The Average Semiconductor Stock Is Down 25% in Six Weeks While the Dow Hits Records. Which One Is Wrong?

Two things are true at the same time, and they are hard to reconcile.

The Dow Jones Industrial Average closed at an all-time high of 53,178.41 on 3 August 2026, gaining 693 points. The S&P 500 ended at 7,600.50 and the Nasdaq at 25,913.9.

And the average semiconductor stock is now in a bear market — down roughly 25% on average in just six weeks.

The July sector data makes the split explicit: information technology fell 8.0% in July while energy gained 12.1%, financials rose 6.2%, health care 2.5%, and consumer staples 2.4%. The S&P 500 edged down 0.13% and the Nasdaq Composite lost 3.20%, while the Dow added 0.32% and the S&P/TSX gained 1.06%.

That is not a market selling off. That is a market rotating, hard, out of one thing and into everything else.

Unwind versus break — and why the distinction is everything

There are two competing explanations, and they lead to opposite actions.

The technical unwind view: piling into AI stocks became a crowded, momentum-driven trade. When too many people hold the same position with the same thesis, an ordinary disappointment forces disorderly exits because everyone reaches for the door simultaneously. Under this reading, the fundamentals are intact, positioning was the problem, and the selloff is a buying opportunity.

The fundamental shift view: the market has begun questioning whether AI capex converts into profit on the timeline embedded in chip valuations. Under this reading, the de-rating is rational and early, and you should reduce rather than add.

The supporting evidence is genuinely mixed. Arguing for "unwind": Korea's memory giants delivered a record single-day rebound after Microsoft's results reinforced expectations for continued AI spending, and Samsung has said the global memory shortage is expected to deepen next year. Arguing for "break": SK Hynix's most profitable quarter in its history still missed consensus, and Nvidia now needs to mobilise over $500 billion of third-party capital to keep its customers buying — which is a statement about how much capital the buildout requires from outside the tech sector.

The three-headed headwind

Underneath the AI debate, three macro variables have all moved the wrong way for equities at once: interest rates, oil prices and the US dollar are all rising.

That combination is a specific problem for long-duration growth stocks. Higher rates compress the present value of distant earnings. Higher oil raises input and transport costs and keeps inflation sticky. A stronger dollar reduces the translated value of overseas revenue, which for semiconductor companies is most of it.

None of that is about AI at all. It is about the discount rate — and chip stocks, which are priced on earnings several years out, are the most discount-rate sensitive part of the market.

What an ordinary investor should take from this

If you own a broad index fund, you already own this. We wrote about how index funds are quietly changing beneath their holders as the market rotates. You do not need to act; you need to know what you own.

Don't treat a 25% drawdown as a signal in either direction. Chip stocks are structurally volatile. Semiconductors are cyclical — that is not a bug that AI fixed, it is the nature of an industry with long capacity lead times and lumpy demand. Drawdowns of this magnitude have happened repeatedly during genuine secular growth phases.

Notice what rotation is doing for you. If your portfolio is broadly diversified, energy up 12.1% and financials up 6.2% has been quietly offsetting technology down 8.0%. That is diversification working exactly as intended, in real time, in a month most people would describe as bad. What "rotation" means covers the mechanic.

Be sceptical of anyone certain right now. We were bullish on the AI rally in February and published an honest reassessment when chips cratered. The truthful position today is that the unwind and break interpretations are both live, and the data over the next two quarters — specifically, whether AI capex converts to disclosed revenue — will settle it.