Valuation test
The rebound from last week’s drawdown never quite shook the sense that it was being supported by borrowed conviction. The S&P 500 once again tested near the 7000 level (6986 as the high watermark) and failed, despite a macro backdrop that would normally be interpreted as supportive of risk. Weak retail sales and softer small-business sentiment reinforced expectations for Federal Reserve easing, pushing rate-cut pricing toward roughly 60 basis points for the year and pulling Treasury yields lower in a clean bull flattening. That impulse worked well for bonds. It did not translate cleanly into equities.
This is an important distinction. The market is no longer responding uniformly to the idea that weaker data automatically lifts stocks. Lower yields are helping duration assets at the margin, but equity investors appear more selective and more fragile in their conviction. Early gains faded, Nasdaq struggled to hold technical support, and leadership narrowed as defensives and value-oriented segments quietly outperformed growth-heavy indices.
Beneath the surface, anxiety around AI-related headline risk is clearly elevated. Recent selling pressure has not been driven by material changes in near-term fundamentals, but by sensitivity to perceived long-term disruption risk. We observed this first in SaaS, then in insurance, and now in parts of financial services, particularly wealth and brokerage firms. Moves in several of these stocks appear disproportionate to the actual news flow, suggesting positioning and narrative risk are doing more work than detailed fundamental reassessment.
Software stabilization later in the session was constructive, but the broader message remains that investors are quick to de-risk first and evaluate second. Retail flows continue to show dip-buying behaviour, which has helped cushion downside risks, but that support has not yet been sufficient to push the index through key resistance. The Dow closing at record highs while the Nasdaq lagged is a clear signal of a more defensive tilt rather than renewed enthusiasm for growth beta.
Technically, the S&P 500 remains capped near 7000, an area that has become a meaningful gamma and positioning threshold. Until the market can establish acceptance above that level, rallies are likely to be treated as tactical rather than structural. Volatility measures have stabilized, but they are not signalling an all-clear. This market is still debating which narrative deserves the premium.
The larger and more consequential issue sits beyond the daily data and headlines. Capex guidance from the hyperscalers has reset expectations in a way that directly challenges the equity support framework of recent years. Aggregate 2026 capex projections across the major platforms now approach $650 billion to $740 billion, implying growth far in excess of prior consensus and absorbing all free cash flow essentially. The implication is straightforward. Buybacks, a critical and underappreciated source of index-level support, are likely to diminish materially.
AI-related uncertainty has heightened investor sensitivity, leading to faster de-risking and reduced tolerance for headline risk.
The key question that first surfaced in 2024 has never really gone away. Is generative AI ultimately a productivity engine, or does it risk becoming a capital sinkhole whose returns struggle to justify the scale of investment required? That debate was never resolved. Instead, the numbers simply kept getting larger.
It is now back at the center of market pricing. Capital expenditure guidance across the hyperscaler complex has reset expectations in a way that forces a more uncomfortable conversation around return on invested capital. Despite broad-based earnings growth, equity markets have been notably unresponsive. Alphabet delivered its strongest Search performance in three years and its best Cloud results in four. Amazon posted its fastest year-over-year AWS growth in roughly three years. Microsoft reported Azure growth of nearly 37 percent. Meta delivered roughly 24 percent topline growth. None were rewarded.
That response is telling. The issue is not demand. It is capital intensity.
If aggregate AI-related capex approaches $700 billion, as current guidance implies, it would effectively absorb all hyperscaler free cash flow. That level of investment would place the AI buildout close to the peak investment impulse observed in the late 1990s, when capital expenditure reached roughly 1.4 percent of GDP. While still below the scale of the Industrial Revolution, it is large enough to warrant a reassessment of capital returns, balance-sheet flexibility, and shareholder support mechanisms.
The market’s anxiety is therefore understandable. Buybacks have quietly been among the most important stabilizers of index-level performance over the past cycle. A sustained shift from capital return to capital absorption materially changes that dynamic.
The debate over the internal rate of return will likely persist for several more years, particularly as contractual commitments begin to mature and the economics of the AI ecosystem are tested under real cash-flow constraints. But even before those outcomes are clear, the scale of announced spending has reopened a critical fault line. Investors are no longer debating whether AI is transformative. They are debating whether the cost of building it is becoming more important than the promise of what it might deliver.
Credit markets are beginning to reflect this uncertainty. The sector has begun trading with a higher risk premium relative to the bond market, a notable shift given how tightly tech credit had traded for much of the past decade. As issuance accelerates and balance sheets absorb continued investment, investors are demanding greater compensation for uncertainty regarding future cash flows.
Elsewhere, cross-asset signals were broadly consistent with a cautious macro read. The dollar softened modestly; gold remained well supported near $ 5,000; silver held steady; crude prices drifted lower despite geopolitical headlines; and bitcoin failed to fully recover from earlier weakness, although leverage conditions have normalized somewhat.
Near-term attention will turn, as usual, to payrolls, though recent changes to the birth/death assumptions raise the risk of increased noise and reduced signal quality in the headline data. But the market’s bigger question is not about one data point. It concerns whether easier financial conditions alone are sufficient to justify higher multiples at a time when the index’s largest constituents are shifting from capital return to capital absorption. Until that tension is resolved, 7000 remains less a launchpad and more a ceiling.
Read the full article here