The second quarter of 2026 was, in many respects, the inverse of what preceded it. Where the first quarter brought broadening participation and a more distributed opportunity set, the second delivered a powerful but deeply narrow rally fueled almost entirely by one theme: artificial intelligence, and the semiconductor stocks supplying it. Brendan Ryan, CFA of Algorithmic Investment Models (AIM) captures the dynamic precisely in his firm's Q2 2026 quarterly commentary: "the broadening we saw in the first quarter gave way to significant narrowing in the second. Outside of this singular theme, there were few opportunities to outperform the major indices."
That framing sets the rhetorical stakes for everything that follows. In a market that essentially demanded a single bet, AIM's models declined to make it, and the results held up.
What Drove the Rally, and Why It Looks Fragile
The catalyst was Anthropic. On February 5th, the release of Claude Opus 4.6 transformed the company's commercial trajectory, turning it into what Ryan describes as "the leading AI lab." What followed was a cascade: massive fundraising rounds, a disclosed annualized revenue run-rate of $47 billion (roughly five times the start-of-year figure), and a tenfold increase in contracted compute commitments. That surge in demand collided with an already tight market for AI hardware. GPU rental prices, which ordinarily decline as hardware generations age, moved sharply higher instead.
Memory chips became the defining bottleneck. The three major suppliers of high-bandwidth memory, Micron, Samsung, and SK Hynix, more than doubled prices per unit and saw combined profits increase nearly tenfold to $96 billion in a single quarter. Ryan puts this in sharp context: the top five AI hyperscalers spent roughly $150 billion on capital expenditure in the same period. "Pure price increases from memory companies," he notes, "are now consuming a large amount of the total spend, a factor that will necessarily force returns on investment lower."
This is the critical pivot in Ryan's argument. The windfall for memory suppliers came directly out of the pockets of the largest AI buyers, intensifying competition among platforms that once operated as near-monopolies within their respective verticals.
A Fever Pitch, and the Question of What Comes Next
Semiconductor stocks marched steadily higher through April, but Ryan flags a historically unusual condition: "many assets are exhibiting high volatility while sitting close to all-time highs." Within AIM's ETF universe, the only comparable episode was the final run-up of the technology bubble, notably the only prior instance where a volatility spike preceded a drawdown rather than marked its recovery. Ryan is measured in his framing of this signal, acknowledging that volatility is not itself a risk, but noting that for semiconductors specifically, "the current setup of high volatility within 10% of an all-time high has yielded negative returns on average."
Crucially, that elevated volatility has not spread to the average S&P 500 constituent. This distinction matters: it helps explain why AIM's models continue to favor equities broadly, while avoiding the AI-related positions that appear "uniquely risky at the moment."
There is also a demand-side question that Ryan raises with evident care. Near-term demand may have been temporarily inflated by what he calls "Claude-fever" — corporate enthusiasm to experiment with powerful models, sometimes without fully appreciating deployment costs at scale. In one notable case, a company reportedly spent $500 million on Claude in a single month. Uber consumed its entire annual AI budget within months of the year starting. "While long-term demand seems likely to be enormous," Ryan observes, "supply could still outpace near-term adoption as implementation becomes the next bottleneck."
Energy, Diversification, and the Underappreciated Hedge
Against the AI narrative, energy equities underperformed during the quarter despite what Ryan characterizes as a compelling fundamental setup. Markets grew increasingly confident that both the U.S. and Iran were incentivized to restore oil flows through the Strait of Hormuz. But tanker traffic through the Strait had not meaningfully recovered as of early July. "From a fundamental standpoint, oil now looks like a fairly cheap hedge," Ryan says. "Energy equities are already discounting lower oil prices, but the companies should continue to make excess profits as long as oil remains elevated in the near term."
More structurally, energy and commodities continue to offer low or even negative correlation with broad equities, at a moment when fixed income has provided little protection or return. These diversification characteristics, Ryan argues, allow the models to carry more equity exposure than they otherwise would.
Five Key Takeaways for Advisors and Investors
- Concentration in AI is a risk, not just an opportunity. Semiconductor volatility near all-time highs has historically preceded drawdowns, not recoveries. Benchmarks now carry meaningful AI weight; portfolios should not follow blindly.
- Memory chips are the canary. The tenfold profit surge at Micron, Samsung, and SK Hynix came directly at the expense of the hyperscalers funding AI buildout. Margin compression for the largest AI buyers is already underway.
- Near-term AI demand may be overstated. Corporate experimentation is not the same as profitable deployment. Implementation bottlenecks and budget discipline will moderate adoption curves before long-term demand is realized.
- Energy remains a credible, underpriced portfolio hedge. With energy equities already pricing in lower oil and Strait of Hormuz disruptions still unresolved, the asymmetry favors continued exposure, particularly given its diversification value relative to both equities and fixed income.
- Diversification across uncorrelated positions is outperforming quiet conformity. Japan, biotech, clean energy infrastructure, and China A-shares generated meaningful contribution in Q2 without requiring semiconductor exposure. The eclectic portfolio, assembled with discipline rather than consensus, held up where concentrated positioning was most vulnerable.
The quarter's lesson, delivered calmly by Ryan's analysis, is not that AI is a bad investment. It is that buying the most volatile assets at the highest prices, driven by fear of falling behind, rarely ends well. AIM's strategies remained ahead of their asset-allocation benchmarks through Q2 without significant exposure to the trade dominating headlines. That is, as Ryan puts it, something to be "extremely pleased" about.
Footnote:
Ryan, Brendan. "A Broad Approach in a Narrow Market." Algorithmic Investment Models, Q2 2026, https://algomodels.com/a-broad-approach-in-a-narrow-market/. Accessed 9 Aug. 2026.
