H200 Green Light: What Nvidia, AMD and Intel Really Gain

$NVIDIA(NVDA)$   just won political cover to ship its H200 AI accelerator to approved customers in China, with President Trump saying the United States will take a 25% cut of revenue on those chips and applying a similar framework to $Advanced Micro Devices(AMD)$   and $Intel(INTC)$   . The Commerce Department now has to turn that headline into actual licenses and compliance rules.

For investors, this is not only about reopening a giant end market. It is also about what happens to the whole system of performance based export limits that has shaped AI hardware for the last two years.

A performance red line that suddenly looks blurry

Export controls originally tried to draw a hard technical boundary. Commerce restricted China to parts like Nvidia's H20 and AMD's MI308 that sat below strict ceilings on compute and bandwidth, and lawmakers are now pushing the SAFE CHIPS Act to freeze that status quo for thirty months.

By contrast, H200 is more capable than H20 and was explicitly named by SAFE CHIPS sponsors as the kind of part that should be barred from China. Allowing H200 exports for a 25% fee effectively swaps a clear performance red line for a toll gate model. Washington gets tax like revenue and keeps Blackwell and Rubin off the table for now, but China regains access to accelerators that are much closer to the global frontier.

Politically this is far from settled. A bipartisan group of senators wants Commerce to deny any new licenses for chips more advanced than today's legal tier for 30 months, directly targeting H200 and future parts. The bill may not pass quickly, but it is a clear reminder that H200 approvals are a policy exception rather than a new stable baseline.

$NVIDIA (NVDA.US)$ : earnings optionality with visible policy risk

In Q3 FY26, Nvidia reported 57.0 billion dollars in total revenue and a record 51.2 billion dollars from its Data Center segment, up 66% year over year and 25% sequentially. Management again noted that sales of the China specific H20 chip were "insignificant" and that the current outlook assumes no meaningful China contribution.

Against that backdrop, any licensed H200 volume into China is incremental. If over the next few years China were to return to even 10%-15% of Nvidia's Data Center revenue on a future annual run rate in the 160 to 200 billion dollar range, that would imply roughly 16 to 30 billion dollars a year of additional sales versus a "no China" base case. This is scenario math, not guidance, but it explains why the stock reacts to every export headline.

The trade off is margin quality. Nvidia currently delivers non GAAP gross margins in the mid-70%s. A direct 25% skim on China H200 revenue plus higher compliance costs will make those units structurally lower margin than Blackwell racks shipped to the United States, Europe or the Middle East. The key question for investors is whether management frames China H200 as pure upside on top of its 500 billion dollar long term Data Center outlook, or partly as a way to fill any pockets of softness elsewhere.

In short, H200 access gives Nvidia a meaningful new lever for top line growth, but with political hair and thinner economics attached.

$Advanced Micro Devices (AMD.US)$ : from write downs to a chance at rebuilding China

AMD has been hit harder than Nvidia by earlier China curbs. The company has warned that tightened export controls on Instinct accelerators could shave about 1.5 billion dollars off its 2025 revenue, roughly 5% of the current Wall Street forecast. China is its second largest market: in 2024, China generated about 6.23 billion dollars of revenue and accounted for more than 24% of AMD's total sales.

On the product side, AMD's Data Center segment is now a genuine scale business. In Q3 2025, Data Center revenue reached 4.3 billion dollars, up 22% year over year, driven by EPYC CPUs and Instinct MI350 GPUs.

CFO Jean Hu has said that if China demand returns to prior levels, future quarterly revenue could rise by roughly 800 million dollars, which gives a sense of the potential uplift if licenses are granted and customers actually buy. Under the new framework, AMD is expected to secure permissions for MI300 on similar revenue sharing terms to H200, with a 25% fee.

For modeling, that means China re entry is a real but high volatility call option for AMD: the upside is meaningful relative to a 4.3 billion dollar per quarter Data Center base, but the fees, license uncertainty and competition from domestic China accelerators mean it should not be treated as a core pillar of the long term story.

$Intel (INTC.US)$ : protecting a large but shifting China franchise

Intel's China exposure is still dominated by CPUs and chipsets rather than leading AI accelerators, but the scale is large. In 2024, Intel generated 53.1 billion dollars of total revenue, and revenue from billings to China contributed 29% of that total, according to its latest 10-K.

The H200 framework gives Intel a template to seek similar licenses for its Gaudi line and future AI accelerators, again likely subject to a revenue share with Washington. If Chinese cloud providers want a non Nvidia option for some clusters and allows foreign AI silicon into those workloads, Intel could layer a modest but growing AI business on top of its existing China PC and server base.

At the same time, local Chinese CPU and accelerator vendors are steadily taking share, and China continues to push state buyers toward domestic designs. For Intel, the H200 style opening is therefore more about slowing erosion than unlocking a brand new growth engine.

The bigger picture for AI semis

Letting H200 into China at scale narrows the gap between US and Chinese AI compute and weakens the deterrent value of the original performance thresholds. SAFE CHIPS and related proposals are a direct attempt by Congress to re tighten that gap with a 30-month freeze on more advanced exports.

For Nvidia, AMD and Intel, investors should treat China re entry as upside optionality rather than a base case:

~It can add billions of dollars in annual revenue for Nvidia and AMD if licenses flow and Chinese demand materializes.

~Those revenues will be structurally lower margin because of the 15%-25% government take and heavier compliance burdens.

~The policy can still be reversed by Congress, by a future administration or by China's own procurement rules.

@TigerStars  @CaptainTiger  @TigerWire  @Daily_Discussion   @Tiger_chat  @Tiger_comments  @MillionaireTiger