The AI trade is rotating to hyperscalers, not breaking
The semiconductor selloff looks more like a rotation in AI leadership than the end of the spending cycle. The market is starting to reward the companies that can absorb massive capex and monetize AI at platform scale, not just the suppliers that enabled the first leg.
The cleanest read on this week’s AI wobble is that the trade is moving up the stack, not falling apart. The chip selloff was painful, but it landed just as Alphabet lifted capex plans again and cloud-driven results reinforced that hyperscalers are still building aggressively. That matters because the debate has changed: investors are no longer just paying for scarcity in compute, they are asking where the profit pool settles once AI becomes a service layered into cloud, software, ads, and enterprise workflows. Our view is that this looks less like a broken capex cycle and more like a handoff from hardware leadership to platform leadership.
Start with the obvious: if AI demand were actually cracking, the biggest buyers would be pulling back. Instead, public reporting this week showed the opposite. Alphabet raised 2026 capex guidance by $15 billion after a cloud-driven beat, and broader consensus framing now points to the largest hyperscalers spending more on capex than they generate in free cash flow by 2027. That is not a sign of retreat. It is a sign that the economics of AI are shifting from the bottleneck layer to the ownership layer — the companies with the balance sheets, customer relationships, and distribution to turn infrastructure into recurring revenue.
The valuation gap is part of why this rotation makes sense. NVDA still trades at 35.62x earnings and 19.87x sales, while MSFT sits at 23.58x earnings with a 39.3% net margin and GOOGL trades at 24.39x earnings with a 54.8% net margin. That is the market telling us something important: semis are still priced like the purest AI exposure, but hyperscalers are not exactly low-growth utilities. Alphabet is posting 15.1% revenue growth and 34.2% EPS growth; Microsoft is growing revenue 14.9% with 15.5% EPS growth. Those are platform businesses with real monetization engines already in place, yet they trade at materially less demanding multiples than the hardware names that depend on their spending.
That is why the recent semiconductor drawdown should be read carefully. Reuters highlighted a sharp one-day hit to chip stocks earlier this month, including a 6.5% drop in AMD, even as the broader semiconductor complex had already been massively extended. Morgan Stanley’s Lisa Shalett and Mike Wilson have been making the same broader point: semis have become crowded, and when the suppliers outperform the spenders by too much, the divergence eventually closes. We think the market is beginning that closing process now. Not because chips stop mattering, but because investors are recognizing that hyperscalers own the customer and therefore have the better shot at capturing durable margins.
The comparative setup across the main AI beneficiaries makes the case more clearly:
MSFT: 23.58x P/E, 14.9% revenue growth, 39.3% net margin
GOOGL: 24.39x P/E, 34.2% EPS growth, 54.8% net margin
The point is not that semis suddenly look bad businesses. NVDA still has extraordinary growth, with revenue up 65.5% and EPS growth at 66.0%, and AVGO and ANET remain critical to AI infrastructure. But the market rarely pays peak scarcity multiples forever once the next layer starts to monetize. AMZN is a good example of why the rotation can broaden: at 3.37x sales and 28.8% EPS growth, it offers direct exposure to cloud demand without requiring investors to underwrite hardware-style valuation extremes. If AI workloads keep moving into enterprise and consumer platforms, the companies that bundle compute with software, advertising, commerce, and cloud contracts should start to look like the more complete trade.
Yes, the bear case deserves respect. Alphabet’s cash burn and the broader free-cash-flow squeeze are real, and investors are right to ask whether hyperscalers are simply inheriting the next margin problem. But that critique misses the distinction between spending pressure and spending failure. A chipmaker can sell into a build cycle; a hyperscaler can sell the service on top of it for years. That does not mean every dollar of capex will earn an attractive return, and Christopher Wood has fairly noted that the market is still sorting out which platforms will monetize best. It does mean the strategic asset is no longer just access to GPUs — it is owning the network, cloud, software layer, and user base that turns those GPUs into products.
That is also why we would not frame this as a clean “sell semis, buy hyperscalers” trade. The better framing is that the first phase of AI rewarded the toll collectors on compute scarcity, while the next phase is likely to reward the companies that can spread that compute across giant installed bases. META, for example, still combines 22.2% revenue growth with a 17.94x earnings multiple, even after heavy AI spending has weighed on sentiment. ANET may also hold up better than some chip names because network ownership sits closer to the platform layer than the pure component layer. Leadership can rotate without the underlying buildout ending, just as internet infrastructure spending continued even after equity leadership moved to the platforms that could monetize traffic and distribution.
The market is not debating whether AI matters anymore. It is debating where the margin pool settles once the infrastructure build is no longer the only scarce asset. We think this week’s setup argues for a rotation toward hyperscalers and adjacent platform owners, not a collapse in the theme itself.
What would change our mind is straightforward: a real slowdown in hyperscaler capex commitments, or evidence that cloud and AI revenue growth is failing to keep pace with the spending surge. Until then, the sharper read is that semis are being de-rated from crowded leadership while MSFT, GOOGL, AMZN, and even META become the next leg of the AI trade. On that basis, the TickerSpark Score lens should matter more on platform monetization and balance-sheet endurance than on who sold the last must-have chip.
Our take, not advice. This is opinion commentary — informational only, not personalized investment recommendations. Markets carry risk. Do your own research and consider your own situation before any trade.
▌The Daily Briefing · Free
A new stock idea, every evening.
One stock worth watching each weekday, plus the analysis behind it. Free, in your inbox.
