Gen AI bubble watch… is it deflating?
AI stocks recently had their worst day in over a year, and American voters have turned against the data centres that power the boom. Is the AI bubble finally losing air - and what does that mean for agencies?
Gen AI bubble watch… is it deflating?
AI stocks recently had their worst day in over a year, and American voters have turned against the data centres that power the boom. Is the AI bubble finally losing air - and what does that mean for agencies?
By Conor Nichols, Special Reporter
Who would you rather live next to: a nuclear power plant or an AI data centre? According to Gallup, 71% of Americans would oppose an AI data centre in their area - well above the 53% who’d oppose a nuclear plant.
You’ve probably seen the headlines.
Here are the numbers behind them.
56%
of US voters told an NBC exit poll that AI and data centres were a ‘crucial’ or ‘major’ factor in how they voted in the midterms
10
states backed restrictions on new data centres
5.5%
the tech-heavy Nasdaq’s fall on 5 November, its worst day since April 2025
9%
Nvidia’s fall on the same day.
$1.7 trillion
wiped off Nvidia’s value between its record high on 6 October and 11 November, the day before Jensen Huang resigned as chief executive amid an investigation into his links to an alleged Chinese spy.
$135-145 billion
what Meta, which makes almost all its money from advertising, expects to spend on capital investment this year, much of it on AI
Put together, these show three pressures building up at once. Investors want proof that the hundreds of billions going into AI will pay off. Voters are pushing back against the physical cost of building it. And the whole trade leans heavily on a handful of companies - in Nvidia’s case, on one man. Days before Jensen Huang resigned, one Bernstein analyst said he “IS Nvidia”. Now its finance chief, Colette Kress, is running the company on an interim basis while the board looks for a replacement.
So, how do we actually know if AI investment is getting out of hand? One simple gauge is the Nasdaq 100 P/E ratio.
The price-to-earnings (P/E) ratio measures a company’s share price relative to its earnings. Aggregated across an index like the Nasdaq 100, it shows how much investors are willing to pay for a dollar of earnings across the market. A high P/E suggests strong optimism about future growth - part fact, part sentiment. But when it climbs too far, that optimism can tip into irrational exuberance.
| Total | |
|---|---|
| Dotcom peak 2000 | 73 |
| End 2023 | 28 |
| End 2024 | 32 |
| January 2026 | 37 |
| March 2026 | 34 |
| November 2026 | 10 |
Here is the historical context: the Nasdaq 100 typically sits in a P/E range of 21-32.
During the Dotcom bubble,
it peaked at 73
| Total | |
|---|---|
| Dotcom peak 2000 | 73 |
| End 2023 | 28 |
| End 2024 | 32 |
| January 2026 | 37 |
| March 2026 | 34 |
| November 2026 | 10 |
At the end of 2023,
it was around 28
| Total | |
|---|---|
| Dotcom peak 2000 | 73 |
| End 2023 | 28 |
| End 2024 | 32 |
| January 2026 | 37 |
| March 2026 | 34 |
| November 2026 | 10 |
rising to 32 by the
end of 2024
| Female | Total | |
|---|---|---|
| Dotcom peak 2000 | 24 | 73 |
| End 2023 | 28 | 28 |
| End 2024 | 32 | 32 |
| January 2026 | 0 | 37 |
| March 2026 | 37 | 34 |
| November 2026 | 0 | 10 |
In January 2026 it stood at 37
| Female | Total | |
|---|---|---|
| Dotcom peak 2000 | 24 | 73 |
| End 2023 | 28 | 28 |
| End 2024 | 32 | 32 |
| January 2026 | 0 | 37 |
| March 2026 | 37 | 34 |
| November 2026 | 0 | 10 |
and by March it had cooled to 34
| Female | Total | |
|---|---|---|
| Dotcom peak 2000 | 24 | 73 |
| End 2023 | 28 | 28 |
| End 2024 | 32 | 32 |
| January 2026 | 0 | 37 |
| March 2026 | 37 | 34 |
| November 2026 | 0 | 26.7 |
Today, as of 13 November 2026 it sits at 26.7
Nvidia's first results since Huang's exit, on 17 November, showed revenue more than doubling year on year to $115.8 billion - a reminder that a falling P/E isn't only about falling share prices. Earnings are catching up too.
For advertising professionals, this is not just financial trivia. Much of agency planning right now - pitches, pricing, hiring, production budgets - quietly assumes AI will keep getting cheaper and better. The market is now testing that assumption.
The cost of AI tools could rise. Many AI firms still lose money, and investors are covering the gap. If that money dries up, prices go up. (See our piece on token pricing)
Margins could get squeezed. If agencies have already passed AI savings on to clients, any price rise comes out of their own pocket - especially on fixed-fee deals.
Production costs could climb. AI video uses huge amounts of computing power. If that gets scarcer, today’s rates may not last.
Platforms will want their money back. Meta and Google fund much of their AI spending from advertising - so expect a harder push on automated buying such as Advantage+ and Performance Max.
Staffing plans and supplier choices built on cheap AI could need a rethink too.
So where does that leave us? The Nasdaq 100 sits at a P/E of 26.7 - the champagne has lost some of its fizz, but nobody has called time on the party yet. Whichever way it goes, the lesson for agencies is the same: plan as if AI might get more expensive, not just cheaper.
Even the shiniest bubble eventually meets a pin. This one might need planning permission first.
