The AI Compute Race Glossary H20

H20

What does “H20” mean?

H20

A downgraded NVIDIA Hopper chip, in the Institute for Progress’s words “designed specifically to comply with export control thresholds,” which went on sale in China in 2024. Washington required a license for it in April 2025. The H200 is almost six times as powerful. Chapter 7.

From The AI Compute Race. Read the book

Where the book uses it

  • Chapter 7 — The Double Blockade

    Two governments deciding one trade: Washington’s export rules and Beijing’s instructions at the border, and what each did to the chip business it meant to shape.

Sources that use it

  • Cyberspace Administration of China, notice of 31 July 2025 summoning NVIDIA over alleged tracking, positioning and remote-shutdown risks in H20 chips (the CAC’s own published notice, citing the Cybersecurity Law, the Data Security Law and the Personal Information Protection Law). In the Sources
  • Donald Trump, “The President’s News Conference,” 11 August 2025, transcript, The American Presidency Project (asked Jensen Huang for 20 percent; Huang asked for 15; payment in return for a “release” limited to H20; on Blackwell: “Although, it’s possible I’d make a deal—a somewhat enhanced, in a negative way, Blackwell. In other words, take 30 percent to 50 percent off of it” and “Now, on the Blackwell, I think he’s coming to see me again about that”). In the Sources
  • NVIDIA Corporation, Form 10-K for the fiscal year ended 25 January 2026, filed 25 February 2026 (“We generated approximately $60 million in H20 revenue under those licenses”; “USG officials expressed an expectation that the USG will receive 15% or more of the revenue generated from licensed sales of our products, but the USG did not publish a regulation codifying such requirement”; a February 2026 license “to ship small amounts of H200 products to specific China-based customers” under which, “to date, we have not generated any revenue”; Note 2, the Groq transaction: “In December 2025, we entered into a non-exclusive license agreement with Groq, Inc., or Groq, for its language processing unit technology and hired certain Groq employees. No customer contracts, existing products, or equity interests were purchased”; “Total consideration consists of $13.0 billion paid at closing and $4 billion, inclusive of imputed interest, payable within one year”; “$14.4 billion of goodwill and a $2.5 billion developed technology intangible asset”, the goodwill “primarily attributable to the workforce and future development of the licensed technology” (accession 0001045810-26-000021, p. 61); the $2.5 billion developed technology intangible asset “valued using a cost-to-recreate methodology with a five-year useful life”; the goodwill “recorded in the Compute & Networking reporting unit” and “tax deductible”; the risk factors call the transaction “an intellectual property license arrangement with Groq, Inc., or Groq, that required significant, nonrefundable payments”; the risk factors, p. 28: “an agency of the Chinese government announced an Action Plan that endorses new standards regarding the compute performance per watt and per memory bandwidth of accelerators used in new and renovated data centers in China”, and should US export controls change “to allow us to return to the market”, the Chinese government “could modify or implement the Action Plan in a way that effectively prevents us from being able to design products to meet the new standard”; Government Regulations: “Over the past three years, we have been subject to a series of shifting and expanding export control restrictions”; the restrictions announced in August 2022 “specifically impact our A100 and H100 integrated circuits”, the requirements of October 2023 reach products “exceeding certain performance thresholds, including, but not limited to, the A100, A800, H100, H800”, “In January 2025, the USG published the AI Diffusion IFR in the Federal Register”, which “would have divided the world into three tiers”, and “In May 2025, the USG announced that it would rescind the AI Diffusion IFR”; of OpenAI: “We are finalizing an investment and partnership agreement with OpenAI. There is no assurance that we will enter into an investment and partnership agreement with OpenAI or that a transaction will be completed.”; under “Facility Lease Guarantees”: “In fiscal year 2026, we entered into agreements to guarantee partners’ facility lease obligations in the event of their default in exchange for warrants. The maximum gross exposure under all agreements is $3.5 billion, which is reduced as the partners make payments to the lessors over terms ranging from 5 to 7 years. The partners have placed $712 million in escrow to mitigate our potential exposure”, no partner named; revenue of $215,938 million for fiscal 2026, and “As of the end of fiscal year 2026, we had approximately 42,000 employees in 38 countries”; “In August 2022, the U.S. government, or USG, announced export restrictions and export licensing requirements targeting China’s semiconductor and supercomputing industries”, reaching exports “to China (including Hong Kong and Macau) and Russia” that “specifically impact our A100 and H100 integrated circuits”; “In October 2023, the USG announced new and updated licensing requirements for exports to China and Country Groups D:1, D:4, and D:5 (including but not limited to Saudi Arabia, the United Arab Emirates, and Vietnam, but excluding Israel)”; read in the held text of the filing). In the Sources
  • NVIDIA, Form 10-Q for the quarter ended 27 April 2025, filed 28 May 2025, the issuer’s own copy (a $4.5 billion charge in the first quarter of fiscal 2026 for H20 excess inventory and purchase obligations, $1.9 billion of it an inventory provision for H20 product inventory and $2.6 billion an excess inventory purchase-obligation charge for H20 product orders; “we incurred a $4.5 billion charge in the first quarter of fiscal year 2026 associated with H20 excess inventory and purchase obligations as the demand for H20 products diminished”; “The $4.5 billion charge was less than what we initially anticipated as we were able to re-use certain materials”). In the Sources
  • NVIDIA, Form 8-K, 15 April 2025 (US government notice of 9 April that H20 exports to China required a license; 14 April that the requirement would remain indefinitely; estimated charge up to $5.5 billion). In the Sources

And 7 more: search the Sources.

Related terms

From the glossary of The AI Compute Race.

How to cite this page

Dryburgh, Lee S. “H20.” In The AI Compute Race. DRYBURGH. https://dryburgh.com/vol1/glossary/h20 (accessed 2026-11-01).

BibTeX

@misc{dryburgh-h20-2026,
  author = {Dryburgh, Lee S.},
  title = {{H20}},
  howpublished = {In \emph{The AI Compute Race}, DRYBURGH},
  year = {2026},
  url = {https://dryburgh.com/vol1/glossary/h20},
  note = {Accessed 2026-11-01}
}

For a reference manager: RIS · CSL-JSON