Nvidia and its partners built a sophisticated system to bypass U.S. restrictions on exporting AI chips to the Middle East, Russia, and China.
The sophisticated system worked as follows: any GPUs that needed to be delivered to China or to serve Chinese companies would go through Singapore as a billing location and/or through Chinese subsidiaries established in Singapore. In these subsidiaries, Chinese control shareholders were given to local Singaporean Chinese, but effective control remained with the parent Chinese company. This setup worked well for ByteDance and others.
Second, any GPUs intended for the Middle East, such as the UAE, or for Russia, would go through India. I covered one distributor that served clients in the UAE, and Bloomberg covered an Indian distributor that helped Russian companies acquire Nvidia GPUs.
Nvidia ensured that this system would operate smoothly to exceed Wall Street expectations, even if it was not fully compliant with restrictions and regulatory guidelines.
Nvidia claims it cannot control the final delivery location of its GPUs. But is that true? Singapore accounted for about 20% of Nvidia's total revenue last quarter which would raise eyes specially for the auditors - PwC San Jose. Nvidia itself has acknowledged that Singapore is merely a billing address, which implies these GPUs are reaching non-Singaporean companies.
Additionally, the law requires Nvidia to actively monitor the destination of its products and to prevent any actions that might violate laws or policies.
This raises the question: what role did Jensen Huang and/or other management play in building this sophisticated system?
