Is the Market Expensive? Earnings, Volatility, and the Capital Behind AI

By Juan Carlos Carvallo

The US equity market enters earnings season at all-time highs, with expectations to match. In the latest episode of Goldman Sachs' The Markets podcast (July 15, 2026), Brian Garrett, head of equities on the firm's Cross-Asset Sales team, offered a diagnosis worth unpacking. But Goldman is not alone: contrasting its view with recent analysis from FactSet, BlackRock, UBS, S&P Global, the Bank of England, MSCI and Cboe reveals a more nuanced consensus than the "bubble" headline dominating the conversation. The relevant divide is not bulls versus bears, but those who see a bubble approaching versus those who see a real growth cycle that has simply become riskier and more selective.

A doubly high bar — that earnings are so far clearing

Garrett identifies two simultaneous demands on US equities. The first is fundamental: Goldman's analysts expect year-on-year earnings growth of roughly 22%. The second comes from the options market: the average S&P 500 stock prices in an implied move of 5.5% around its earnings print — 60 to 70 basis points above the long-run average. In this environment good results are not enough: companies must surprise, on both earnings and price reaction.

FactSet's data, however, suggests earnings have so far justified much of the rally. The estimated Q2 growth rate for the S&P 500 stands at 23.6%, and given the historical pattern of positive surprises, FactSet projects final growth could exceed 29% — the best print since Q4 2021. Of the first 18 companies to report, 89% beat estimates, with aggregate surprises of 14.5%.

My read: the risk is not that earnings will be bad. The risk is a company reporting good results but soft guidance; robust growth but higher capex; strong revenue but weaker free cash flow. In this market, "good" may not be enough.

There is a positive counterweight: per Goldman's prime brokerage team, about 75% of global equities bought off the April lows had been sold heading into earnings. Hedge fund positioning is clean; if results deliver, there is dry powder to buy back in.

Is this an AI bubble?

Opinions divide here, and the contrast is instructive.

The constructive case is led by BlackRock. Rick Rieder argues this is an earnings-led rally, not a multiple-expansion rally: since October, tech's forward multiple has compressed from 30.7x to 23.9x and semiconductors' from 28.7x to 25.7x, while forward EPS growth runs at 45% and 88% respectively. Cumulative data center investment has reached roughly $1.5 trillion, annualizing at $214 billion per year — more than the inflation-adjusted cost of the US Interstate Highway System. Not a speculative boom, says BlackRock: an infrastructure buildout with tangible demand behind it.

The more cautious view comes from MSCI, with a data point worth attention: in the US, AI-exposed companies increased capex by nearly 60% year-on-year through May — almost ten times the pace of non-AI peers — while their revenue grew at a fraction of that rate. In Taiwan and China, by contrast, sales growth is keeping pace with capex. MSCI's conclusion: over-investment risk is, for now, a predominantly US phenomenon.

My synthesis: this does not look like a repeat of 1999 — there are real revenues, solid balance sheets, enterprise customers and substantial profits — but there are segments with bubble characteristics: companies without cash flow, small leveraged suppliers, data centers built on aggressive projections, leveraged exchange-traded products, and stocks priced for perfection for years. The bubble, for now, is more likely in the ecosystem's peripheral layers than in the large leaders.

The AI bill: where consensus is strongest

Goldman estimates the AI ecosystem will need to raise between $5.5 and $6 trillion between 2025 and 2030. Other houses agree on direction even while measuring different things: UBS forecasts AI capex of $820 billion in 2026 and $990 billion in 2027, and market estimates put big-hyperscaler capex around $750 billion for 2026, up more than 60% year-on-year. The figures are not directly comparable — some measure only corporate capex, others include external financing, energy and related assets — but all point to historically extraordinary investment.

The implication is a profound transformation: Microsoft, Amazon, Google and Meta — for two decades asset-light businesses with massive buybacks — are becoming capital-, debt-, energy- and asset-intensive businesses. Increasingly they will need to be analyzed with the tools applied to utilities, telecoms and infrastructure: return on invested capital, capacity utilization, equipment life and investment payback, rather than headline capex growth.

Debt is the main emerging risk

Here Goldman, S&P Global, the Bank of England and others align almost completely.

The most telling data point comes from the Bank of England's July 2026 Financial Stability Report: the five large hyperscalers accounted for just 3% of outstanding US investment-grade debt at end-2025 — but over 15% of year-to-date issuance through early May. Their 2026 issuance is comparable in scale to UK gilt issuance over the same period. The market is feeling the weight: Goldman's hyperscaler bond basket widened 22 basis points in a single week, and Garrett notes an unusual divergence between credit protection (CDS) and put skew on these names — credit and equity markets are not pricing the same risk the same way.

The risks go beyond volume: concentration (bond funds can only absorb so much of one sector), free-cash-flow erosion while capex lasts, growing refinancing dependence, off-balance-sheet financing (SPVs, guarantees, data center securitizations, private credit) that obscures where risk actually sits, duration mismatch between 30-year debt and short-cycle technology assets, and a degree of circularity — companies investing in customers who then use that capital to buy their compute. The Bank of England warns that if AI debt financing scales as expected, an adverse shock to AI companies could materially affect global financing conditions. None of these houses sees a solvency crisis at the leaders — their balance sheets remain strong; the risk lies in supply absorption and structure.

The "capex taper tantrum": a double-edged risk

UBS adds the most interesting nuance: the danger is not only that spending is excessive, but that it stops growing at the expected pace. It calls this a "capex taper tantrum": a deceleration in capex growth could hit hard the stocks that price in continuous investment increases — semiconductors, memory, servers, electrical equipment, cooling, data center construction. UBS expects hyperscalers' cash capex requirements to overtake operating cash flow as early as Q3 2026, forcing them to fund increasingly through debt and equity.

The market can punish two opposite scenarios: capex too high (hurting cash flow, margins and credit) or capex lower than expected (hurting the entire supplier chain). That two-sided risk explains much of the volatility ahead.

The volatility the VIX doesn't show — confirmed by Cboe

Garrett insists the VIX doesn't tell the whole story, and Cboe's data confirms it precisely: average single-stock volatility (the VIXEQ index) sits near 45%, while the VIX hovers around 15.8% — and the spread between the two hit a record 29 points. The cause is historically low correlation: stocks are moving a lot, but in different directions. Cboe also reports a record 35% of S&P top-100 stocks trading with inverted call skew, confirming the speculative appetite for call options Garrett describes.

The extreme case is Korea: KOSPI implied volatility exceeds global financial crisis levels, driven by index concentration in a couple of extremely volatile names and the proliferation of leveraged products. And that is a global phenomenon: one in five existing ETFs incorporates a leveraged or inverse component; among those issued in 2026, one in three. As these vehicles gather assets they amplify the market's short gamma — selling more as the underlying falls, buying more as it rises — accelerating both rallies and corrections. They don't create the AI thesis, but they amplify its moves.

Goldman's two trades

Garrett closed with two concrete ideas. First, a collar on concentrated stock positions: keep the stock, buy a 10% out-of-the-money put, fund it by selling a 15%–20% out-of-the-money call. What's unusual is that in roughly 10% of S&P names and 15% of Nasdaq names, call implied volatility exceeds put implied volatility — something Garrett hadn't seen in twenty years. The market is effectively subsidizing the cost of insurance.

Second, leveraged index protection via digital ("one-touch") options: if the S&P falls 7% at any point before August expiry, the payout is roughly 5 times the premium; a 10% drop pays about 10 times. A necessary warning: binary options are sophisticated instruments where 100% of the premium can be lost; they are tactical hedges for qualified investors, not structural positions.

Where each house stands

Goldman is tactically cautious: very high expectations, extreme single-stock volatility, cheap index protection. BlackRock is constructive but selective: the rally is earnings-backed, and it recommends harvesting elevated volatility by writing covered calls on high-conviction names. UBS is structurally positive — double-digit 12-month return potential for the AI theme — but tactically careful on taper-tantrum risk; it recommends staying invested with selectivity and diversification. S&P Global and the Bank of England watch from the credit side: volume is still being absorbed, but the growth, complexity and concentration of the debt could breed fragility. MSCI recommends geographic diversification: over-investment risk is highest in the US, while Taiwan, Korea and parts of Europe own essential links of the chain with better alignment between investment and sales.

My consolidated view

Assigning subjective weights to scenarios: in my base case (55%), the AI buildout continues and the market advances, but with periodic 7%–12% corrections, strong sector rotation, and outperformance by suppliers with visible contracts, revenue and cash flow. In the bull case (25%), compute demand keeps outrunning supply and monetization ends up justifying the capex. In the adverse case (20%), capex far outruns monetization, spreads widen and projects get deferred — and the correction would likely start in private credit, suppliers dependent on one or two customers, momentum stocks and leveraged ETFs, not necessarily in Microsoft, Alphabet or Amazon.

The practical implications I draw: keep exposure to the AI theme but reduce concentration; favor companies with visible monetization, long-term contracts and cash flow; broaden exposure toward infrastructure (energy, electric grids, memory, automation, and selectively Asia); be demanding with AI credit — scrutinize collateral, covenants, off-balance-sheet structure and refinancing dependence rather than the issuer's name; and use protection selectively, taking advantage of expensive calls and relatively cheap index protection.

The consensus is not that AI is about to collapse. It is more nuanced: the revolution is real, but its financing is changing the market's structure and raising risk. After the strong rally, the right strategy does not appear to be exiting the market, but moving from indiscriminate exposure to a more selective, diversified and partially hedged portfolio. When insurance is cheap and expectations are at record highs, buying it isn't pessimism: it's discipline.

---

Sources: Goldman Sachs, "The Markets" podcast (Jul 15, 2026, Brian Garrett and Chris Hussey); FactSet Earnings Insight (John Butters, Jul 10, 2026); UBS CIO, "Capex taper tantrum" (Michael Bolliger, Jul 10, 2026); BlackRock, "The AI Boom and Two-Speed Economy" (Rick Rieder, Jun 2026); Bank of England, Financial Stability Report (Jul 2026); Cboe Macro Volatility Digest (Mandy Xu, Jun 2026); MSCI, "Mapping AI Exposure Across Global Markets" (Jul 2026); S&P Global Ratings, CreditWeek.

This article is for informational and educational purposes only. It does not constitute investment advice or a recommendation to buy or sell any security. The opinions expressed are personal and do not necessarily represent those of LifeInvest Wealth Management. Every investor should assess their particular situation with a financial advisor before making investment decisions. Past performance is not indicative of future results.