Global restrictions on exporting cutting-edge AI technology were two days away from implementation in the US when the Bureau of Industry and Security (BIS) announced it was rescinding them. Known as the ‘AI diffusion rule’, BIS warned the restrictions would have ‘stifled American innovation’ and ‘undermined US diplomatic relations’.
A replacement rule is on the way, with an expected path to country-by-country negotiations. The United Arab Emirates has already emerged as among the first to secure greater access to advanced AI chips as part of an agreement to build an AI campus, according to reports of President Trump’s trip to the Middle East.
While we await details, it’s important to understand the significance of the shift and the possible implications of the new setting.
King & Wood Mallesons partner Daryl Cox and L.E.K. Consulting partner Aakash Gandhi met on 13 May 2025 – the day the BIS announced the rescission – to share timely insights on the export restrictions, including tips on meeting new standards and mitigating potential risks.
In this post we share key takeaways from their discussion. You can watch the webinar here.
‘This shift in policy is both a challenge and a chance for nations to redefine their positions in the global AI race. For the industry, it’s about building capacity and being ready to meet the market’s evolving needs. As new rule unfolds, the ability to understand and adapt will determine who thrives in this new era.’ Daryl Cox
‘This policy shift has opened a new playing field for countries like Singapore. While there’s relief, the game is far from over. China is ramping up its domestic capabilities, and everyone else must keep pace.’ Aakash Gandhi
The ‘3D chess’ game behind AI restrictions
Restrictions on AI technology have accelerated over the past few years, driven by national security concerns and a bid to win the global AI race. The US agency behind the rules (BIS) and US chip makers have engaged in a strategic back-and-forth game of restrictions and launches. The latest regulatory effort – the AI diffusion rule – targeted high-performance AI chips and model weights (the ‘brains’ behind AI).
The Trump administration turn-around means countries now have the chance to strike deals directly for AI technology access.
Simplified framework: A new direction
A leaner framework is promised, scrapping the ‘tiered’ system and allowing countries to negotiate access to AI technology on a case-by-case basis.
Existing chip export limits remain strictly enforced, with particular attention on countries redirecting technology to China.
Expect some countries to welcome this change. Others might find themselves navigating a maze. Watch this space as countries start bargaining to bring AI capabilities home.
Winners and losers: Who stands to gain?
The shake-up in AI policy offers new openings for countries like Singapore. But while these nations might breathe a sigh of relief, challenges still loom.
China is intensifying its investment in domestic AI capabilities and Huawei is making waves with its advances in AI chip development. China’s focus on self-reliance means it’s working hard to fill any gaps left by US restrictions. That’s both an opportunity and a challenge for other players in the region.
Implications for investors and operators: Staying agile
Investors and data centre operators must adapt quickly to the changing demands and regulations. AI demands high-end GPU* power, and while demand is shifting, it’s not disappearing. There are new opportunities in places like Johor, Batam, and Sarawak.
For operators, it’s about gearing up for high-density designs. Liquid cooling and cutting-edge infrastructure are key. The market is splitting into those who can adapt and thrive versus those who might struggle.
* ‘GPU’ stands for graphics processing unit. Originally built for gaming and 3D graphics, GPUs are designed to handle many tasks at once, making them ideal for training AI models. GPUs are now widely used for tasks like AI training and data analysis because they can process large volumes of data faster than regular computer processors. Well known providers of GPUs include NVIDIA and AMD.
For more AI technology key terms, check out our short list of definitions in the ‘Story so far’ section of our recent AI diffusion rule insight.
Pockets of Southeast Asia: Continuing restrictions
While the policy shift offers some relief, the US is keeping a close watch on countries that might divert technology to China.
Strict enforcement of existing chip export rules is still the norm. Any Southeast Asian nation that has acted as a channel for AI technology to reach China in recent years needs to tread carefully. These countries must navigate a delicate path between compliance and opportunity.
Navigating grey areas: Compliance challenges and opportunities
One major challenge is the GPU-as-a-Service model. This approach lets companies rent computing power instead of buying chips.
For cloud providers, compliance requires vetting customers and managing AI services with precision. It’s about balancing access with accountability. The promise of ‘strict’ enforcement reinforces the need to understand any restrictions and apply them as intended.
What next? Watch this space (and watch the webinar…)
The discussion on AI export controls is far from over. As policies evolve, we’ll keep the conversation going. Stay tuned as we navigate these changes together.
Want to know more about AI and data centres? Subscribe to our Tech & Data insights to stay up to date.
- AI race heats up: What US export controls mean for Asia
- Enforcement starts now: Grace period for defence export controls is over
- Can the data centre goldrush go green? Malaysia’s Johor shows we can move fast, but more is needed
