Nvidia budgets zero China data centre sales into its $108 billion quarter

The guided step from $96.2 billion to $108 billion is $11.8 billion of new revenue in one quarter, or 12.3% sequentially ^[3][1][10]. Cryptobriefing.com puts Chinese AI chip demand at around $50 billion a year, a figure it attributes to unnamed analysts ^[19]. That is $12.5 billion a quarter ^[11]. So one quarter of guided growth from everyone else is worth about one quarter of the market Nvidia has been shut out of ^[10][11].

Run the old share against the current base. China was 13% to 20% of data centre revenue before the controls bit, and data centre revenue is now $89 billion a quarter, 92.5% of the top line ^[5][4][12]. That share applied to today's base would be $11.6 billion to $17.8 billion a quarter, $46 billion to $71 billion a year ^[15]. The top of that range sits $21 billion above the estimate for the whole Chinese AI chip market ^[16]. Either the $50 billion number is stale, or China's share would not have scaled with the data centre build-out.

The two China figures do not fit together. Take $89 billion of data centre revenue out of $96.2 billion and $7.2 billion of gaming, automotive and everything else is left ^[13]. China across all product categories came to about $7.88 billion ^[7]. For both to hold, essentially every non-data-centre dollar Nvidia booked anywhere in the world would have to have been Chinese, plus $680 million of data centre sales, which is 0.76% of $89 billion and squeaks in below the 1% share reported for the quarter ^[14][5].

The zero rests on a sales record. Nvidia shipped a small batch of H200s to roughly 10 approved Chinese firms and saw minimal uptake, with Beijing's purchasing restrictions layered on top of the US export controls ^[6]. Earlier rounds of US rules had already closed off the downgraded parts Nvidia designed for that market ^[9]. According to cryptobriefing.com, Huang has publicly framed the loss as temporary while the financial plan treats the revenue as gone until proven otherwise ^[8].

Cryptobriefing.com also argues that the current valuation reflects a world of essentially zero Chinese sales, so any policy shift becomes pure upside ^[17]. That is a claim about price, and the account does not include a share price or a multiple to test it against. The same piece grants the decay: Huawei's Ascend chips are gaining traction while sitting generations behind Nvidia's architecture in most benchmarks, and the longer controls run, the smaller the addressable market becomes even if they lift ^[21][22].

I'd price the China channel as an option with a $50 billion-a-year ceiling that loses value each quarter it goes unexercised, and a Trump-Xi summit is the event that would exercise it ^[19][20]. The case against: if Nvidia books material China data centre revenue in the October quarter and still clears $108 billion, the zero was conservatism and the channel was live the whole time ^[1].