Deutsche Bank has taken a deep look at the great fear hanging over markets: whether artificial intelligence is inflating a new investment bubble.
Its conclusion, set out in a 24-page report titled Would the real AI bubble please stand up?, is that while warning signs are emerging, it is too soon to declare the boom unsustainable.
The research argues that the debate is muddled because there is not one AI bubble but potentially several, spanning valuations, investment flows and even the technology itself.
The starting point is valuations. Deutsche highlights that the Shiller price to earnings ratio for the US market has climbed above 40, nearing levels last seen at the peak of the dot com boom.
That is a red flag. Yet the bank stresses that today’s surge has been far more earnings led, with big tech delivering years of profit growth of more than 20% rather than relying on speculative projections.
Public market valuations look stretched but defensible. The froth, they argue, is mostly in private AI start-ups such as OpenAI and Anthropic, now valued at 38 to 44 times sales.
Investment trends tell a similar two-sided story. AI data centre spending could reach $4 trillion by 2030, a scale Deutsche likens to ten times the cost of the Apollo space programme.
But unlike the telecoms boom of the late 1990s, today’s capital expenditure is being funded largely from free cash flow. Hyperscalers are keeping investment comfortably below operating cashflow levels, with companies such as Google generating tens of billions of dollars each quarter.
Returns on invested capital have also been rising. All of this earns a green flag.
The biggest question is technological. The report sets out concerns about persistent flaws such as hallucinations, scaling bottlenecks and physical limits on data movement between chips.
Yet Deutsche Bank also highlights significant improvements from new models such as Google’s Gemini 3, which shows progress in reasoning, vision and multimodal performance. Demand is rising quickly, with Google now processing 1.3 quadrillion tokens a month.
Still, there are genuine risks. Deutsche Bank flags increasingly complex financing arrangements between chip designers, cloud operators and private AI labs, raising concerns about circular valuations.
Companies are issuing more debt. Public attitudes are turning more sceptical, with rising concerns over jobs, privacy and control. And the biggest constraint may be energy, with global electricity demand from AI projected to quadruple by 2030.