The market is cooling under the semis.

The market is strong, but not for broad reasons

The most important question this week is not simply whether the market is going up. It is why the market is going up. That answer is more nuanced than the index chart suggests.

If someone only looked at the major indices, they could reasonably conclude that investors are still trading a soft landing. Earnings are better than feared, the labor market has not rolled over, and the AI capex cycle remains one of the cleanest growth stories available. That is all true. But the market underneath is telling a more selective story. Semiconductors have become the release valve for investors who still want growth exposure but do not want to own everything tied to the old economic cycle.

J.P. Morgan notes that semiconductors are up 24% year to date and have outperformed hyperscalers by 19 percentage points, while semiconductor EPS has grown far faster than hyperscaler EPS as AI spending flows down the supply chain.¹ That does not make the rally fake. It makes the rally specific. Investors are not buying every cyclical business because they think the economy is reaccelerating. They are buying the parts of the market where demand is still being driven by a structural investment cycle rather than by the next consumer spending print.

That is why the “market keeps pumping because of semis” explanation is directionally right, but incomplete. Semis are not the only reason the market is holding up, but they are the clearest expression of the market's current preference. Investors want businesses tied to AI infrastructure, durable earnings, and capital spending that does not disappear just because GDP slows from “good” to “fine.” The market is cooling underneath, but the AI complex is still hot enough to keep the index from looking as tired as the macro data feel.

The risk is that this setup can look healthier than it is. A narrow rally can last longer than skeptics expect because the winners keep compounding. But it also leaves the market more vulnerable to two things: higher yields and disappointment in the AI capex narrative. If Friday's jobs report comes in too hot, the rates market can pressure long-duration growth. If it comes in too weak, investors may start asking whether AI leadership is strong enough to offset a broader earnings slowdown.

The macro data are cooling, not collapsing

This week's data were not bad. They were just not clean enough to let investors relax.

The ISM Services PMI came in at 53.6 for April, down slightly from March but still in expansion for the 22nd consecutive month.⁴ Business activity rose to 55.9, which says the service economy is still moving, but new orders fell to 53.5 from 60.6, suggesting some of March's strength may have been pulled forward ahead of higher prices.⁴

The price line was the uncomfortable part. The ISM Services Prices Index held at 70.7, matching March and remaining at its highest level since October 2022, with diesel, gasoline, oil, and related costs among the most frequently cited sources of price pressure.⁴ That is not the combination the Fed wants. Growth is slowing, but input costs are not giving policymakers a clean disinflation signal.

Labor data told a similar story. JOLTS showed job openings unchanged at 6.9 million in March, hires rising to 5.6 million, quits little changed at 3.2 million, and layoffs little changed at 1.9 million.⁵ That is not a labor market in free fall. It is a labor market where employers are becoming more cautious, workers have less leverage than they did during the hottest part of the cycle, and the balance is shifting slowly rather than suddenly.

Weekly claims also refused to validate a recession narrative. Initial claims rose to 200,000 in the week ending May 2, but the four-week average fell to 203,250, and continuing claims declined to 1.766 million for the week ending April 25.⁶ That matters because claims are usually where a true labor-market break starts to show up. For now, the economy looks slower-hiring, not firing-cycle.

The best summary is that the economy is moderating, not cracking. That is still a constructive setup for equities if inflation keeps moving lower. The issue is that inflation is not cooperating fast enough. The market wants a softer labor market because it wants lower yields. The Fed wants a softer labor market only if it comes with lower wage pressure and cleaner services inflation. Those are not the same thing.

What we are looking for in Friday's jobs report

The market wants Goldilocks, but Goldilocks has a narrower definition now.

The ideal payroll report would show job growth close to consensus, unemployment steady at 4.3%, wage growth cooling but not collapsing, participation stable, and revisions that do not quietly rewrite the last two months lower. In plain English, the market wants evidence that employers are slowing hiring because they are cautious, not because demand is falling apart.

The headline payroll number matters, but the revisions may matter more. March payrolls rose 178,000 after February was revised to a 133,000 loss, while January and February combined were revised 7,000 lower in the last report.³ If April lands near consensus but prior months are revised sharply lower, the market may treat the report as weaker than the headline suggests.

The wage line is the second hinge. ADP's April report showed private employment rising by 109,000, with annual pay up 4.4% for job-stayers and 6.6% for job-changers.⁷ That is not alarming by itself, but it keeps the Fed focused on whether wage gains are cooling enough to reduce services inflation pressure.

The sector mix will matter too. Health care carried a large share of March job growth, while federal government employment continued to decline and financial activities softened.³ A report supported only by defensive hiring and government noise would not have the same quality as broader private-sector resilience.

The best case is not a boom. A blowout number would likely push yields higher and make the Fed's higher-for-longer stance easier to defend. The best case is a slow, boring, internally consistent report that lets investors say: growth is cooling, inflation pressure is easing, and earnings are not about to fall off a cliff.

What Wall Street might not be thinking about yet

The consensus debate is still too focused on whether Friday's number is good or bad. The better question is what kind of good or bad the market can actually absorb.

Issue one: a soft print is not automatically bullish. A modest miss with stable unemployment, cooler wages, and clean revisions probably helps the soft-landing narrative. A weak print with rising unemployment, lower hours, and ugly revisions becomes an earnings risk before it becomes a Fed-cut story.

Issue two: semiconductors are carrying more than just the tech trade. They are carrying the market's belief that a structural AI cycle can overpower cyclical cooling. That belief is reasonable, but it is not risk-free. Goldman Sachs estimates that consensus 2026 hyperscaler capex has moved up to 527 billion dollars, while Fortune reports that Alphabet, Amazon, Meta, and Microsoft are projected to surpass 700 billion dollars of combined 2026 capex.⁸⁹ Those numbers support the thesis, but they also raise the bar. Once the market prices in a durable boom, “still spending” may not be enough. Investors will eventually ask who earns the return on that spending.

Issue three: the AI trade is moving from chips into power. Wood Mackenzie estimates the U.S. data center electrical equipment market can grow from 20 billion dollars to 65 billion dollars by 2030, while U.S. data center capacity is expected to rise from roughly 24 GW in 2026 to 110 GW in 2030.¹⁰ That turns AI into a grid, equipment, transformer, switchgear, cooling, and utility-interconnection story. The market may still be underestimating how much of the next leg of AI spending happens outside the most obvious semiconductor names.

Issue four: Fed governance is still a real market variable. Powell is expected to remain on the Board after his chair term ends, Warsh is expected to face confirmation the week of May 11, and the most recent FOMC vote included four dissents, the highest count since October 1992.¹¹ If sticky services prices meet a politically pressured Fed transition, the long end of the curve may start pricing credibility risk before the front end prices a dramatic change in policy.

Investment thesis

The investment thesis is that this is still a constructive market, but it is no longer an easy market. The index can grind higher if the labor market cools in an orderly way and earnings estimates remain intact. But the leadership profile argues for selectivity rather than broad risk-taking.

Our view is that broad market beta is less compelling than exposure to companies with three traits. First, durable earnings that do not require a perfect macro backdrop. Second, balance-sheet flexibility, because higher-for-longer rates punish businesses that need cheap capital to defend growth. Third, structural demand drivers that are not fully dependent on the next quarter of GDP.

AI infrastructure still fits that framework, especially semiconductors, memory, networking, custom silicon, and power-related beneficiaries. MU remains one of the cleaner ways to express the memory and HBM part of the AI trade. NVDA remains the compute anchor. AMD remains the earnings-sensitive challenger. AVGO remains important because custom silicon and connectivity are becoming more central as hyperscalers try to improve cost and performance across the stack.

The second-order AI power trade may be even more interesting for investors who do not want to chase the most crowded front-end names. Electrical equipment, power distribution, backup systems, cooling, utility interconnection, and data-center construction all sit behind the same capex cycle but may have different valuation and positioning dynamics.

Cybersecurity and defense also deserve attention. Both themes benefit from a world where geopolitical risk, infrastructure vulnerability, and government spending visibility remain elevated. They may not have the same upside as the hottest AI names, but they can offer cleaner earnings durability in a market where macro visibility is getting worse.

Quality healthcare and defensive growth belong on the watchlist as well. If Friday shifts the market from inflation risk toward growth risk, cash flow, pricing power, product-cycle durability, and balance-sheet quality should matter more than pure momentum.

Macro-level options ideas for this setup

These are not recommendations, but they are reasonable structures traders often consider when the market is waiting on labor data, rates volatility, and concentrated leadership.

A long straddle or strangle on SPY or QQQ is the cleanest event-volatility expression if payrolls can trigger a larger-than-priced move in either direction. The risk is paying too much premium into a known catalyst and then getting hit by post-event volatility compression.

A put spread on QQQ or another duration-sensitive growth vehicle makes sense if the larger risk is a hot payroll number or sticky wage data. The point is that higher yields can pressure long-duration equities even if the AI earnings story remains intact.

A call spread on semiconductors or high-quality AI infrastructure makes more sense if the report lands in the Goldilocks zone: soft enough to contain yields, but not so weak that recession risk takes over.

For investors who want a less directional expression, the relative-value version may be cleaner: own quality AI infrastructure, cybersecurity, defense, or durable healthcare against weaker cyclicals and companies that need a rapid decline in rates.

Practical takeaway

The market is not pumping because everything is fine. It is holding up because semiconductors, AI infrastructure, and quality earnings are offsetting a cooling economy. That is a bullish story, but a selective one.

Friday's jobs report matters because it will tell us whether the slowdown is still orderly. A boring report keeps the soft-landing narrative alive. A hot report creates a rates problem. A very weak report creates an earnings problem. The worst mix would be weak growth with sticky wages, because that would leave investors with slower demand and a Fed that still cannot comfortably ease.

The better opportunity is not to chase the whole market blindly. It is to own the parts of the market where structural demand, earnings durability, and balance-sheet strength still matter: AI infrastructure, semiconductors, memory, power, cybersecurity, defense, and selective defensives.

The market is waiting for Friday morning. The better investors are already thinking about what the data will mean by Monday.