The maths that should keep you awake
OpenAI is currently valued at US$500 billion. It has roughly US$13 billion in annual revenue and US$1.5 trillion in infrastructure commitments ahead of it. It is losing money at one of the fastest paces in Silicon Valley history. And it is not alone.
Nvidia briefly touched US$5 trillion in market capitalisation. The five largest AI-investing tech companies plan to spend US$3 trillion over five years on AI infrastructure, yet their combined annual revenue, excluding Amazon’s retail arm, is only US$1 trillion. Three times annual revenue committed to capex is not a strategy. It is a bet that the revenue will materialise to justify the spending after the fact.
Fisher Funds’ Harry Smith put it directly: “the amount of money that the market expects to be generated over the next five years is quite huge and that’s where a lot of the bubble chat is really coming from.” Michael Burry, who correctly called the 2008 financial crisis, has taken short positions against Nvidia and Palantir. The Federal Reserve, the IMF, and the Bank of England have all flagged concern about overblown AI stock prices.
The circular financing structures make it worse. The US$30 billion deal between Microsoft, Nvidia, and Anthropic is the kind of cosy related-party arrangement that has historically characterised bubbles. Revenue flowing between co-investors is not the same as revenue flowing from customers.
Spending mistaken for investment
University of Auckland Professor Rod McNaughton has drawn the dot-com parallel carefully. In 2000, success was measured in eyeballs and clicks. Today the metrics are “tokens processed” and “model queries” but the belief that scale leads to profit has not changed. His verdict is blunt: “Then, as now, spending is mistaken for investment.”
The critical difference from 2000 is that today’s AI surge is led by Microsoft, Google, and Amazon rather than fragile startups. That makes a sudden catastrophic collapse less likely. But it does not make the valuations rational, and it does not protect the broader ecosystem of AI-branded companies riding the wave. McNaughton argues the real question is “whether AI’s real productivity gains will ultimately justify today’s valuations, as the internet, after a painful correction, eventually did.”
Monash University Professor Chris Marsden went further at a University of Auckland conference, warning there is a real risk the AI investment bubble will burst in 2026.
Why New Zealand cannot shrug this off
The Spinoff’s analysis described AI as a “meta-bubble” combining attributes of real estate, technology, loose credit, and government backstop. Critically, as much as half of all US economic growth has been driven by the AI boom. A correction would ripple through US consumer confidence, capital markets, and trade flows, all of which hit a small open economy that has already endured two recessions.
The fiscal foundation underneath the AI story is itself cracking. America’s national debt is projected to reach US$64 trillion within a decade, with interest payments hitting US$2 trillion annually by 2036. As CFO Tech noted, the entire AI growth story sits on top of an increasingly fragile fiscal base. Market strategist Greg Boland of Moomoo ANZ added that NZ’s own inflation surprise of 3.1% against an expected 2.7% complicates the Reserve Bank’s room to respond if a correction hits.
For a local illustration, consider Being AI Limited on the NZX. Revenue barely moved at $40.993 million in FY25, but the net loss before tax blew out from $372,000 to $11.98 million. A capital raise attracted subscriptions for just 6% of shares offered. Rebranding around a hot technology narrative, failing to convert it into earnings, then failing to raise capital when the story stops working is a compressed version of what bubble corrections look like at scale.
Distinguish earnings from narrative
None of this means AI is fake. Government projections suggest generative AI could add $76 billion to the NZ economy by 2038, though that figure comes from Microsoft-commissioned research and should be read accordingly. The adoption data tells its own story: 67% of larger NZ businesses now use some form of AI, but 68% of SMEs have no plans to evaluate or invest in AI, compared to just 38% in Australia.
The AI Corner offered the most useful framework: a correction “will clear out weak unit economics. Durable cash flows and defensible systems will survive. Story stocks will not.” For NZ business owners, the practical task is not deciding whether AI is real but learning to tell the difference between a vendor selling productivity and a vendor selling narrative. The first category will survive a correction. The second will not, and it will take your contract spend with it.
Sources
- The Spinoff: The impact on NZ if the AI bubble bursts (2025-12-11)
- Newsroom: Fears of bubble trouble with AI shares (2025-11-12)
- interest.co.nz: The AI boom feels eerily similar to 2000’s dotcom crash
- NZ Herald: The AI boom feels eerily similar to 2000’s dotcom crash, with some important differences
- RNZ: Risk the AI investment bubble will burst this year, Australian professor says
- CFO Tech: Inflation, US debt and AI – key risks for NZ investors
- The AI Corner: How New Zealand can build value in the AI bubble