AI compute access rules could change how lenders assess data centre revenue. The United States is weighing legislation that would extend export-control authority to certain forms of remote access, while Malaysia plans to require data centres to disclose more about their operations, imported chips and electricity use. For lenders, the question is whether customers can continue to use the computing capacity that supports projected cash flow, and how quickly an operator can replace that demand if access rules change.

How US AI compute access rules could reach cloud services

The US House passed the Remote Access Security Act, H.R. 2683, by 369 votes to 22 on 12 January 2026. The Senate received the bill the following day and referred it to the Banking, Housing, and Urban Affairs Committee. The bill would give US authorities a basis to regulate certain remote access by foreign persons to controlled items, including through cloud services. Its practical effect would depend on the final legislation and how regulators implement it.

A customer can rent computing capacity in another country without taking delivery of the chips. Reuters reported that some Chinese AI companies were seeking access to Nvidia-powered capacity overseas amid concern that Washington could tighten restrictions on remote use. For data centre operators, the exposure depends on who ultimately uses the capacity and whether that access remains permitted throughout the contract.

The policy debate drew renewed attention after Nvidia announced on 12 November that co-founder Jensen Huang had resigned as president, chief executive and a director. Its SEC filing says the board accepted his resignation on 11 November, effective immediately. Chief financial officer Colette Kress is serving as interim CEO while the board searches for a permanent successor. The legislative question for lenders remains which customers, technologies and forms of remote access any future rules would cover.

What Malaysia plans to monitor at data centres

Malaysia’s investment, trade and industry minister, Johari Abdul Ghani, said on 12 November that the government plans to put data centre disclosure requirements into law. He identified facility locations, operators, hyperscalers, electricity use and the number of imported GPUs as information the government wants to track. Bernama reported that 55 of Malaysia’s 117 approved data centre projects had started operating. The proposal would extend monitoring beyond the initial approval process. Final legislative text, implementation dates and the treatment of existing projects remain open.

Malaysia already applies strategic-trade controls to the export, transit and transshipment of specified advanced AI chips under a 2025 directive. The proposed disclosures would give authorities more information about facilities during operation. Johari also illustrated how electricity reservation charges can affect project economics. A facility that books 100 megawatts and uses only 50 could still be billed for the unused 50 under its supply arrangement. The financing implication depends on the power contract and on whether customer payments cover reserved capacity during a slower ramp-up.

How compute access could shape data centre credit covenants

Tenant credit remains important, but the named tenant may resell capacity to other users. Lenders can assess that exposure by examining who ultimately uses the equipment, where users connect from and whether the operator can verify eligibility when capacity is resold. Information rights, permitted-use clauses and obligations to notify lenders of regulatory changes can give them time to assess the effect on contracted revenue before a missed payment.

Equipment traceability also affects recoveries. Inventory records can establish where GPUs sit, who owns them and which assets the security package covers. A lender’s ability to repossess, relocate or resell the equipment depends on enforceable rights and any applicable trade restrictions. If a rule change limits the pool of lawful users or buyers, the project could face pressure on both utilisation and collateral value.

Power commitments can amplify that pressure. If an operator loses a customer while remaining liable for reserved grid capacity, its costs may stay elevated while it looks for replacement demand. Debt sizing can account for that scenario by testing how long the operator can service debt at lower utilisation, what liquidity is available and when the power commitment can be resized.

Where financing terms could diverge

Projects with identifiable end users, auditable equipment records and a diversified pool of eligible customers may be easier to finance. Projects that rely on a single reseller or assume unrestricted cross-border demand may face closer scrutiny of leverage, reporting and liquidity reserves. The effect on valuation will depend on whether operators can replace restricted demand quickly enough to maintain revenue and cover their fixed power costs.

The Perennial view

Perennial will watch the progress of H.R. 2683, Malaysia’s legislative text and whether lenders begin requiring end-user reporting, equipment inventories or specific responses to a change in access rules. The credit test is whether a facility can maintain lawful utilisation and service debt if its original customer mix changes. Evidence that operators can replace affected customers without weakening cash flow would limit the impact on financing terms.

Explore more analysis on AI infrastructure and private credit at Perennial Markets.

Where Insight Meets Opportunity.

Sources

Disclaimer. This material is general financial information and market commentary only. It does not constitute personal financial advice, an offer, a recommendation or a solicitation. Consider objectives, financial circumstances, eligibility, legal restrictions and independent professional advice before acting.