Skip to content
United States · China · NVIDIAEpisode 058 · 18 May 2025 · 39:28

The US Decided to Sell NVIDIA to the World So the World Would Not Learn to Live Without NVIDIA

Central question

Why is the United States willing to sell NVIDIA technology globally so other countries do not learn to build compute systems without it?

What you take away

Evaluate NVIDIA and Amazon as a physical and industrial system rather than only as software. The decision requires the reader to account for chips, energy, cloud capacity, manufacturing, and the cost of scaling.

Main threads

What to watch for

1Compare “A strict ban looks like a powerful weapon until the buyer starts seeking an alternative” with “The USA bars states from regulating AI”: they provide different criteria for judging the same issue.
2Test the conclusion from “Half a trillion for data centers: Microsoft, Amazon, Oracle” in your own use case—what actually changes in the process and what remains a promise.
3Before choosing a product or approach, record the constraint identified in “The world's first digital AI goddess in a temple”.
4Define the owner of the outcome and the quality metric for the situation described in “AI in everyday life”.
Signals to track afterwards
Watch for actions by Amazon and Apple that confirm or challenge the episode’s central claims.
Compare new launches and policy changes with “Half a trillion for data centers: Microsoft, Amazon, Oracle”: have access, quality, price, or constraints changed?
Check whether the scenario in “AI in everyday life” becomes repeatable practice rather than a one-off demonstration.
Most useful for
EntrepreneursInvestorsTechnical leadersStrategy teamsEveryday usersProduct teams

Key takeaways

00:00Why an announcement is not enough: a strict ban looks like a powerful weapon until

The discussion of “A strict ban looks like a powerful weapon until the buyer starts seeking an alternative” yields a practical test: if countries cannot buy NVIDIA, they invest in Huawei and their own solutions. Several years later, the American company risks losing not only China but the entire market that has learned to operate without it.

02:45The new strategy therefore looks more pragmatic: sell American chips more broadly while controlling supply chains and restricting Huawei

The “The USA bars states from regulating AI” topic becomes clearer once this point is included: the new strategy is more pragmatic than an outright ban — sell American chips more broadly while controlling supply chains and restricting Huawei, to preserve the world's dependence on the CUDA ecosystem, clouds, and US suppliers.

06:23The boundary between value and constraint: the US strategy against China

The working conclusion from “The US strategy against China” is that the American move is not permission for “anything, anywhere” but a fight to make the next developer build on an American standard; if control is too strict or too weak, China gains the incentive and the market for a platform of its own.

09:14Microsoft, Meta, Oracle, Amazon, Google, and Apple can together invest hundreds of billions in data centers

The “Half a trillion for data centers: Microsoft, Amazon, Oracle” topic becomes clearer once this point is included: Microsoft, Meta, Oracle, Amazon, Google, and Apple can together invest hundreds of billions in data centers, and xAI is already building enormous clusters — that scale gives the US an advantage for years, but only with access to engineers, visas, and intellectual property.

12:22What changes in real work: openAI outsweighs CEO Instacart

The working conclusion from “OpenAI poaches Instacart's CEO” is that hiring strong executives is part of the same talent race: infrastructure scale gives an advantage only with access to people, so bringing in a CEO-level leader matters no less than another cluster.

15:44Why context matters more than one metric: saudi Arabia's AI perspective is a chance

The practical meaning of “Saudi Arabia's AI prospects: is there a chance?” is that the country has capital, but its chance in AI is decided by access to chips, engineers, and intellectual property — money without an ecosystem does not build a compute platform, it only buys someone else's dependence.

19:30How the issue moves from news to product: Elon Musk's new robots and the future AI economy

The “Elon Musk's new robots and the future AI economy” topic becomes clearer once this point is included: the technology is moving fast into the physical world — Musk's robots and self-driving cars in San Francisco show a real level of adoption; but the future AI economy is tested not by a demo but by a repeatable result and a clear cost.

24:20The practical meaning of the issue: aI Alive — does TikTok's new feature have a future

The working conclusion from “AI Alive: does TikTok's new feature have a future?” is that AI Alive animates photos right inside TikTok, but the feature will have a future only where it actually changes the habit of creating content rather than remaining flashy marketing.

27:26The technology is moving rapidly into the physical world

The discussion of “The world's first digital AI goddess in a temple” yields a practical test: musk’s robots, self-driving cars in San Francisco, AI Alive in TikTok, automatic audiobook narration, and even a digital goddess in a temple show different levels of adoption. In some places AI saves labor, in others it becomes a cultural object, and in still others it remains marketing.

38:10The American move against China is not permission for “anything, anywhere.” It is a fight to make the next developer build on an American standard

The boundary of the “AI in everyday life” case is defined by this point: in daily life AI already saves labor — automatic audiobook narration and similar scenarios — but elsewhere it stays a cultural object or marketing; the value shows where the technology genuinely removes manual work.

What this episode is about

A new strategy relaxes some export restrictions while increasing pressure on Huawei. The logic is simple: if allies cannot buy American chips, China will build a substitute and capture the market. With half a trillion dollars of data-center investment in the background, the race is becoming a fight over standards, infrastructure, and dependence.

A strict ban looks like a powerful weapon until the buyer starts seeking an alternative. If countries cannot buy NVIDIA, they invest in Huawei and their own solutions. Several years later, the American company risks losing not only China but the entire market that has learned to operate without it.

The new strategy therefore looks more pragmatic: sell American chips more broadly while controlling supply chains and restricting Huawei. It is an attempt to preserve the world’s dependence on the CUDA ecosystem, cloud services, and US suppliers. Sales revenue flows back into new generations of equipment.

Microsoft, Meta, Oracle, Amazon, Google, and Apple can together invest hundreds of billions in data centers. xAI is already building enormous clusters, and OpenAI is hiring strong executives. That scale gives the United States an advantage for several years, but only if it has access to engineers, visas, and intellectual property.

The technology is moving rapidly into the physical world. Musk’s robots, self-driving cars in San Francisco, AI Alive in TikTok, automatic audiobook narration, and even a digital goddess in a temple show different levels of adoption. In some places AI saves labor, in others it becomes a cultural object, and in still others it remains marketing.

The American move against China is not permission for “anything, anywhere.” It is a fight to make the next developer build on an American standard. If the strategy works, the United States will earn money and preserve influence. If control becomes too strict or too weak, China will gain both the incentive and the market for a platform of its own.

If control turns into too strict or too weak, China will gain both the incentive and the market for a platform of its own.

Episode transcript

The episode is in Russian; below is an English reading guide to the transcript (the full EN transcript is a machine translation). Voice matching applied to 101 segments: 51 identified, 0 mixed, 29 probable, and 21 unresolved.

Loading…