On April 17, a report from Morgan Stanley made Nvidia once again the eye of the storm in global capital markets.

In its latest research report, the Wall Street investment bank lowered Nvidia's revenue forecast for the next few quarters. The core logic points to the "unexpected destructive power" of the U.S. government's chip export restrictions to China-especially the H20 series chips. Sales orders may cause Nvidia's data center business revenue to drop by 8%-9%, and related inventory impairment and purchase commitments losses of up to US$5.5 billion.

After the news was announced, Nvidia's share price plunged nearly 7% in a single day, triggering a collective plunge in the U.S. stock chip sector.

Judging from the evolutionary path of technical control, U.S. chip export restrictions to China have evolved from an initial "precision strike" to systematic containment.

As a "special version" chip specially designed by NVIDIA for the China market, the H20 is based on the previous generation Hopper architecture, but relying on the balance of performance and cost, it will still contribute US$12 billion to US$15 billion in revenue in 2024, accounting for more than 70% of its revenue in the China market.

Now the U.S. government has extended the export threshold from the calculation index to comprehensive parameters such as memory bandwidth and interconnection bandwidth, which means that even such "downgraded version" products need licenses to ship, essentially blocking the operating space for enterprises to evade control through technical parameter adjustment.

This precision in policy design reflects the regulatory authorities 'deep understanding of the path of technology spillover-just as the new BIS rules expand the scope of restrictions from China to the world for the first time, trying to build a three-tier control system of "Tier1 allies-Tier3 high-risk countries" is precisely to cut off the possibility of China re-export through third countries.

The penetration of geopolitical risks into business decisions is particularly evident in Nvidia's financial statements-a potential loss of $5.5 billion includes not only inventory impairment, but also liquidated damages on customer procurement commitments. Supply chain disturbances of this magnitude are extremely rare in the semiconductor industry.

It is worth noting that despite the short-term pain, most Wall Street institutions still maintain their optimistic ratings on Nvidia.

Morgan Stanley interprets the target price of US$162 as "long-term value retention under short-term pressure", and investment banks such as Bank of America and Jefferies have given potential room for growth of 53%-78%.Behind this seemingly contradictory logic is actually capital's determination of the underlying logic of AI computing power demand-the construction of global data centers is moving from the "kcal era" to the "ten thousand card cluster". Meta alone plans to purchase AI chips worth tens of billions of dollars within the year, and NVIDIA's technological moat built by the CUDA ecosystem will still be difficult to shake in the short term.