H2 opens with the AI trade losing its monopoly

The AI trade losing its monopoly

The S&P 500 is still near the highs. The Dow is making fresh ones. From thirty thousand feet, the market looks broadly fine.

But beneath the index level, H2 has opened with a fairly serious change in leadership.

The AI, memory and semiconductor trades that carried much of the market higher are being marked down hard, while financials, consumer discretionary and some of the prior laggards are finding a bid. The market is not selling everything. It is changing seats.

That distinction matters.

This is not primarily a payrolls story for equities. The weaker jobs data mattered for bonds, FX, gold and Bitcoin. The rotation in stocks came from a different question altogether: whether the AI build-out can keep running at the same pace without eventually running into its own capacity.

Reports that Meta may be looking to sell excess AI compute brought that question closer to the surface. No one is saying the AI capex cycle is over. Hyperscalers are still spending heavily, data-centre demand remains strong and AI infrastructure remains one of the better-supported earnings stories in the market.

But the trade had been built around a very clean assumption. Compute would stay scarce. Capacity would always be absorbed. Every new memory chip, accelerator and server rack would find a buyer.

Once investors begin to wonder whether supply is arriving faster than demand can absorb it, the maths changes.

The long-term AI story does not need to collapse for the shares to struggle. It only needs to become less perfect than the market had priced.

That was the pressure point.

Momentum fell more than 18% over two days, its worst two-day decline since November 2020. A closely watched AI beneficiaries-versus-AI-at-risk basket dropped roughly 16%, its worst move in the cited data. Memory stocks fell more than 18%, their sharpest two-day decline in at least 12 years, while AI semiconductors suffered their worst two-day move since Liberation Day.

This was not a casual pullback in a few overheated names. The move had the familiar shape of a crowded trade being forced to shrink.

The winners fell sharply. The shorts rallied. Realized volatility jumped. That is when the market stops debating the long-term thesis and starts dealing with position size, VaR limits and where the next liquidity is likely to be.

The memory complex made the point visible.

SanDisk fell roughly 22% from its recent high, breaking its steep trendline and 21-day moving average. Kioxia weakened through similar short-term support. Micron broke below its trendline and 21-day average. SK Hynix, one of the clearest Asian expressions of the AI boom, also lost momentum through comparable levels. Samsung then followed by breaking its short-term trendline and 50-day moving average.

When one leader weakens, the market can call it company-specific.

When the leaders and laggards begin to crack together, the market starts questioning the pricing structure of the whole trade.

That is the more important point here. Investors are not suddenly concluding that AI demand has vanished. They are questioning whether the shares had already discounted too much uninterrupted scarcity, too much uninterrupted hyperscaler spending and too little chance of capacity catching up.

The reported possibility that Apple may consider Chinese memory suppliers for products sold domestically adds another layer. The immediate attraction is clear enough: lower costs and more supply flexibility. The longer-term issue is whether that sort of validation gradually improves the standing of domestic Chinese suppliers in the world’s largest electronics market.

That is not necessarily tomorrow’s earnings problem. But it is the kind of competitive shift investors begin to care about once the market stops treating the incumbents as untouchable.

The reason the S&P 500 has held together is that money has not left the market. It has rotated.

Financials and consumer discretionary outperformed. The Dow held firm. The Nasdaq found some support around its 50-day moving average. Small caps remained softer, while technology, energy and utilities lagged.

The headline index looked orderly because the selling was concentrated. The market was not collapsing. It was redistributing risk away from the most crowded part of the room.

That makes the next few weeks more important than the last few days.

Second-quarter earnings expectations are still demanding. Bottoms-up consensus is looking for roughly 22% year-on-year S&P 500 earnings growth. AI infrastructure remains one of the strongest areas of that outlook, but the burden of proof has gone up.

The easy trade was owning the story.

The next trade is working out which companies can still deliver the orders, margins and spending momentum needed to justify the valuations.

That is particularly relevant because the S&P 500 has been stuck in a range since mid-May. The market has largely priced in the end of the US-Iran war and is now waiting for earnings to provide the next real catalyst. In a range-bound market, broad beta becomes less useful. Stock selection matters more.

While equities were reassessing AI scarcity, macro markets were trading a different message.

June payrolls rose by just 57,000, well below expectations, while prior months were revised lower. The unemployment rate fell to 4.2%, but the detail was less reassuring than the headline suggested. The labour force shrank by roughly 720,000, with around 700,000 of that decline coming from workers aged 25 to 34.

Unemployment fell because fewer people were looking for work, not because hiring suddenly accelerated.

That does not amount to a nascent recession signal on its own. Manufacturing is still expanding, with the ISM index at 53.3, and the 2Q GDP tracker remains around 2.1%. But it does suggest that the labour market is becoming less forgiving around the edges.

That was enough for macro markets to adjust.

Two-year Treasury yields fell around 4bp to 4.14%. The dollar softened. Traders trimmed expectations that the Fed would need to raise rates again in the near term. Gold found a bid and Bitcoin posted its strongest two-day run since February.

Warsh’s softer tone on inflation risks had already taken some heat out of the tightening debate. The payroll report gave markets another reason to lean in that direction. The message was not that inflation has been beaten. It was that the economy may be less able to absorb another tightening cycle without a clearer cost to jobs and households.

Three numbers capture the wider backdrop: 57,000 payrolls, 53.3 for ISM manufacturing and $70 Brent.

That is less a classic K-shaped economy than an apples-and-oranges one.

The apples are the enterprises. Corporate investment remains firm. Businesses are still spending on AI, automation and productivity, and may be increasingly willing to use those tools to lift output without adding headcount at the old pace.

The oranges are workers and consumers. Consumer confidence has been revised lower. Labour-force participation has weakened. The jobs data suggest that finding work may be getting harder even while the stock market remains close to record levels.

One side of the economy is still buying servers and software. The other is becoming more cautious about income and jobs.

Oil fits the same broader picture.

Brent has fallen toward $70 from around $112 in mid-May, while WTI has dropped more than 40% from its war-era peak. The reopening of the Strait of Hormuz is bringing Gulf supply back faster than many expected.

Flows through Hormuz reportedly reached around 14 million barrels a day on July 1. Add Saudi Red Sea exports and UAE barrels moving through Fujairah, and the market is again looking at a significant volume of crude that needs to be absorbed. Saudi flows have returned toward 90% of pre-war levels, while emergency reserve releases and weaker Chinese imports remain part of the backdrop.

The war premium has come out of oil. The market is now looking at whether demand can absorb the returning barrels without a more visible surplus.

The front of the crude curve has shifted into its widest contango since December 2022, which is a more bearish signal than the outright price move alone. Lower oil helps the inflation story at the margin and gives bond bulls another reason to lean toward a less hawkish Fed path. But it also raises questions about demand at a time when the labour market is losing some of its earlier resilience.

Bitcoin’s bid alongside the Korean equity unwind fits the same broader rotation. It would be too neat to say one caused the other, but the timing is notable. As leveraged equity exposure in Asia was being reduced, bonds, bullion and crypto all found buyers.

Capital was not leaving the market. It was becoming far more selective about where it wanted to sit.

That is the real H2 opening message.

The broad index remains resilient. Manufacturing is still expanding. AI spending has not disappeared. But the market is no longer willing to pay for every part of the story in the same way.

The leadership trade has become less comfortable. The labour market has softened enough to change the Fed debate. Oil has moved from war scarcity toward a looser supply outlook. And the AI complex is heading into earnings season with much less room for disappointment.

The next phase will be decided by delivery, not narrative.