Goldman Hints at AI Trade Rotation: The Real Winners Are Now Platform and Productivity Stocks, Not Just Chips and Data Centers

Recent market pullbacks are a normal part of a maturing cycle, not a sign the boom is over. The AI trade has stumbled at the start of 2026, with several distinct tech stock pullbacks over the course of 2025. This mirrors the dot-com era, where tech stocks delivered positive returns over 80% of the time in the last 20 years despite significant volatility. The recent rotation away from AI infrastructure companies reflects a market becoming more selective, not abandoning the theme.

This pattern is familiar. In the late 1990s, tech stocks outperformed global equities by a wide margin, yet the fear of a dramatic drawdown haunted investors. The AI boom, like the dot-com wave before it, is a multi-year technological shift that requires investment across every layer of the tech stack. Such waves don't end abruptly; they evolve. The dot-com bubble was the only compute wave to crash, but it was also the only one to reach a crescendo after six years of extreme returns before collapsing. The current AI cycle is still in its early, scaling phase.

The divergence in stock performance this year is evidence of that evolution. Investors have rotated away from AI infrastructure companies where operating earnings growth is under pressure and capex spending is debt-funded. At the same time, they have rewarded companies demonstrating a clear link between AI investment and revenue. This selective rotation is healthy-it's the market separating durable productivity beneficiaries from over-leveraged spenders. The consensus estimate for 2026 capital expenditure by AI hyperscalers is climbing, but analyst estimates have consistently underestimated this spending. The market is now focusing on which companies can convert that spending into profits.

The bottom line is that volatility is the price of participation in a structural wave, not a signal of its demise. While the AI trade may be more volatile in 2026, the underlying momentum from earnings, adoption, and strong balance sheets persists. The market isn't abandoning the theme; it's refining its bets within it.

The Wrong Way: Focusing Only on Infrastructure Spending

The boom is not just about building the machine; it's about who gets to use it. The market's recent rotation away from pure infrastructure is a clear signal that investors are looking past the data centers and chips to where AI's value is actually being captured. This shift is structural, not a temporary pause.

Analysts have consistently underestimated the scale of the build-out. The consensus estimate for 2026 capital expenditure by AI hyperscalers has been revised up to $527 billion, continuing a trend of upward revisions. This persistent underestimation shows how hard it is to forecast the true pace of investment, but it also highlights that the infrastructure phase is already well underway and being funded. The divergence in stock performance this year proves the market is no longer willing to reward all big spenders equally. It's now focused on which companies can convert that spending into real revenue.

That focus is now turning to the next beneficiaries. According to Goldman Sachs Research, the next phases of the AI trade will involve AI platform stocks and productivity beneficiaries. This means shifting attention from semiconductor and data center operators to the software and services companies that provide the tools and platforms for AI adoption. The early signs are there: while some AI platform stocks have struggled to show AI-enabled revenue growth, they are proving to be an exception to the broader trend of underperformance.

The most concrete evidence of this monetization is in the numbers. Last quarter, hyperscaler cloud revenues grew 27% year-over-year on average. That's the clearest signal that AI investment is translating into commercial activity and top-line growth. It shows the boom is moving from a capital-intensive build-out phase into one where the value is being realized through services and productivity gains. The market is simply following that money.

The Wrong Way: Ignoring Structural Differences from Dot-Com

The bubble narrative persists, but it fails to account for fundamental differences in how AI is being built and funded. Unlike the dot-com era, where the boom was driven by speculative consumer spending and venture capital, today's AI adoption is powered by corporate capital expenditure. This shift is structural. According to recent analysis, AI-driven capital expenditure is now the dominant contributor to US growth, with functionally all economic growth in the United States coming from these investments in late 2025. This is a different engine-one that starts with balance sheets, not business plans.

The funding model is also more robust. While the dot-com bubble was fueled by a surge in venture capital, today's build-out is supported by a broader mix of private equity and credit. This has enabled massive raises for established companies and funded the trillions in data center projects that are the physical backbone of AI. The industry's dependence on key materials and geopolitical stability, however, introduces a unique systemic risk not present in the 1990s. The global economy has become dependent on the AI industry, with much of its supply chain-chips, data centers, and fuel-relying on key materials produced in or transported through just a few places, particularly the Middle East. A regional conflict there could disrupt the entire build-out, creating a fragility that was absent in the dot-com era.

This creates a new kind of vulnerability. The boom is now so deeply embedded in the global economy that its disruption would be more systemic. As one observer noted, the interlocking points of fragility are clear, from advanced chips to the energy that powers them. The risk is not just a market correction but a potential halt to the investment cycle itself. In this light, the historical comparison is less about valuations and more about the scale and interconnectedness of the modern build-out. The dot-com crash was a bubble bursting in a specific sector. The AI build-out is a global economic project with a single point of failure in a volatile region.

A Clearer Framework: The AI Boom's Real Drivers

The AI boom is now a central engine of the global economy, but its trajectory is being shaped by a new set of drivers and risks. The core force is corporate capital expenditure, with AI-driven investment now the dominant contributor to US growth. In the final months of 2025, functionally all economic growth in the United States came from these investments, a scale not seen in previous tech cycles. This is a structural shift from the dot-com era, where the boom was fueled by venture capital and speculative consumer spending. Today's build-out is powered by balance sheets, creating a more robust but also more interconnected foundation.

The next key signal to watch is the market's rotation from infrastructure to platform and productivity stocks. This divergence is a natural marker of the cycle's maturity. Investors have already rotated away from AI infrastructure companies where earnings growth is under pressure and capex is debt-funded. The focus is now on companies that can demonstrate a clear link between AI investment and revenue, such as major cloud platform operators. This shift means the market is no longer just betting on the machine being built; it's betting on who gets to use it profitably. The next phases of the AI trade, as identified by Goldman Sachs, will involve these platform and productivity beneficiaries, signaling that the boom is moving from a capital-intensive build-out phase into one where value is being captured through services and efficiency gains.

The primary risk to this entire setup is a global energy shock stemming from Middle East instability. The AI industry has become dependent on a supply chain that relies heavily on key materials produced in or transported through just a few regions. A conflict there could disrupt the entire build-out, halting the capital-intensive investment cycle. This introduces a systemic fragility absent in the dot-com era. The risk is not just a market correction but a potential halt to the investment engine itself, which would be devastating for the leveraged tech firms and private lenders that have funded this race. In this light, the historical comparison is less about valuations and more about the scale and interconnectedness of the modern build-out. The boom is now so embedded in the global economy that its disruption would be more systemic.