Whilst the US market still looks strong on the surface, but the leadership has had some dramatic shifts. The S&P 500 has held up well in 2026 even as many large technology and software stocks have fallen sharply from their highs, a change that may mark the start of a longer term market rotation.
The Big Picture
Whilst the S&P 500 is still up almost 10% in 2026, 30 of its members had already fallen more than 30%, showing that the market’s gains have been far narrower than they first appear.
That matters because many investors think they are diversified when in practice they are heavily tied to a small number of mega-cap technology names. The world’s most valuable technology stocks have lost large amounts of value over the last few months as investors question whether very heavy AI spending will generate returns quickly enough to justify the valuations.
And this is not just a US story; South Korea’s tech-heavy market has also suffered violent pullbacks as chip stocks rolled over.
Recent highs versus now
| Company | 2026 high | Close on 31 July | Fall from high |
|---|---|---|---|
| Tesla | c.$490.00 | $311.21 | −36.5% |
| Salesforce | $274.00 | $184.02 | −32.8% |
| Meta Platforms | c.$735.00 | $556.71 | −24.3% |
| Alphabet | c.$403.00 | $356.13 | −11.6% |
Why have we seen this market reaction
There is a number of reasons, valuations, worries about rates and a growing debate over whether the AI boom had run too far too fast.
1. The AI bill has become enormous
The largest technology companies are spending hundreds of billions of dollars on data centres, power, networking equipment, cloud capacity and AI models and applications.
Amazon, Alphabet, Microsoft and Meta have all announced substantial investment programmes. Earlier in 2026, analysts estimated that the major US technology groups could collectively spend more than $500 billion during the year.
This spending may eventually create significant value. But shareholders are asking a straightforward question:
How much additional profit will each dollar of AI investment generate?
2. Expectations had become extremely high
A company’s share price reflects what investors expect to happen next.
If a business is valued on the assumption that profits will grow exceptionally quickly, merely reporting good growth may not be enough. The company may need to produce outstanding results every quarter.
That is why a strong company can still be a disappointing investment when it is bought at too high a price.
3. Software faces its own AI threat
Investors previously assumed software companies would be major winners from AI. That may still prove correct, but the picture is more complicated.
AI could help software businesses create better products and charge customers more. It could also make it easier for competitors to develop rival software, automate coding and reproduce existing services.
Companies such as Salesforce, Microsoft, Adobe, Accenture and other IT service providers are therefore being judged in two ways:
- Can they make money from AI?
- Could AI undermine parts of their existing business?
That uncertainty helps explain why some software shares have fallen much more heavily than the wider market.
Microsoft: when strong results are not enough
Microsoft is the best company to bring this story to life. A fall in the share price from roughly $483 to $389 is about 20%,.
The key point is that Microsoft did not fall because the business is struggling. Microsoft expected strong cloud growth and said its AI revenue run rate had surpassed $37 billion, but investors were also confronted with a 2026 capital spending plan of around $190 billion, well above analyst expectations.
That is where valuation comes in. If a company is already priced as a near-perfect winner, then even strong results may not be enough if investors fear cash is going out faster than clear returns are coming in. Microsoft reported $35.8 billion in operating cash flow in one fiscal quarter while recording $37.5 billion of capital expenditures including finance leases, a striking figure because it shows how intense the infrastructure build-out has become.
It was also reported in January that Microsoft fell after only modest growth versus the market’s elevated hopes, and that investors worried about the scale of spending and the concentration of backlog linked to OpenAI. That is a very useful point for the article: when expectations are extreme, “good” can still disappoint.
The AI race is not one race
A helpful way to explain the shift is to say the AI boom has layers. Jensen Huang has often described the AI stack in layers, from chips and systems to cloud infrastructure, models, platforms and the applications built on top; the investment case is that different winners can emerge at each layer, and the profit pool may shift over time as the stack matures.
That fits neatly with the view that the winners of today may not be the winners of tomorrow, because leadership can move as the market shifts from building the infrastructure to developing the software and services that sit on top. The practical point for investors is that owning “AI” can be done in many different ways.
For now, the market appears to be asking a sharper question at every layer of the stack: who is earning the return, who is just spending, and who may be over-owned? That is why the rotation matters. It is not only about whether AI grows, but about where the economics finally settle.
It is happening outside the US too
The reassessment of AI shares has spread well beyond America.
South Korea’s KOSPI has fallen by around 40% from its late-June peak, including declines of almost 11% on 28 July and a further 6% on 29 July. The retreat has centred on the country’s two memory-chip giants: Samsung Electronics is approximately 45% below its recent peak, while SK Hynix has fallen around 56%.
This is not because demand for AI memory has suddenly disappeared. SK Hynix reported a sixfold increase in quarterly profit, while Samsung’s semiconductor profits surged, but results failed to match the exceptionally high expectations already reflected in share prices. Investors are also worried about rising capital expenditure, increasing Chinese competition and whether today’s extraordinary memory-chip margins can be sustained.
The decline was intensified by borrowed money and leveraged single-company ETFs, which forced further selling as prices fell. Samsung and SK Hynix together represent more than half of the KOSPI’s value, showing how a country index can become a highly concentrated technology bet. Even after the sell-off, however, the Korean market remains substantially higher in 2026, another reminder that this is a sharp reversal after an extraordinary rally, rather than evidence that the AI investment story has disappeared.
Action Plan
- Check what is actually inside your portfolio. Many investors own technology exposure several times over through global funds, US trackers, dedicated tech funds and individual names, without realising how concentrated the overlap has become.
- Separate “index exposure” from “mega-cap exposure.” An S&P 500 tracker can still leave a portfolio heavily influenced by a handful of giant technology shares, so it is worth comparing that with an equal-weighted approach that spreads risk more broadly across the index.
- Review software and cloud holdings with fresh eyes. In this market, the key question is no longer just who has an AI story; it is who can convert spending into revenue growth, margins and free cash flow.
A final thought
The AI story is far from over. The real debate is not whether AI matters, but whether today’s spending, valuations and leadership positions are sustainable at the prices investors recently paid.
A healthy dose of realism.
