Goldman Sachs says markets are moving into what it sees as one of the most capital-hungry investment cycles in history, with several large demand pools competing for funding at the same time.
In his latest weekly note, Mark Wilson, Goldman Sachs’ head of European hedge fund business, said the old backdrop of excess savings, abundant capital and low rates is fading. In its place, he sees simultaneous demand from AI infrastructure, reindustrialization, renewed defense investment, power-system rebuilding, supply-chain reconfiguration under deglobalization pressure, and sovereign borrowing needs tied to rising interest bills and expanding welfare spending.
Wilson’s conclusion was blunt: this competition for capital is likely to remain a durable medium-term feature, lifting both the pricing and the cost of capital and changing the investment framework seen over modern market history.
The Fed is a passenger, not the driver
Wilson tied that argument to this week’s move in the US 30-year Treasury yield, which he said decisively broke to levels not seen before the global financial crisis.
The move came as the new Federal Reserve chair, Warsh, has intentionally reduced forward guidance, leaving markets with greater uncertainty around the policy path. Wilson cited historical data showing that of the six Fed chairs since 1970, Ben Bernanke and Janet Yellen each saw a 10% drawdown in their first year, while the other four saw first-year drawdowns of 20% to 36%. Markets have often struggled in the early phase of a new chair’s term.
Even so, Wilson said he agrees with the view that the Fed is more passenger than driver in the current debate. In his reading, structural capital demand matters more than monetary policy in explaining why yields are moving higher. He added that, given the scale of the capital competition he described, investors should not expect the latest breakout in yields to reverse quickly.
Indexes looked calm in July, but internals did not
For investors focused only on broad indexes, July may have looked quiet. Wilson argued the real action was underneath the surface, and that the moves were historic.
He pointed to two trends running at once. First, single-stock dispersion became extreme. On one Friday alone, Amazon rose 15% while Apple fell 10%. Wilson noted that such a divergence in the same session between the world’s two largest companies by market value is itself highly unusual.
Second, momentum factors broke sharply. A market-neutral portfolio fell 40%, exceeding the extreme factor rotation seen in March 2000 during the tech bubble unwind. Wilson said the move created major problems for effective risk management across many strategies.
The result was broad de-risking. Goldman Prime data showed fundamental managers cutting gross exposure to a one-year low, while net long positioning fell back to the bottom quartile. After that washout, Wilson said, market structure is now much cleaner.
Why August may not be an easy buy-the-dip month
Wilson said the cleanup in positioning does not automatically mean investors can turn bullish right away. He laid out several reasons why July’s price action may not reverse quickly.
- The effects of the rate breakout have not yet been fully absorbed, and markets are still recalibrating.
- The extreme volatility seen in July has changed the inputs used in risk models. Wilson cited a one-day 18% jump in the KOSPI and a 26% one-day rise in SK Hynix, moves that can prevent many institutions from rebuilding exposure quickly.
- On a three- to six-month view, attention is likely to shift after summer to the US midterm elections. Wilson cited data showing that in the 13 midterm election years since 1974, the median return for the S&P 500 from early August to election day was 0%.
His bottom line: August is likely to be a digestion period.
Fundamentals remain firm, but the split is widening
Despite the sharp trading swings, Wilson said the fundamental picture has held up better than expected. Unlike a typical year, earnings-per-share estimates for 2026 and 2027 have been revised higher through the year.
He said US single-quarter EPS growth is expected to peak this quarter at about 26%. Europe, by contrast, posted 13% EPS growth in the first half and could accelerate to 19% in the second half, an unusual back-half strengthening pattern.
Memory chips were the exception in his note. Even though the segment has been one of the best performers year to date, Wilson said marginal signals are worsening: spot DRAM prices have stabilized, low-memory-consumption model technology is advancing, and Chinese memory-chip company CXMT has seen its stock rise sixfold from its IPO price, pushing its market value above $550 billion. In his view, that points to a large future increase in supply.
Hyperscalers are spending huge sums, and Goldman says the returns are showing up
A central question this earnings season, Wilson said, was whether hyperscale cloud companies could raise capital spending and still show enough revenue growth and return-on-investment signals to justify it. For Amazon and Microsoft, his answer was yes.
Goldman’s current capex forecasts are:
- Alphabet: $350 billion in 2027 and $415 billion in 2028
- Amazon: $325 billion in 2027 and $366 billion in 2028
- Microsoft: $262 billion in 2027 and $284 billion in 2028
Wilson called the scale staggering.
Operating results, he said, have been just as striking. Google Cloud growth accelerated to 82% year over year. Microsoft said enterprise customers are moving from “frontier models” to “frontier ecosystems,” meaning infrastructure that routes requests across the most suitable model capabilities. AWS revenue growth accelerated to its highest level since the pandemic period.
Amazon says AI annualized revenue has topped $25 billion
Wilson highlighted Amazon management’s comments on the earnings call. The company said: “AI annualized revenue has climbed sharply on a quarter-over-quarter basis and now exceeds $25 billion, with triple-digit year-over-year growth.”
Management also said: “We are seeing AI margins and returns tracking slightly ahead of where the cloud business was at the same stage.”
On spending and capacity, Amazon was even more direct: “Even with 2026 capital expenditures reaching $220 billion, we still will not have enough capacity in 2026 to meet all demand. That will likely also be true in 2027, and the scale of demand in 2028 is already stunning... We have long believed AWS could become a hundreds-of-billions-of-dollars revenue business. We now believe it can be at least twice that, and it could very well become a $1 trillion annual revenue business, with highly attractive free cash flow and returns on invested capital.”
A transition period between private capital and the state
Wilson ended the note with a broader macro frame. He said the market is in a transition period: hyperscale private-sector companies are racing ahead with AI-related investment, while the state sector is becoming more constrained by capital. The tension between those two forces, he said, is likely to become more visible.
At some point, he wrote, global economic reality will force a reckoning with the political choices that come with a repricing of capital. For now, his view is that the AI supercycle still lies ahead, while August is more likely to be a relatively calm period of digestion.