The AI Trade Has Split: Market Breadth, Sector Leadership and the Rotation From Hyperscalers to Enablers

For most of the past three years, "the AI trade" was treated as a single position. Own the megacap platforms building the infrastructure, and you owned the theme. That assumption is now breaking down. Beneath a headline index that continues to print near record highs, participation has narrowed and leadership has quietly rotated. The companies that sell the inputs of the build-out — chips, memory, networking — are behaving very differently from the companies that spend on it.

The analogy is the California gold rush. History remembers the prospectors who rushed west chasing fortune, but the people who reliably grew rich were the merchants who sold them the picks, the shovels, the pans and the tents. Whether any single miner struck gold or went home empty-handed, the suppliers were paid either way. The AI cycle is rhyming with that pattern: the hyperscalers are the prospectors staking enormous, uncertain claims on future demand, while the enablers who sell the essential tools of the build-out are the ones booking dependable revenue from every participant in the race.

This is not a story about a market top or a market bottom. It is a story about market structure: which industries are leading, how broad that leadership is, and whether the relative strength beneath the surface is consistent with the calm at the index level. Those questions matter more than any single price target, and they are the questions a disciplined process is built to answer.

The split beneath the index

The clearest way to see the rotation is to separate two baskets. On one side, the AI enablers — the semiconductor, memory and networking suppliers that sit at the bottlenecks of the build-out. On the other, the hyperscalers — the large platforms funding the capital expenditure: GOOGL, AMZN, META and MSFT.

Measured across horizons, the two baskets have diverged sharply. The enabler group has led over the quarter and the year, while the hyperscaler group has lagged, giving back ground over the same windows even as the broad index held up. The most recent week shows both cooling, but the medium-term structure is unambiguous: relative strength has been concentrated in the sellers of the AI build-out, not its buyers.

This pattern is consistent with what strategists have flagged at the index level. According to Nomura's Charlie McElligott, the pressure on the broad market has come less from macro fears and more from a rotation out of hyperscaler capital-expenditure exposure and into the bottleneck "enablers" — with the large spenders effectively acting as a drag rather than a support. When the biggest weights in the index stop leading, the tape can look stable while the internals are doing something else entirely.

Why the spenders are being questioned

The market's caution toward the hyperscalers is grounded in cash flow, not sentiment. Goldman Sachs Research estimates that the largest cloud operators are on course to spend in the region of $770 billion on capital expenditure in 2026 — roughly equivalent to the entirety of their operating cash flow. To sustain that pace, the group has leaned increasingly on debt and equity issuance and has pulled back on share buybacks, with aggregate net debt rising materially since the start of 2025.

That is the structural reason the "big spenders" have been treated with more scepticism than the "picks-and-shovels" suppliers. The suppliers are converting the build-out into revenue and margin today; the spenders are funding it against future returns that remain uncertain. The same research house notes that AI-infrastructure beneficiaries are expected to account for roughly half of S&P 500 earnings growth this year, with semiconductors as the primary direct beneficiary. Leadership, in other words, is not only narrow — it is narrowing toward one link in the chain.

Breadth is the warning that precedes the rotation

None of this is visible if attention stays on the index level. It becomes visible through breadth — the share of the market actually participating in the advance.

• Goldman Sachs Research has described market breadth as among the narrowest since the dot-com era, with the aggregate index sitting far closer to its highs than the median constituent.
• RBC Wealth Management's work on the "Great Narrowing" puts the combined weight of the ten largest companies at a record level of roughly 40 per cent of the index, a concentration that raises idiosyncratic-shock risk and quietly turns broad exposure into a thematic bet.
• Charles Schwab's mid-year work notes that, despite a calm index, the average member's maximum drawdown has been around −21 per cent year-to-date — evidence of significant churn and rotation beneath the surface.

Narrow breadth does not time a reversal. What it does is change the meaning of an index level. A market carried by a small cohort is more fragile than the same market carried broadly, because there is less underneath to absorb a stumble in the leaders. This is why participation deterioration often appears before index weakness — it is a structural tell, not a forecast.

Portfolio implications: dispersion, relative strength and the long/short lens

A split market of this kind is, first and foremost, a dispersion environment — and dispersion is where relative positioning matters more than direction.

• Selectivity over exposure. When leadership is concentrated in one part of a value chain, broad exposure to "the theme" dilutes the signal. The relevant question is not whether to own AI, but which link — enabler or spender — is carrying the relative strength.
• Concentration risk in what looks diversified. With the top handful of names dominating the index, passive exposure increasingly behaves like a single concentrated position. Portfolio balance has to be assessed at the level of industry participation, not index membership.
• A long/short structure fits the regime. Persistent dispersion between a leading group and a lagging group is the classic setting for a long/short approach — net bias expressed through the leaders, with awareness of the laggards, and diversification that comes from low correlation between the two sides rather than from simply holding more names.
• Drawdown awareness. A −21 per cent median-member drawdown behind a stable index is a reminder that position-level risk can be far higher than the headline suggests. Sizing and patience matter more, not less, in narrow tapes.

Behavioural risks: mistaking a theme for a position

The main hazard in a market like this is narrative chasing — buying "AI" as a slogan rather than as a structured position. The theme is real; that is precisely what makes the reflex dangerous. Several recurring mistakes tend to cluster around narrow, high-momentum tapes:

• Treating the index as diversified when its returns are driven by a handful of correlated names.
• Anchoring to the most familiar leaders — the platforms — while the relative strength has migrated to the suppliers.
• Extrapolating the sharpest part of a momentum move, when history suggests such moves are frequently followed by unstable, mean-reverting periods.
• Ignoring breadth entirely, and so being surprised by volatility that the internals had already signalled.

The antidote is not prediction. It is structure — a repeatable way to check participation, leadership durability and portfolio balance before adding risk.

The ImGeld framework: market, then industry, then stocks

This is the discipline ImGeld is built around. The framework works top-down — market, then industry, then stocks — because leadership and its durability are visible at the industry level long before they resolve into individual names.

The ImGeld Industry Heat Map ranks US industries by strength each trading day, so that a rotation like the one now underway — enablers strengthening, spender-adjacent groups softening — shows up as a change in industry participation and relative strength, not as a headline after the fact. From there, the framework assesses how durable that leadership is and how it should be balanced at the portfolio level, so that exposure reflects where participation actually is rather than where the narrative says it should be.

The point is not to call the next move in the AI complex. It is to read the structure beneath it — breadth, leadership, dispersion — with enough discipline that the position reflects the data rather than the story.

Closing

The AI trade has not ended; it has split. Enablers are leading, the largest spenders are lagging, and the index is calmer than its internals. For disciplined traders, that is a structural signal to study before committing capital:

• Process over prediction.
• Structure over narratives.
• Risk over conviction.

Read participation and leadership first. The position should follow the structure — not the headline.