Right about AI. Wrong about the investment?
Imagine being right about AI—and still losing money. More businesses use it. Customers buy more services. The technology keeps improving. Yet the shares you bought fall.
That is possible because a useful technology and a rewarding investment are different things. Your return depends on the profits a company earns, the money it spends to earn them, and the price you pay for its shares.
So, is AI a bubble? The evidence does not support one answer for every company. Some businesses may earn enough to justify investors’ hopes. Others may struggle. The practical question is how much success is already built into the price.
You can believe in AI’s future and still question the price of an AI stock.
Record profits. A falling share price.
SK Hynix makes memory used in AI infrastructure. On 29 July 2026, it reported record quarterly results: operating profit rose 557% from a year earlier, supported by strong AI demand.
Yet Reuters reported that its shares fell 9.6% that day as the results fell short of investors’ expectations. Leverage and forced selling also intensified the wider Korean rout.
The lesson travels beyond Korea: exceptional growth can still disappoint when the price assumes even more. This episode illustrates expectations risk; it does not establish that every AI stock is in a bubble or that earnings alone caused the sell-off.
The company can break records while the stock breaks investors’ expectations.
Priced for 250 mph. What happens at 200?
Our concern is overvaluation: paying so much for an AI business that the investment needs exceptional growth to continue just to justify today’s price. A company does not have to stop growing for its shares to fall sharply.
Imagine a car priced as though it will keep travelling at 250 mph. It slows to 200 mph—still extraordinarily fast. But buyers were paying for 250. The disappointment lies in the gap between what was expected and what happens next.
For a stock, that gap might be slightly slower sales growth, lower profit margins, or customers delaying orders. Investors can respond by paying less for each dollar of profit. That repricing can magnify the damage: earnings disappoint, and the price investors place on those earnings falls too.
High expectations are not a valuation measurement by themselves. AI stocks must be assessed individually; the BIS noted that large cloud-company valuations had become more moderate after the correction. Our warning concerns prices that still require too much future success, not a claim that every AI company is overpriced.
Even rising profits may not be enough. In the hypothetical example below, profit per share increases from $5 to $6. But investors reduce what they pay for each $1 of annual profit from $40 to $25. The share price falls from $200 to $150.
When a stock is priced for perfection, a slowdown can be enough. The business need not fail.
The company earns more. Do you?
Starting point: $5 profit per share × 40 = a $200 share price.
This is the price-to-earnings ratio, or P/E: the share price divided by annual profit per share.
-25.0% share price change
Profits grew, but a lower price multiple pulled the share price down.
Hypothetical illustration, not a forecast or actual investment result. Excludes dividends, fees, and taxes.
Who wins when AI gets cheaper?
Imagine AI services becoming much cheaper as providers compete. Customers could save money, even as some providers earn less on each sale. More users may offset that pressure—but that is not guaranteed.
This is why wider adoption does not automatically mean larger profits for every AI company. Some of the benefit may go to customers or new competitors.
There may also be opportunities in businesses that use AI to improve their work without paying to build the most expensive infrastructure. Their earnings and share prices still need careful assessment.
What could upset the forecast?
Even a sensible profit forecast rests on assumptions. Will customers keep buying? Will vital equipment arrive? Could a change in trade rules interrupt sales or production?
FWI examines where companies earn revenue, where essential suppliers operate, and how political or regulatory changes could affect the business. These questions can expose weaknesses that a convincing AI story leaves out.
This research cannot stop a share price from falling. It can help identify business risks before deciding what price and level of exposure make sense.
How FWEG seeks to reduce hidden AI bubble risk
Free World Efficient Growth (FWEG) combines FWI’s geopolitical research with Running Oak’s Efficient Growth investment approach. Its purpose is to pursue equity growth with discipline about the price paid and the risks taken.
For AI investors, the value of that combination is practical: examine both the valuation and the business assumptions supporting it. Four parts of the approach matter.
1. Apply a price test to the growth story
Efficient Growth seeks earnings growth while avoiding overvalued companies. Applied to AI, this means enthusiasm for the technology is not enough: the investment must also make sense at its price.
Why it matters: paying less for a stream of profits reduces the amount of future success needed to justify the investment. It does not eliminate the risk of overestimating those profits.
2. Focus on business strength
The underlying approach favours profitability and seeks to avoid unprofitable or financially fragile businesses. That brings attention back to earnings and financial resilience when the market is rewarding ambitious promises.
Why it matters: if AI spending slows, a business able to support itself is less dependent on investors continuing to fund its plans.
3. Diversify across industries—and look beneath the labels
FWEG holds companies across different industries, spreading exposure beyond a single technology theme. But different industry labels do not guarantee different risks. A chipmaker, a data-centre equipment supplier, and an electricity provider could all depend on the same AI spending boom. FWI’s research looks through companies to identify shared revenue, supply-chain, and geopolitical dependencies, complementing disciplined portfolio management.
Why it matters: industry diversification and attention to shared exposures seek to limit how much of the portfolio depends on one outcome. If AI investment slows, the aim is to avoid having too many holdings affected by the same setback. Diversification can reduce concentration risk, but cannot prevent losses in a broad sell-off.
4. Test the assumptions behind AI growth
An AI company’s forecast may assume continued access to overseas customers, advanced chips, and particular manufacturing locations. FWI’s geopolitical screening examines these dependencies. Export restrictions, tariffs, or disruption at a critical supplier can reduce sales, raise costs, or delay deliveries—making an ambitious earnings forecast harder to achieve.
Why it matters: a highly valued stock has less room for such disappointments. A geopolitical shock can hurt expected profits and prompt investors to pay less for those profits. Screening helps inform which business risks the portfolio takes, alongside valuation analysis. It does not predict an AI bubble’s timing or protect against every fall in valuations.
FWEG’s role is to make growth, price, and hidden business risks part of the same investment decision.
Stay curious. Stay disciplined.
These principles seek to manage risk while pursuing growth. They do not guarantee protection from losses. FWEG can lag when expensive stocks keep rising, its screening can exclude profitable opportunities, and its holdings can fall with the market.
You do not need to predict when—or whether—an AI bubble will burst to ask better questions today. Would the investment still make sense if adoption took longer, competition grew stronger, or customers paid less than expected?
AI may change the world. An investment still has to earn its price.
Research sources
- SK hynix — Q2 2026 results (29 July 2026)
- Reuters — SK Hynix earnings and the Korean sell-off (29 July 2026)
- Free World Investment — research and FWEG strategy
- Running Oak Capital — Efficient Growth approach
- Running Oak — Efficient Growth factsheet (August 2026)
- BIS — September 2026 market review
This material is for informational and educational purposes only. It does not constitute investment advice, a recommendation, or an offer to buy or sell any security.
Research as of September 19, 2026. Company examples do not imply FWEG holdings or recommendations. Investing involves risk, including possible loss of principal.
