business 5 min read

Nvidia Chip Smuggling: The Public Face Versus the Real Operator

A WSJ investigation reveals that Alice Huang, the woman linked to Nvidia CEO Jensen Huang, served as only the public face of a sophisticated chip-smuggling network run from Singapore. The real architect — a 52-year-old Chinese national named Li Ming — operated behind the scenes through rotating shell companies.

  • Artificial Intelligence
  • Semiconductors
  • Export Controls
  • China-US Trade

The Face and the Architect

Alice Huang, whose romantic entanglement with Nvidia CEO Jensen Huang has drawn persistent headlines, appears in recent reporting primarily as a prop — a credential to lend legitimacy to a business that had no right to exist.

The Wall Street Journal reported Monday that she served as the public face of MegSpeed, a Singapore-registered company that held Nvidia cloud-partner status and allegedly funneled restricted American AI chips into China. Behind her, according to multiple sources, stood Li Ming, a 52-year-old Chinese national based in Singapore, who ran the actual operations.

This distinction matters more than it might first appear. The difference between the figurehead and the operator is not just a biographical detail. It is the operating model of an entire evasion ecosystem — one that has adapted faster than any enforcement regime has tracked it.

How the Network Worked

Li Ming, now released on bail and living in a villa near the Chinese embassy in Singapore, is charged with fraud. According to Singapore prosecutors, he deceived US suppliers into believing his local firms would use Nvidia servers domestically. Instead, the servers moved through Malaysia — a well-known transit corridor for redirected electronics — and disappeared into a third country that has not been publicly identified. China is widely understood to be the end market.

MegSpeed, originally a Singapore subsidiary of a Chinese video-game company, acquired preferred-purchaser status through Nvidia’s cloud program in 2023. That designation gave it priority access to advanced chips at a time when the Biden administration was tightening export controls on precisely those products.

Alibaba-owned Century Liantong, a Shanghai technology firm, appears multiple times in the filing as a downstream recipient of those diverted servers. Two Singapore nationals were listed as officers of another entity connected to Century Liantong’s Singapore arm — and that entity’s director was separately indicted in Singapore on charges similar to Li Ming’s.

When Nvidia flagged MegSpeed for internal review in June 2024, it placed the company on a “no sell” list. Alice Huang resigned two months later. MegSpeed effectively ceased operations in Singapore.

But the network did not dissolve. By last year, it had incorporated a new vehicle called Novagate Cloud, again staffed with Singapore-based personnel who hold positions at firms tied to the same Century Liantong affiliate. The rotation continues.

What This Reveals About Enforcement Gaps

The most striking feature of this case is not that smuggling occurred — restricted chips reaching China through third countries has been documented repeatedly since 2022. The striking feature is the structural sophistication of the response.

A single raid or indictment disrupts one entity. Rotating to a new company, with familiar principals and a freshly minted partnership application, is far harder to prevent. The network absorbed a “no sell” designation and a fraud indictment without stopping its core activity. That level of resilience suggests either deep institutional knowledge of how to navigate export-control compliance or significant patience backed by serious capital.

Singapore plays a structural role here. It is not a direct transit point like Malaysia but rather a clearance hub — a jurisdiction where US-designated cloud partners can obtain legitimacy, file paperwork, and build relationships with American suppliers. The geographic proximity to Malaysia and Thailand creates a short physical loop. The legal system provides enough procedural friction to delay seizures but not enough to block purchases outright.

The result is a gap: US export controls regulate outbound sales from America, but they have limited reach once those goods enter a supply chain governed by a different legal jurisdiction.

Who Wins and Who Loses

Li Ming wins, apparently. He remains free on bail, housed in an upscale residence near the Chinese diplomatic mission, and his enterprise continues operating through successive corporate shells. If convicted, the penalties remain to be seen — but the cost of doing business under these conditions is low relative to the revenue flowing from supplying China’s AI data centers.

Nvidia loses visibility and, increasingly, credibility. The company knew MegSpeed was an NVIDIA cloud partner and sold chips accordingly. It acted when alerted, placing the firm on its internal restriction list. But the fact that a company tied to the CEO’s personal life obtained that designation in the first place raises uncomfortable questions about due diligence that extend beyond one client.

China’s AI sector gains what it cannot buy legally. The very chips Washington restricts — H20, A800 variants, and earlier models still deemed viable for training large language models — move through circuits that American regulators struggle to trace past Singapore.

The United States loses policy credibility. Export controls lose force when every restriction generates a new bypass mechanism within months. The pattern across multiple enforcement actions — Hong Kong, the Netherlands, Singapore — is consistent: controls are reactive rather than anticipatory.

What Comes Next

Three developments are likely. First, the Biden-era framework will face pressure to expand not just the list of controlled chips but the scope of secondary sanctions targeting non-US facilitators. Companies in third countries that knowingly enable diversion could face consequences for their US-market access. That is the logical extension of the Treasury Department’s existing guidance, though no broad rule change has been announced.

Second, Nvidia and other American chipmakers will tighten cloud-partner vetting. The MegSpeed case — a gaming company subsidiary that somehow qualified for priority chip access — should never have reached that stage. Expect longer approval timelines, deeper beneficial-ownership disclosure requirements, and potentially mandatory downstream audits for partners sourcing restricted components.

Third, China will accelerate domestic alternatives. Every chip that enters China through these channels buys time but also increases the incentive to build indigenous capability. Huawei’s昇腾 series and SMIC’s foundry expansions are already progressing; the smuggling network exists precisely because demand outstrips legal supply.

The deeper lesson, however, applies to everyone who sells into or buys out of the American semiconductor ecosystem. Enforcing export controls requires tracking not just shipments but corporate formations, partnership applications, and directorship records across multiple jurisdictions. No single agency — not Commerce, not Justice, not Interpol — currently has that capacity. The network around Li Ming understands this. Its ability to reconstitute itself so quickly is evidence of systematic planning, not improvisation.

The headline characters may shift. Jensen Huang’s name will draw scrutiny. But the architecture that let him operate at arm’s length from the actual supply chain was designed by someone else entirely.