July Market Commentary | The AI Trade Cracks as the Market Broadens

Nicola Wealth Market View

July was a month of rotation beneath the surface. While the headline index was virtually flat (the S&P 500 total return in U.S. dollars was -0.1% in July), the broader market (the S&P 500 Equal Weighted Index was +1.0%) strengthened, even as technology and momentum (the tech-heavy NASDAQ was –3.2%) weakened.

The trade that has dominated equity markets for the better part of two years, companies geared to artificial intelligence spending, came under significant pressure. The Philadelphia Semiconductor Index fell nearly 21%, its worst month since October 2008, while the equal-weight S&P 500 ended the month just below its June 2 all-time high.

This combination deserves attention. A market in which the leaders fall, and the average stock rises is rotating rather than breaking down. Breadth improved through the month, credit spreads remained reasonably tight, and defensive sectors did not take over leadership. Corporate earnings, which ultimately drive returns, continued to come in ahead of expectations. Taken together, these factors provide a supportive backdrop for markets.

The more important question may sit in the bond market. Long-term U.S. yields moved higher again, with the 30-year Treasury spending more days above 5% this year than in any year since 2007. The Federal Reserve held its policy rate steady, though three officials dissented in favour of an increase. Chair Kevin Warsh made clear he is content to let markets do some of the tightening for the Fed. Higher long-term yields raise the cost of capital and reduce the value investors place on distant profits, which is precisely the environment in which the most expensive parts of the market get tested.

In our view, July is not a signal that the cycle is ending. Instead, it serves as a reminder that concentration cuts both ways. Portfolios heavily exposed to the leaders of the past two years may carry more risk than their recent returns suggest. The same dynamic might be at play in the oil market, with crude oil prices potentially underpricing the risk that the Strait of Hormuz remains closed longer than the market anticipates, further depleting global reserves.

Key Takeaways

• The AI and semiconductor trade unwound sharply, with the Philadelphia Semiconductor Index down 21% for its worst month since October 2008
• Market breadth improved even as the headline index stalled, with 73% of S&P 500 companies trading above their 200-day average
• Earnings remain the market’s strongest support, with 2026 S&P 500 earnings growth expected at 28.5%
• Long-term bond yields, rather than the Federal Reserve, did the tightening in July
• The Fed held rates steady for a fifth consecutive meeting, though three officials dissented in favour of a hike
• Canadian inflation cooled materially, while a new U.S. tariff threat reopened trade risk

Chief Economist Rob Edel works through what drove the rotation, why long-term yields did more tightening than the Fed, and what both mean for portfolio construction. Read the full commentary below.

Momentum unwound sharply, and the leaders took the damage

The Philadelphia Semiconductor Index (SOX), which tracks 30 of the world’s largest chipmakers, fell 21% in July, its worst month since October 2008. According to Bloomberg, the decline erased US$2.2 trillion of market value. Taiwan Semiconductor fell 15%, removing more than US$380 billion. Micron Technology fell 29%, its steepest monthly decline in more than a decade, and Intel fell 35%, its largest monthly drop since September 2000. Nvidia and Broadcom rose over the month, a useful reminder that this was not a uniform sell-off.

The volatility was as notable as the direction. On nearly half of July’s sessions, the SOX index closed up or down by at least 4%. All 22 sessions saw intraday swings of at least 2%, something last seen in 2020. Even after an 8.3% two-day rally to finish the month, the index ended 23% below the record high it set on June 22.

This was a factor event as much as a sector event. Scotiabank reports that its US top-momentum screen underperformed its lowest-momentum counterpart by 19% in July, a margin last seen in April 2020 and before that April 2009. Both of those earlier episodes came as markets rebounded from a trough. This one occurred near a high.

For long-term investors, the lesson is less about semiconductors than about portfolio construction and concentration risk. A move of this magnitude is severe for a concentrated portfolio and more manageable within a diversified one.

The charts show the increase in semiconductor stock volatility in July and how narrow the damage was in dollar terms and how uneven it was across companies, with a handful of large names accounting for most of the decline while others gained.

Beneath the index, the market broadened

While the cap-weighted S&P 500 churned, the equal-weight version of the index traded to a fresh high. 73% of S&P 500 constituents finished the month above their 200-day moving average, up from 56% at the cap-weighted index’s peak and the highest reading in nearly two years. Small companies told the same story, with 76% of the S&P 600 Small-Cap Index constituents above their 200-day averages. The share of S&P 500 companies reaching a 20-day high expanded to 38%, the best reading since the initial recovery from the late-March and early-April lows.

Leadership rotated toward health care, financials, and, notably, smaller consumer discretionary companies. Semiconductors, by contrast, ended the month with no constituent trading above its 50-day average. Strategas reads the pattern as rotational rather than distributive, noting that real risk usually announces itself through leadership and credit. Both were supportive through July.

Bloomberg made a similar observation in mid-July, describing a market that had broadened beneath the surface of the headline indices. The dispersion is striking: a basket of companies geared to AI spending, including semiconductors, trailed a basket of companies seen as vulnerable to AI disruption by a record 42 percentage points during the month.

Performance by size helps illustrate the shift in leadership as well. The Magnificent 7 plus Broadcom were down 1.1% on average year to date at month-end, while the remaining 92 largest companies were up 11.6% and the next several hundred were up between 6.8% and 17.5%.

This chart matters because it separates a rotation from a decline. The index stalled, but participation improved, which historically looks more like leadership changing hands than like a market topping out.

The questions investors are now asking about AI spending

Apollo framed the shift well late in the month, reducing the debate to three questions. Will AI capital spending generate attractive returns, and how quickly? How is that spending being financed, and at what spread? And will demand for computing power prove effectively unlimited, or will it peak?

The answers are genuinely uncertain, and several sources made that case in July. The Economist argued that AI revenues are growing quickly but not quickly enough, with adoption lagging the scale of investment. The Wall Street Journal noted that analyst models for the largest technology companies now assume cost efficiencies that look difficult to achieve. The Financial Times reported that Big Tech AI spending has passed US1$ trillion, and that credit risk across the sector has risen sharply as that spending is funded. Bloomberg puts outstanding future spending commitments at roughly US$2 trillion.

At the same time, the spending itself has not slowed. Amazon and Microsoft both reaffirmed substantial AI investment plans in late-month results, which Bloomberg noted helped ease some of the concern. The competitive picture is also changing: China’s Kimi K3 model demonstrated rapid progress at materially lower cost, which sharpens rather than settles the question of returns on infrastructure spending.

Positioning likely amplified the volatility. Bloomberg reported in mid-July that U.S. corporate insiders were selling shares at close to a record pace, among the fastest in 20 years. On the other side, Bloomberg Intelligence recorded US$12 billion of net inflows into semiconductor exchange-traded funds in a single week, roughly a quarter of all ETF inflows over five days from a group representing about 1% of ETF assets.

We do not believe the AI investment cycle has ended. We do think the market has moved from paying for spending to asking about returns on that spending. AI revenue is growing, but is it growing fast enough?

 Earnings remain the market’s strongest support

For all the volatility, the fundamental earnings backdrop continued to improve. According to FactSet data, S&P 500 earnings are forecast to grow 28.5% in 2026 and 14% in 2027. In mid-July, Strategas noted that expected earnings growth over the next twelve months, at 31.1%, ranked third highest on record, and that it is occurring in the middle of an economic cycle rather than in a recovery from a recession. Bloomberg reported second-quarter estimates as high as 29% and broadening beyond the largest technology companies. Barron’s was more cautious, noting that while second-quarter earnings are set to rise more than 27%, the result still depends heavily on technology profits.

Still, the breadth of that strength remains encouraging, with seven of the 11 economic sectors delivering double-digit earnings growth in the first quarter, and all eleven reporting higher revenue. Deutsche Bank’s Binky Chadha projects profits for the rest of the S&P 500, excluding technology and AI, to grow 14.3%.

There remain valid reasons to be cautious, however. Operating margins are at record levels, with the estimate for the next 12 months at 21.0%. Such strong growth also creates difficult comparisons next year. Higher energy costs and the coming wave of depreciation tied to the AI build-out could also make further margin expansion harder to sustain. Valuation offers no cushion either: the market trades at roughly 25 times trailing earnings against a long-term average of 16 since 1950, and 19 times forward earnings.

The chart shows both why the market has held up and where the risk sits. Earnings estimates have kept rising, and the gap between the largest technology companies and everything else is expected to close.

Long-term yields did the tightening

The most consequential move in July happened in long-dated bonds. The 30-year Treasury yield traded above 5% on 27 days in 2026, including 12 consecutive sessions in July. That is the highest number of days and the longest run since 2007, when it spent 50 days above the 5% level.

The long bond closed July at 5.27%, its highest level since June 2007. The 30-year inflation-adjusted, or real, yield has risen roughly 50 basis points this year toward 3%, an area it last traded in 2008. Notably, the Fed’s benchmark rate is about 150 basis points lower than it was in 2007, so investors are demanding more compensation for holding long-dated Treasuries now than they were at the start of the financial crisis.

Two forces appear to be driving the move. The first is fiscal. Since 2007, the Treasury market has grown to US$31 trillion from US$4.5 trillion, debt has passed 100% of gross domestic product, and annual interest costs have risen above USD $1 trillion. Strategas expects a federal deficit of roughly US$1.9 trillion this year, and notes interest costs absorbed 19.5% of tax revenues in June. Fitch Ratings recently warned that the U.S. debt burden sits far above that of other countries sharing its AA rating. Hoisington Investment Management, a long-standing bull on long-term bonds, reversed its position in July, citing a structural backdrop of larger deficits and higher capital demand.

The second is increasing competition for the same buyers. More than US$500 billion of AI-linked financing is now coming to the debt market, which means governments and hyperscalers are bidding for the same pool of long-duration capital.

Importantly, this is a real-rate story rather than an inflation-expectations story. Rosenberg Research argued in late July that the increase in yields reflects a higher real risk premium tied to uncertainty about Fed policy, not AI-related issuance. Inflation breakeven rates, a market-based measure of expected inflation, have been range-bound for roughly five years. Still, it’s hard to dismiss the role future inflation may play in yields, and it’s likely the market is starting to discount the scenario of inflation and rates remaining higher for longer.

As we noted last month, both Morgan Stanley and Strategas have identified 4.5% on the 10-year level above which equity valuations have historically compressed. Given that U.S. 10-year yields ended the month at 4.74%, we would expect any move higher in yields to be met with continued investor angst and market volatility.

This chart puts the move in historical context. Long-term borrowing costs above 5% are no longer a brief episode to be bought quickly, and that matters for the value of every long-dated asset, including equities.

The Fed held, and three officials dissented

On July 29, the Federal Open Market Committee voted nine to three to hold the federal funds rate in a range of 3.5% to 3.75%, its fifth consecutive hold. Dallas Fed President Lorie Logan, Cleveland’s Beth Hammack, and Minneapolis’s Neel Kashkari dissented in favour of a quarter-point increase. The statement was otherwise identical to June’s, again describing activity as expanding at a solid pace, noting strong capital investment and productivity growth, and characterizing inflation as elevated relative to the 2% target.

Chair Warsh was direct on the inflation objective, telling reporters, “There is no soft inflation target.” Asked why the Fed had not moved, he pointed to the rise in market rates since the June meeting, suggesting investors are doing some of the central bank’s work. He attributed that in part to his own decision to pare back the guidance the Fed normally offers. Two-year Treasury yields fell after the decision. In the days beforehand, futures markets had put the odds of a hike as high as 40%.

The shape of the curve and the message it is sending to investors remain the question. Société Générale observed in late July that the two-year to 10-year curve remains positive even with two hikes priced in, implying that a larger hawkish surprise would be required to invert the yield curve. Bear steepening (a steeper curve caused by long rates rising more than short rates) typically signals rising growth, while bear flattening (flatter curve caused by short rates rising more than long rates) signals rising growth and inflation. July delivered both, ending with a bear steepener after the hold and the absence of guidance.

Views on whether the Fed will ultimately move remain divided. Alpine Macro argues it will not, because current inflation is concentrated on the supply side while monetary policy works mainly on demand. Goldman Sachs suggests rate increases could offset some supply-shock inflation by reducing resource utilization and by influencing expectations. It concludes, however, that the drivers monetary policy influences most effectively are not the problem, pointing instead to tariffs, the Iran war, and the mismeasurement of AI effects. Strategas notes that the market is pricing one to two increases by year-end, and that a hike has become its base case.

We suspect the reduction in forward guidance from Chair Warsh is itself part of the story. When a central bank offers less certainty about the path of rates, investors price a wider range of outcomes into long-dated bonds. Warsh believes past guidance has encouraged increased risk-taking and leverage, and that the risk premium priced by the market has become too tight. The market reacted accordingly and sold longer-term Treasuries, thus driving yields and risk premiums higher. In effect, Warsh’s lack of guidance is itself a form of tightening.

Inflation cooled in June, though the relief may prove temporary

The June inflation report, released mid-July, was better than expected. Consumer prices rose 3.5% from a year earlier, down from 4.2% in May and below the 3.8% economists surveyed by the Wall Street Journal had expected. Core prices, which exclude food and energy, rose 2.6%. On a monthly basis, core prices were broadly flat. Shelter costs posted their smallest monthly increase since 2021, and prices fell outright for apparel, used cars, car insurance, and medical care. Producer prices also rose less than expected. Market-implied odds of a July rate increase dropped from close to 40% to roughly 15% after the release.

Two details temper the good news. Car insurance, wireless data services, and hotel prices together pulled the monthly rate down meaningfully and are unlikely to repeat. The price index for computer software and accessories rose more than 17% over the year, the largest gain on record, as AI adoption spreads through households and businesses.

More significantly, June benefited from a temporary reprieve in gasoline prices that has since begun to reverse. The benchmark U.S. oil price was already 12% higher in July at the time the report was published. The Federal Reserve’s preferred inflation measure, published by the Commerce Department, has accelerated in recent months and remains well above the 2% target.

Oil: a fraying ceasefire and a thinner cushion

The June memorandum of understanding between the U.S. and Iran unraveled during July. Brent crude climbed past $100 a barrel following the re-escalation, then eased again late in the month, trading near $90 on July 31. Goldman Sachs has said Brent could exceed $120 a barrel by the fourth quarter if disruptions to the Strait of Hormuz persist.

Bloomberg counts supply disruptions on four separate fronts: the conflict in the Middle East, Ukrainian strikes against Russian infrastructure, the closure of the Strait of Hormuz, and the shipping attacks in the Bab el-Mandeb.

Global supply buffers appear thinner than they were, though the degree of tightening remains difficult to quantify with precision. The International Energy Agency reports three-quarters of the 400-million-barrel emergency stock release announced in March has been executed, leaving only a few weeks before those supplies dry up. The Wall Street Journal reported that the U.S. Strategic Petroleum Reserve has been drawn down to its lowest level since 1983. RBC estimated that exports during the ceasefire added an inventory buffer worth roughly 40 days.

China has increasingly become the swing factor, to the point of being described as the OPEC of oil demand. It has curtailed imports, switched to other energy sources, and drawn on its reserves. Citing Kpler data, the Wall Street Journal reported in late July that Beijing could comfortably suppress crude imports for another six months.

There is a credible case for lower prices as well. Citi believes Brent could fall to US$60 a barrel by Christmas as the market returns to surplus, arguing that both the U.S. and Iran have good reason to uphold the memorandum, eventually. Bloomberg has also argued that the global oil cushion is not genuinely near empty, noting China’s remaining reserves and a depletion in developed economies that has been less severe than feared. Strategas points out that continued escalation has not produced a new high in crude, which in itself might confirm supply concerns are not so severe. Then again, the market might be overly optimistic in believing the Strait of Hormuz has to reopen because the consequences would be so dire if it doesn’t.

The range of plausible outcomes for oil prices remains unusually wide, from $60 to above $120, or even higher. We would treat confident forecasts in this area with caution.

The U.S. economy still looks solid

Against that backdrop, the U. S. economy has remained notably resilient. A Bank of America Fund Manager Survey conducted between July 2 and July 9 found 54% of investors expecting no landing for the U.S. economy (meaning continued growth) and only 2% expecting a hard landing. A Wall Street Journal survey of economists put the probability of recession over the next 12 months at only 25%.

Supporting that resilience has been a labour market that continues to perform better than many had expected. Weekly jobless claims reached their lowest level since 1969, and the Wall Street Journal reported that large companies have started hiring again, contrary to predictions of an AI-driven employment decline. The picture is not uniformly good, however. The Wall Street Journal also notes that roughly two million workers now face long-term unemployment.

Strategas offers a useful framework for understanding the resilience. The U.S. economy has absorbed three major supply shocks in the past year: tariffs, reduced labour supply following changes to immigration policy, and the conflict in the Middle East. Growth still looks adequate, with consumer and fixed investment numbers holding up. Strategas nonetheless describes “a gathering storm” of investor concerns, and its own weighted assessment of market conditions sits only slightly positive.

Canada: cooler inflation, and a new tariff threat

Canadian inflation improved more than expected. Statistics Canada reported that consumer prices rose 2.8% in June, down from 3.2% in May and below the 2.9% economists surveyed by Bloomberg had expected. The average of the Bank of Canada’s preferred median and trim core measures fell to 1.85%, the lowest since September 2020 and the first reading below 2% in nearly six years. Gasoline prices fell 10.2% over the month, the largest decline since April 2025. Shelter inflation eased to 1.5%, and grocery price growth slowed to 3.9%, while traveler accommodation rose 10.1% as the FIFA World Cup began.

The Bank of Canada held its policy rate at 2.25% for a sixth consecutive meeting, projecting inflation to average 2.5% in 2026 and to return to the 2% target by early next year, while noting it is prepared to adjust policy as needed. Toronto-Dominion’s Leslie Preston expects the gasoline relief to fade in July’s data but still believes Canadian inflation has peaked this year.

Trade-related uncertainty returned. On July 20, the U.S. announced tariffs of 50% on a wide range of Canadian goods, effective in 30 days, under Section 338 of the Tariff Act of 1930, a provision never previously used for this purpose. Exemptions cover energy, potash, critical minerals, and fish, and goods already subject to national security tariffs such as steel and aluminum. Goods complying with the terms of the 2020 Canada-United States-Mexico Agreement will not be exempt, which means a substantial volume of trade currently crossing the border duty-free would face the higher levy. Prime Minister Mark Carney said Canada “stands ready to engage intensively” to resolve the outstanding issues. The Canadian dollar fell to roughly C$1.41 following the announcement.

Investors should recognize that Canadian and U.S. economic conditions continue to diverge in important ways. Canadian core inflation is now below target while U.S. inflation remains well above it, and the two central banks face different problems. While we question whether either central bank will ease in the coming months, we are more confident that the Bank of Canada will stay on the sidelines than the U.S. Federal Reserve.

This chart shows that the inflationary effect of the war in Iran has so far remained contained to fuel rather than spreading through the Canadian economy, which is why the Bank of Canada has been able to wait.

What this means for long-term investors

July served as a reminder that headline market returns often conceal important shifts beneath the surface. The companies that led markets for much of the past two years came under pressure, while participation broadened, earnings expectations improved, and long-term borrowing costs continued to rise.

Take together, these developments reinforce several long-standing investment principles. When leadership becomes this concentrated, the cost of being wrong about a small number of companies rises significantly, and diversification becomes even more important. When yields are above 5% at the long end, the price investors should be willing to pay for distant profits is lower, and high-quality fixed income offers both income and ballast that it did not offer a few years ago. When policy guidance is deliberately thin, a wider range of outcomes needs to be respected, though not necessarily forecast.