Chipmaker Nvidia reported Wednesday that it brought in $96 billion in revenue during the second quarter of its fiscal year — more than double the figure from a year earlier — and guided for $108 billion in the following quarter. CEO Jensen Huang declared that “AI has reached its inflection point” and described the infrastructure buildout as proceeding “at full steam.” The data centre division alone generated $89 billion last quarter, up 117% year on year, as every major technology company building AI tools — Amazon, Meta, Google, Microsoft — relies on Nvidia’s hardware to do so. The results beat Wall Street forecasts; shares rose about 4% in after-hours trading. Nvidia’s market capitalisation now stands above $5 trillion, making it the world’s most valuable firm. The company has also begun investing in the companies that depend on its chips, providing funding to OpenAI, Anthropic, and SpaceX. Analysts noted that the guidance for the next quarter points to “revenue comfortably above $110 billion.”
The received wisdom
The optimistic framing of Nvidia’s results is that they validate the largest and most consequential technological buildout since the internet: AI infrastructure that will, in time, produce productivity gains, medical breakthroughs, scientific discoveries, and a general expansion of human capability. Jensen Huang’s language about an “inflection point” reflects a genuine consensus view among technologists that the last two years have seen AI cross from impressive laboratory demonstration to pervasive deployment. The argument runs that Nvidia’s extraordinary financial performance is, in this sense, a leading indicator of civilisational-level investment — that the billions flowing into data centres are rational bets on a future of enormous economic value. Progressive critics of Big Tech tend to focus on concentration and labour displacement, while AI boosters in both parties treat the buildout as an unqualified good. On the numbers, this is a hard case to dismiss: $96 billion in quarterly revenue does not happen by accident, and the customer list — essentially every significant technology company on earth — is not a speculative cohort.
A different read
The history of transformative technology booms is, however, also the history of infrastructure concentration, rent extraction, and strategic vulnerability that only became visible in retrospect. Standard Oil looked like a marvel of industrial efficiency before it looked like a monopoly. Microsoft looked like an unstoppable engine of the information economy before the DOJ noticed that its control of the operating system was a bottleneck that stifled competition for a decade. The analogy to Nvidia is imperfect — Nvidia competes in a market where Amazon, Google, and others are actively designing their own chips — but the underlying dynamic is instructive.
Matt Britzman at Hargreaves Lansdown called this “another monster set of results,” and financially it is. But consider the strategic exposure embedded in those numbers. With around 40% of the US stock market concentrated in ten companies “heavily invested in AI,” as the BBC noted, a significant share of American household retirement savings is now a bet on the continued performance of a single technology cycle built on a single firm’s chips. This is not a new observation — it was made about the FAANG cohort a decade ago — but the concentration is now considerably more extreme, and the underlying technology is considerably less mature.
There is also a geopolitical dimension. The Taiwan AI server smuggling case reported earlier this week — in which Nvidia and Super Micro employees were charged with helping to export 74 restricted AI servers to China through third-country intermediaries — illustrated that export controls on Nvidia’s most advanced chips are porous in practice, even as they are formally tightened. Nvidia’s commercial success depends on building and selling the most powerful processors available; US national security policy depends on those processors not reaching adversaries who will use them to close the AI capability gap. These objectives are in structural tension, and no amount of export compliance programming eliminates it.
Moreover, Nvidia’s growing practice of investing in the companies that depend on its chips — OpenAI, Anthropic, SpaceX — creates an unusual supplier-customer-investor relationship that has few precedents in the history of technology manufacturing. If Nvidia is funding OpenAI’s infrastructure costs while selling OpenAI the chips to run on that infrastructure while also holding equity in OpenAI, the standard checks of arms-length commercial relationships have broken down in a way that makes it difficult for regulators or even sophisticated investors to assess where Nvidia’s value actually resides. This is not an allegation of wrongdoing — it is an observation about structural complexity that compounds as the company’s tentacles extend.
None of this changes the immediate reality: the demand for AI compute is real, Nvidia’s technology advantage is real, and the quarterly results are not a fiction. But the history of technology booms suggests that the moment of maximum exuberance — “AI has reached its inflection point,” says the man whose company profits most from that inflection — is also the moment to ask what the dependency costs are, who is positioned to extract them, and whether the political class overseeing this transition has the technical literacy to know what questions to ask. Congress barely understood cloud computing when it was already a decade old. AI is moving faster, and the stakes are higher.
What to watch
- Antitrust scrutiny: whether the DOJ or FTC treats Nvidia’s simultaneous roles as chip supplier, infrastructure investor, and equity holder in AI labs as a basis for structural investigation.
- Competition from custom silicon: whether Amazon’s Trainium, Google’s TPUs, or Chinese domestic alternatives (where export controls have accelerated development) begin to erode Nvidia’s data-centre dominance, and on what timeline.
- The export-controls enforcement gap: whether the Taiwan smuggling case prompts a tightening of the intermediary-country oversight regime, or whether it simply illustrates the limits of what export controls can achieve against determined circumvention.
- Stock market concentration: whether the Federal Reserve or Treasury begins monitoring systemic risk in the AI equity concentration, given its exposure to a single technology cycle and a single dominant supplier.
— J