AI Bubble Fears: How Exposed Is Your Portfolio?
A $20bn gap in OpenAI's revenue hit AI stocks on 8 October. How to measure your real AI bubble exposure across stocks and ETFs, with a worked euro example.
The AI bubble question got a live test on Thursday 8 October. The Financial Times reported that OpenAI's annualised revenue was close to $50 billion, about $20 billion below the figure investors had seen a month earlier. The Nasdaq fell 1.25%, chip stocks fell 3.4%, and a private company that you cannot buy on any broker moved a large part of a normal European portfolio.
By Friday most of it had bounced back. But the day showed something useful: your AI exposure is probably bigger than the tech stocks you picked yourself. This post shows how to measure it in euros, with a worked example.
What happened to AI stocks on 8 October?
The trigger was a number, not a loss. According to Euronews' summary of the FT report, the older $70 billion estimate came from OpenAI's backers. They had recounted its sales the way Anthropic counts them, which includes customer spending that runs through cloud platforms such as AWS and Google Cloud. OpenAI's own method leaves those sales out.
So the $20 billion did not disappear. It was never in OpenAI's own books. The market sold first and read the footnotes later.
Here is how far the shock travelled:
| Part of the AI chain | Company | Move after the report |
|---|---|---|
| Cloud and data centres | CoreWeave | about -8% |
| Cloud and data centres | Oracle | -5.5% |
| Chip designers | Broadcom | -4.35% |
| Chip designers | AMD | about -4% |
| Chip designers | Nvidia | almost -3% |
| Big platforms | Microsoft, Amazon, Alphabet, Meta | more than -1% each |
| OpenAI backers | SoftBank (owns about 13%) | down as much as 7.3% in Tokyo on Friday |
Source: Euronews, 9 October 2026. Moves are for Thursday 8 October unless stated.
On Friday, Bloomberg reported that OpenAI expects annualised revenue to reach $70 billion or more by the end of 2026, mostly from enterprise customers. Stocks recovered and all three main US indexes closed the week higher.
What is an AI bubble?
An AI bubble is a market phase where share prices of companies linked to artificial intelligence rise faster than the profits that AI actually produces, on the expectation that those profits will arrive later. If the profits arrive, the prices were early. If they do not, the prices fall back to what the profits support.
The hard part is that the profits sit in different places. Nvidia, TSMC and Samsung already earn record money from AI. The companies paying them, the model makers like OpenAI and Anthropic, are private and do not publish audited accounts. A large part of the AI trade rests on revenue figures that nobody outside those companies can check. Thursday showed how fast prices react when one of those figures changes.
Bubble or not: what the numbers say
Both sides have real data this week.
The case that it is not a bubble. On 5 October, the CIO of Singapore's DBS Group argued that Nvidia trades at about 17 times its expected earnings for the next 12 months, according to Bloomberg. Cisco traded near 100 times before the dot-com crash. Demand is also visible in hard numbers: TSMC said its September sales rose 54.6% year on year, and Samsung estimated a record third-quarter operating profit of 107.4 trillion won (about $80 billion).
The case for caution. The chip makers' profits depend on customers who spend more than they earn. OpenAI was valued at $852 billion in March and is reportedly raising money at $1.4 trillion, on revenue of roughly $50 billion a year. Kathleen Brooks of broker XTB said the revenue news raised questions about how long the AI spending boom can last, right before earnings season.
We do not know which side is right, and neither does anyone else. What you can know is how much of your money depends on the answer.
Why your AI exposure is bigger than you think
Most investors count their AI exposure as "the Nvidia shares I bought". The real number has three layers.
- Direct AI stocks. Shares you own in chip designers, cloud companies and AI platforms.
- The AI supply chain. Companies that are not "AI" by name but earn most of their growth from AI spending. ASML is the clearest European example. We covered the wider group in European AI stocks and the infrastructure rally.
- AI inside your index funds. The seven largest US tech companies, most of them heavy AI spenders, made up 23.6% of the MSCI World at the end of August. An S&P 500 or Nasdaq-100 fund holds even more. Our post on ETF overlap and Magnificent 7 exposure shows the full look-through method.
Layer three is the one people forget. In August, five ECB economists estimated that euro area households hold about €440 billion of US technology stocks, mostly through funds, often without knowing how concentrated that is.
How to measure your AI exposure in 5 steps
You need your positions, the top holdings of each fund, and 20 minutes.
- List every position with its current value in euros. Include every broker. A tracker like Zune.Money imports your DEGIRO transactions CSV and shows each holding with its weight, so this step takes seconds.
- Tag each single stock as AI or not AI. Use a simple rule: does more than half of its recent growth come from AI spending? Nvidia, ASML, TSMC and Broadcom are yes. A food company is no.
- Look through each ETF. Open the fund's factsheet and add up the weights of the AI-linked companies in its top holdings. Multiply that share by the euro value of the fund.
- Add it all up. Direct AI stocks plus the AI share of each fund, divided by your total portfolio value.
- Run a stress test. Multiply your AI total by 30%. It is a rough figure for a sharp correction (chip stocks as a group lost more than that in 2022), and it shows the euro loss you would see.
A worked example
Take a €50,000 portfolio. The ETF shares below are assumptions for the example. Use your own factsheets.
| Position | Value | AI share (assumed) | AI exposure |
|---|---|---|---|
| MSCI World ETF | €30,000 | 25% | €7,500 |
| Nasdaq-100 ETF | €6,000 | 50% | €3,000 |
| ASML | €4,000 | 100% | €4,000 |
| Nvidia | €3,000 | 100% | €3,000 |
| European dividend stocks | €7,000 | 0% | €0 |
| Total | €50,000 | €17,500 (35%) |
This investor owns two "AI stocks" worth €7,000, or 14% of the portfolio. Their real AI exposure is 35%, more than double. A 30% fall in AI-linked stocks would cost them about €5,250, or 10.5% of the whole portfolio, even if every other company stayed flat.
What to do with the number
None of this is advice to buy or sell. A 35% AI share can be a deliberate bet. The problem is only when it is an accident.
Pick a ceiling before the next headline, not during it. If you decide that AI-linked stocks should stay under 30%, you know what to do when the number drifts to 38%. Our guide to rebalancing a portfolio covers how to trim without paying more tax and fees than needed.
Do not sell on one day's news. Thursday's fall reversed by Friday. Investors who sold at Thursday's close locked in the drop and paid transaction costs to do it.
Add what is missing instead of cutting what is there. New money can go to parts of the market with little AI exposure, such as European value stocks, small caps or bonds. See our diversification guide for the options.
Watch the demand signal, not the private-company gossip. ASML reports on 14 October and TSMC on 15 October. Their order books tell you more about real AI spending than any leaked revenue figure. Anthropic could list as early as November and OpenAI next year, which would put audited numbers in public for the first time.
Tax rules for selling differ by country, so check with a tax advisor before you rebalance a large position.
See your AI share without the spreadsheet
The five steps work in a spreadsheet, but the numbers go stale the day after you finish. Import your DEGIRO CSV into Zune.Money and your holdings, weights and sector allocation update on their own, so you can see your tech and semiconductor share at a glance when the next AI headline hits. The free plan covers the import and the full holdings list.
Frequently asked questions
Can I invest in OpenAI from Europe?
Not directly. OpenAI is a private company, so no broker lists its shares. You get indirect exposure through listed companies that own a stake, such as Microsoft and SoftBank, which owns about 13%. Both are much bigger than their OpenAI stake, so the link is loose.
Is ASML an AI stock?
ASML makes the lithography machines that every advanced chip factory needs, so its orders follow the spending plans of chip makers like TSMC, Samsung and Intel. AI chips are now a major driver of that spending. ASML does not sell AI products, but its revenue depends on the AI build-out continuing.
What happens to an MSCI World ETF if AI stocks crash?
It falls with them, by roughly the weight AI-linked companies hold in the index. The seven largest US tech companies alone were 23.6% of the MSCI World at the end of August 2026. A 30% drop in that group would cost a world ETF about 7%, before any effect on the rest of the index.
How often should I check my AI exposure?
Once a quarter is enough for most long-term investors, plus once after every large purchase. AI stocks move fast, so the share can drift by several percentage points in a single quarter without you buying anything. Daily checks mostly add stress, not information.
