The defining feature of today's market is not enthusiasm for artificial intelligence. It’s the growing insistence that profits justify enthusiasm.

Benjamin Graham’s old observation that markets are voting machines in the short run and weighing machines in the long run feels increasingly relevant.

For much of the past two years, markets rewarded AI exposure first and asked questions later.

That phase is ending.

Strong earnings are still being rewarded, but expectations have risen with share prices. Increasingly, companies are being judged not by the scale of their AI spending but by whether that spending produces tangible economic returns.

That distinction matters because the AI investment cycle has become more important to markets than many investors appreciate.

One of the striking features of the current earnings season is the degree to which profit growth is being driven by AI-related investment. Binky Chadha at Deutsche Bank has noted that AI beneficiaries contributed about half of S&P 500 earnings growth in the second quarter. More broadly, the 10 largest contributors accounted for almost three-quarters of total growth. Semiconductor manufacturers, memory producers, networking companies, power suppliers and data centre operators have all benefited from an extraordinary wave of investment spending.

This helps explain why earnings have remained resilient despite softer signals from other parts of the economy. Since the current bull market began in late 2022, roughly three-quarters of global equity returns can be explained by earnings growth rather than investors simply paying higher valuations for the same profits, based on expected profits over the next 12 months. That is encouraging: markets supported by earnings are generally healthier than those driven purely by optimism.

The question, however, is how much of those earnings ultimately depends upon the continuation of the AI investment boom itself.

Technology has led the rally

Technology-heavy US equities have outpaced the broader US market since late 2022, raising the bar for future earnings delivery.

A tale of two economies

One useful way to understand the current environment is to view it as a K-shaped economy. The upper branch consists of businesses benefiting directly from the AI investment cycle: semiconductors, data centres, power infrastructure, and cloud computing. The lower branch is more subdued. Hiring has slowed, housing remains weak, and manufacturing is mixed. The result is not a broad-based boom, but an economy in which growth is concentrated in a narrow set of activities linked to technology, infrastructure and productivity.

This helps explain why equities have climbed despite unremarkable economic data: investors are focusing on where profits are being generated.

Encouragingly, leadership is beginning to broaden into financials, industrials, and parts of Europe. That gives markets a healthier foundation rather than dependence on a handful of AI winners.

Reflexivity and the AI cycle

The most important question for investors may not be whether the economy slows. It may be whether the AI investment cycle slows.

George Soros’s idea of reflexivity is useful here: market prices do not just reflect reality; they can shape it. Rising prices encourage investment and reinforce the trend. Falling prices can reverse the loop.

That matters because today’s market is unusually dependent on a single investment theme: high share prices support capital expenditure, capital expenditure supports earnings, and earnings, in turn, support share prices.

The late-1990s technology boom offers a warning. It didn’t end because the recession arrived first. Confidence weakened, share prices fell relative to expected profits, capital spending slowed, and economic weakness followed.

Forecasts for cumulative AI infrastructure investment now exceed $5 trillion by the end of the decade, making this one of the largest waves of corporate spending on long-term infrastructure in modern history. While investors remain willing to fund the build-out, that spending supports earnings growth, employment, infrastructure demand, and broader activity. 

If investors lose faith in the long-term economics of the AI build-out, the consequences could extend beyond share prices. Lower valuations would make financing more expensive, capital raising harder, and investment plans more vulnerable. Markets wouldn’t merely be reacting to economic weakness; they could help cause it. That is not the most likely near-term outcome, but it’s the principal vulnerability.

The AI build-out meets financial reality

The AI story is no longer simply about technological possibility. It’s increasingly about capital allocation – how companies decide where to invest their money.

Like previous investment booms – railways, telecommunications and the internet – AI requires financing. What matters is the quality of that financing and whether the returns justify the capital employed.

Some recent developments suggest that parts of the AI ecosystem are becoming more financially interconnected. Debt issuance – companies raising money by borrowing – linked to AI infrastructure has grown rapidly, while more sophisticated financing arrangements are emerging throughout the supply chain. Multi-year purchase commitments, vendor financing, and debt-funded data-centre expansion can all reinforce the cycle as demand expectations rise. Still, they may also amplify disappointment if demand falls short.

That may prove rational if demand evolves as expected, but it creates vulnerability if expectations change.

Vendor financing and leverage – borrowing or other financing used to support investment – tend to look safest when growth is abundant. They become more visible when growth disappoints.

Recent earnings reports suggest that investment spending is still translating into real revenues and profits. Yet the market is becoming more discerning. Investors increasingly want evidence that spending today will generate attractive returns tomorrow.

This is not necessarily a sign of weakness; it’s evidence that the investment cycle is maturing.

A healthier market

The most striking feature of today’s market is not persistent enthusiasm for AI, but that the enthusiasm is becoming more discriminating.

Investors are rewarding evidence rather than ambition, profits rather than promises, and execution rather than aspiration.

The current cycle still appears to have solid foundations. But those foundations are tied to one of the largest capital investment programmes the corporate world has undertaken. The question is no longer whether AI will matter. It’s whether the economics of AI can meet the expectations already embedded in markets.

For investors, the next phase of the AI trade will be judged less by the scale of ambition than by the discipline of capital allocation and the durability of returns.