policy has moved from a trade matter to a core piece of AI policy. Our look at the China dilemma facing US firms under the remote access rule https://shattered.io/nvidia-7-percent-apple-15-percent-china-dilemma-remote-access-rule/ shows how that pressure lands on the companies that sell into these markets. The Financing Question Nobody Has Answered Johari said he was unclear on how the rules would affect companies’ financing requirements. Financiers will not share that patience. A data-centre project typically lines up capital against two things: the hardware it expects to install and the customers it expects to serve. The new rule puts a question mark on both at once. Chips need a licence, and some customers may fall under the China-headquartered test. Lenders have three ordinary responses to that kind of uncertainty. They can raise pricing to reflect the risk, they can attach conditions precedent that require licences before drawdown, or they can pause new commitments until the text settles. The reports do not say which path any bank has chosen, and we do not claim any has acted. The minister’s comment simply shows that the government sees the issue and has no answer yet. Operators with diversified customer lists have the easiest road. A facility whose clients are mostly local enterprises, regional cloud buyers and global firms with no China headquarters has a clearer case for licences. A facility that built its business plan around a handful of large Chinese tenants faces a harder conversation, both with regulators and with the banks behind it. How the Rule Reaches Compute Rental and Cloud Operators The services test is where the package gets interesting. Earlier export controls focused on the box, meaning the chip or the server. This one reaches the rented hour. A company that never imports a single GPU can still fall into scope if it sells a large volume of compute capacity to China-headquartered customers. We covered the early signs of that shift in AI compute rental facing curbs across five countries and 90 days https://shattered.io/us-ai-chip-curbs-asia-data-centers-compute-rental-five-countries-90-days/ . That matters for the loophole debate. Critics of earlier rules argued that Chinese firms could reach restricted hardware by renting it abroad. Our past reporting on the remote access chip loophole claim https://shattered.io/nvidia-remote-access-chip-loophole-240000-gpus-7-5-billion-claim-unproven/ showed how contested the size of that gap was, with a headline figure that nobody had proven. The November 16 package answers the argument in policy terms by attaching a licence duty to the service provider. Cloud operators will now need customer-screening processes that go beyond a credit check. They must know where a customer is headquartered, who owns it, and how much capacity it consumes. That is a compliance function many regional providers did not need to build before. Expect hiring and tooling spend on this front, even though no company has announced figures. Historical Context: Export Controls Move From Chips to Services US limits on advanced chips for China did not begin this month. They have tightened in several rounds over the past few years, and each round closed a path that the previous one left open. The first wave focused on the most capable accelerators. Later waves pulled in more chip variants and equipment. Third-country shipments came next, because rerouted orders blunted the direct limits. The November 16 package continues that pattern. By naming five countries for chip licences, it addresses the transshipment route. By adding a services test, it addresses the rental route. What remains is enforcement, and that is the weakest link in any export regime. A licence only works if somebody checks who actually uses the hardware. Washington has also faced questions about where its chips went after sale. Our coverage of the 600,000 Nvidia H200 chips reported idle in Hong Kong https://shattered.io/nvidia-h200-chips-hong-kong-600000-idle-beijing-import-approval/ shows how complicated the picture gets when supply, demand and Beijing’s own import approvals collide. Policy on one side of the Pacific rarely produces tidy results on the other. Competitive Comparison: Where Each Layer of the Rule Bites The five countries on the list share one chip rule, but the burden differs by role. The second table sorts the package by the layer of the market it touches. It relies on the confirmed rule text and flags the open point in each row. | Market layer | What the rule requires | Who acts | Open question | |---|---|---|---| | Chip buyers in the five countries | US export licence for covered advanced chips | Buyer and supplier | Which chip models count as covered | | Data centres serving China-headquartered firms | Licence within 90 days to continue service | Data-centre operator | What counts as a significant volume | | Malaysian approved projects | Government notification, per Bernama | Malaysian authorities | Deadline for the 117 projects | | Lenders and investors | No direct duty reported | Banks and funds | How financing terms will shift | | Chip suppliers | Licence applications for shipments | Supplier | Approval timelines and volumes | Malaysia stands out only because its minister spoke on the record. The reports here do not include statements from Singapore, Thailand, Indonesia or Brazil, so we cannot say how those governments will handle the same rule. What the table does show is that Malaysia’s role sits in the middle layer. It is not the rule-maker, and it is not the buyer. It is the host government that decides how quickly and how firmly the message reaches operators. Market Impact for Operators, Suppliers and Lenders No company has published an impact estimate, and we will not guess one. The direction of pressure is clear enough to describe in plain terms. Operators with US-chip clusters and China-headquartered customers face the sharpest short-term strain. Operators with clean customer lists may gain, since licensed capacity in a compliant jurisdiction becomes more valuable when rivals lose access. Chip suppliers face a paperwork load. Each shipment to a covered country needs a licence, and approval speed will shape delivery schedules. Delays ripple into data-centre commissioning dates, which then ripple into revenue timing for the operator. Suppliers also face reputational risk if a licensed chip later turns up in a restricted use. For the AI model market, the effect is indirect but real. Labs and startups that rent regional capacity may see pricing shifts as supply tightens or reroutes. Teams building on open-weight models and hosting them in Southeast Asia should check their provider’s licence plans. The Vance remarks we tracked in his concern over China chips in third countries https://shattered.io/vance-concerned-china-chips-third-countries-zero-shipments-confirmed/ already hinted that Washington was watching these hubs closely. What Data-Centre Operators Should Do Now Operators in Malaysia do not need to wait for a formal government letter to start preparing. Several steps cost little and help in any scenario. We frame these as practical analysis, not legal advice, and firms should consult trade counsel for their own facts. - Map every installed and ordered accelerator by model, vendor and delivery date, so a licence review can start quickly. - List customers by headquarters country and by share of capacity, which gives a first read on the China-headquartered test. - Check contracts for clauses that let you suspend or reprice service if a licence is denied. - Brief lenders early, before a covenant review forces the conversation. - Watch for the formal notification from Malaysian authorities and log the date it arrives. Smaller operators face a harder job, because they lack in-house trade lawyers. Industry groups and the relevant ministry are the likeliest places for shared guidance, though the reports name no such programme. Until one appears, the 117 projects will each work from the same incomplete picture. Five Predictions for Malaysia Data Centres and US AI Chip Rules These are our forecasts, not reported facts. We base them on the confirmed rule text and on how earlier export regimes behaved. - Notification arrives in waves. Malaysia will likely contact larger, better-documented projects first and smaller ones later, because the 117 vary widely in maturity. - Customer screening becomes a product. Regional cloud providers will market verified-compliant capacity to global buyers who want a clean audit trail. - Lenders add licence conditions. New data-centre loans in the five countries will probably carry export-licence conditions before funds release. - Neighbouring governments echo the message. Other covered countries may issue their own guidance, since Malaysia’s public comments set a template. - The 90-day clock drives a rush of filings. Operators serving China-headquartered customers will either file for licences, restructure contracts, or wind down that business before the window closes. Any of these could fail. A clarification from Washington on the definition of significant volume could shrink the affected group overnight. A broad definition could do the opposite. The honest position is that the next two weeks of guidance will matter more than any forecast today. Key Dates and Signals to Watch The third table collects the dates and signals that will tell readers whether this story is settling or escalating. The dates come from the reporting, and the signals are our suggested checkpoints. | Date or signal | Why it matters | Status | |---|---|---| | November 16, 2026 | US restrictions announced | Confirmed | | November 18, 2026 | Malaysia’s trade minister comments reported | Confirmed | | Formal notice to the 117 projects | Starts the practical compliance process | Pending, no date reported | | Start of the 90-day window | Sets the real deadline for China-serving data centres | Not specified in reports | | Definition of significant volume | Decides how many operators fall in scope | Not specified in reports | | First licence decisions for Malaysian sites | Shows how strict reviewers will be | None reported | Readers tracking the story should look for three things first. The first is the formal text, which should define covered chips and the services threshold. The second is any statement from other covered governments. The third is the first sign of how fast licensing offices move, since speed will matter as much as policy. The Takeaway on Malaysia and US AI Chip Rules Malaysia chose compliance, and it said so in public. Johari Abdul Ghani told Reuters that operators should follow the rules, Bernama reported that all 117 approved projects will hear from the government, and the underlying US rule gives China-serving data centres just 90 days to secure licences. Those are the confirmed facts. Everything else remains open. The covered chip models, the affected companies, the financing impact and the project-level outcomes have no public answers yet. For readers following AI infrastructure, the story belongs in the wider AI and machine learning coverage https://shattered.io/ai-machine-learning/ on this site because chip access sets the pace for model training and deployment everywhere. We will update this analysis as formal guidance and licence decisions arrive. FAQ: Malaysia Data Centres and the US AI Chip Rules What did Malaysia say about the new US AI chip rules? Trade Minister Johari Abdul Ghani said Malaysian data-centre operators should comply with the new US restrictions on advanced AI chips, according to Reuters. He called the chips strategic assets. How many Malaysian data centre projects are affected? Bernama reported that Malaysia will inform all 117 approved data centre projects to ensure compliance. The reports do not say how many of the 117 hold or expect covered chips. Which countries need a US licence for covered chips? Shipments of covered advanced chips to Singapore, Thailand, Malaysia, Indonesia and Brazil require a US export licence under the restrictions announced on November 16, 2026. What is the 90-day requirement? Overseas data centres that supply a significant volume of computing power to companies headquartered in China must obtain US export licences within 90 days to continue those services. The reports do not define significant volume or state when the 90 days begin. Which chips and companies are covered? That is unconfirmed. The reports do not name the chip models, the manufacturers or the data-centre companies affected. Will the rules change data-centre financing https://shattered.io/softbank-7-billion-quarterly-loss-us-data-centre-financing-costs-2026/ ? Nobody knows yet. Johari told Reuters he was unclear how the rules would affect individual companies’ financing requirements, and no lender has commented in the reports we reviewed. Is there a compliance deadline for the 117 projects? No exact deadline has been reported. Bernama said notification will follow once the requirements are communicated.