Jensen Huang’s exit is a body blow to a company he has embodied since co-founding it in 1993. Jensen is Nvidia and Nvidia is Jensen.
Under his relentless leadership and aggressive drive to stay ahead of rivals, Nvidia has climbed to Olympian heights in the AI boom.
Its graphics processing units are the dominant chips for training and running AI models, and they turned the company into a juggernaut that booked US$215.9 billion in revenue in its last fiscal year. His spectacular downfall leaves many threads hanging.
First, how it unravelled. A news article reported Huang’s relationship with Alice Huang, an executive at Singapore-based Megaspeed, whom US officials alleged is a Chinese intelligence asset. Beijing denies the allegation. Huang confessed to the affair in a letter to staff, and soon the board opened an inquiry into whether he had been targeted by a foreign intelligence operation.
Then came reports of a second relationship, with an Nvidia executive who reported to him. The United States Justice Department has opened its own inquiry, and three House committee chairs have asked him to brief lawmakers privately on November 18.
What began as extramarital affairs has become a scandal with an enormous blast radius. It reaches beyond his family and the future of the AI industry to investments in AI companies and to US competitiveness.
Strip out the personal drama, though, and what remains is more troubling – a customer that, on the public record so far, hasn’t broken any law.
Megaspeed was spun out of a Chinese games company in 2023 and quickly became an official Nvidia cloud partner, with priority access to chips. Through its Malaysian subsidiary, Speedmatrix, it bought at least US$4.8 billion of Nvidia-based equipment in under two years.
Much of this came through the US arm of Inspur, a Chinese firm Washington has blacklisted. The computing power was rented out to Chinese customers, including Alibaba.
Every link in that chain appear to be defensible, on paper. An American chip company sells to an American server maker. The server maker ships to a data centre in Malaysia, which is not China. The Malaysian operator then rents out computing time over the Internet.
Under current US rules, renting AI chips to Chinese companies for use outside China is allowed. It becomes unlawful only if the hardware is shown to have moved into China, or the operator is shown to be acting on behalf of a Chinese entity.
Megaspeed denies wrongdoing and said checks by US and Malaysian authorities and by Nvidia have found no violations. Nvidia said its site visits confirmed the GPUs were where they were supposed to be. This is where it gets complicated.
Controls built for a world of boxes
Export controls were written for objects that cross borders. AI capability travels as a service. A model trained on data centre racks in Johor, Malaysia, is as useful to an engineer in Hangzhou as one trained down the road. Washington has grasped this. It has not yet closed the gap.
In January, the US House of Representatives passed the Remote Access Security Act that would treat remote access to controlled chips as a regulated export. The bill has not been passed into law. In August, the Commerce Department was reported to be drafting its own remote-access rule, but none has been published.
In the meantime, policing has fallen by default to the vendor. Nvidia selected its cloud partners, ran the site inspections and vouched for the result. When the seller certifies the buyer, and the seller’s chief executive is personally close to the buyer’s executive, what you have is an honour system. Compliance takes a back seat.
Policing this properly means regulating compute as a service, not only chips as goods. That needs the statutory authority still stuck in the US Senate. Even then, a US law has limits.
A statute in Washington still has to reach a server hall in Malaysia, so the countries hosting the racks must carry part of the load.
Neither Malaysia nor Singapore is equipped for that today. Malaysia began requiring permits in July 2025 for exports and transfers of high-performance American AI chips, and its trade ministry says it has since tightened oversight and keeps up a dialogue with Washington.
Singapore police are investigating Megaspeed for suspected breaches of domestic law, as they did an earlier case of Nvidia servers declared for Malaysia that may have ended up in China.
Notably, Singapore stresses that it enforces its own laws, not America’s. Both regimes watch the box leaving the building. Neither watches access leaving over fibre.
To close the gap, first follow the money. The trail matters as much as the hardware. Who ultimately owns, funds and controls a company, behind the shells and nominee directors?
Banks must know their customers before opening an account. Operators renting top-end compute need not. The Financial Times reported that Alibaba, through a Singapore subsidiary, lent Speedmatrix money in June 2024, while Speedmatrix was renting compute to Alibaba. A customer financing its own supplier is precisely what such checks exist to surface.
Next, make lenders and landlords gatekeepers. Those who finance and house the racks are chokepoints too. Bridge Data Centres, Bain Capital’s data centre arm, removed Megaspeed from its Malaysian campus in April while raising billions in loans. Make them share liability, and they will move faster than any legislature.
Finally, track usage, not just location. Verification has to come out of the seller’s hands. The Chip Security Act would require exported AI chips to verify their location, typically by timing how long a signal takes to travel between the chip and known servers.
That calls for usage telemetry: a tamper-resistant log, generated by the chips or the cluster itself, of which accounts ran workloads, how much compute they used and for how long. Training a large model leaves a distinctive footprint, with thousands of graphics processing units (GPUs) running flat out for weeks.
Those logs should go to an auditor without a sales quota, not only to the vendor. Advocates of the bill say Nvidia’s GPUs already have the hardware to support location checks through firmware.
What Singapore and Malaysia should do
For Asean countries, none of this comes cheap. Compliance means negotiating with Washington, with data centre operators and with the companies renting their capacity.
Malaysia has marketed itself as the neutral data centre hub, and Johor’s boom rests on that pitch. Singapore is where the holding companies sit. But neutrality that consists of not asking who the tenant is will not survive the US’ next rule.
Both countries should act before Washington moves for them. They should amend their strategic trade laws so that supplying high-end compute to a restricted end user counts as a controlled transfer, whether a chip crosses a physical border.
And they should place know-your-customer duties on operators above a compute threshold: identify the tenant, its owners and its funders, and keep usage logs that regulators can inspect.
Singapore, where many of the holding companies are registered, should require those operators to disclose their beneficial owners.
And the two neighbours should inspect jointly, because an operator headquartered in one and racked in the other currently falls between both. The alternative is a rule written in Washington, enforced on their soil, which isn’t great for a data centre pitch.
Jensen Huang built the company that made the AI boom possible. His departure fixes none of what Megaspeed exposed. The data centres are still running, the US Senate has still not voted, and the next Megaspeed may already be registered somewhere in Singapore. Until the rules catch up with how computing is actually sold, the next scandal is always brewing.
