Goldman Sachs' chief global equity strategist, Peter Oppenheimer, suggested in early August that technology stocks may be facing an earnings problem, rather than a valuation problem. In a research note released on Thursday, he presented evidence to support that assertion. The new report, titled "Competition for Capital," doesn't overturn his August thesis — it reinforces it.
The report links the risk of an AI-driven "earnings bubble" to a specific transmission mechanism, a historical stress test, and a near-term trigger, which Oppenheimer says has already appeared in this month's bond market turbulence. While he still stops short of declaring a bubble is certain, six weeks after first floating the possibility, that cautious judgment is now supported by capital expenditure-to-cash flow data, record credit issuance volumes, and a lowered short-term outlook for equities.
Oppenheimer wrote in early August: "Tech doesn't appear to be in a valuation bubble, but it may be brewing an earnings bubble." That statement drew attention because his firm is one of Wall Street's most prominent long-term bull research desks. At the time, his argument was largely observational. He cited the dramatic stock reactions to quarterly earnings: Microsoft jumped 17% in a single day on stellar results, Meta fell nearly 10% despite beating estimates, and the equal-weight S&P 500 outperformed its market-cap-weighted counterpart by the widest margin since 2009. To him, these signs suggested investors were beginning to doubt that the earnings growth driving the market rally was too narrowly concentrated.
He broadly attributed this risk to "larger government deficits, increased debt issuance and persistent inflation" pushing up the cost of capital, but didn't fully articulate the specific transmission mechanism to tech earnings. Thursday's report upgrades that loose connection into a central thesis. Oppenheimer argues that AI infrastructure spending and government borrowing are competing for the same pool of capital: private companies are issuing debt and equity to finance AI data centers, while governments are ramping up borrowing for infrastructure, energy security, and defense spending — and inflation pressure from higher energy prices pushes policy rates up further. It is this clash of capital demand, he argues, that raises global capital costs.
What was only implied in the August report is now the title and the through-line of the current one. He supports the view with more granular data: capital expenditures by AA-rated tech issuers grew 65% year over year in the second quarter, marking the tenth consecutive quarter that overall capex growth for AA-rated issuers has exceeded 35%. US convertible bond issuance so far this year has reached $135 billion, with AI-related borrowers accounting for 44% of total issuance. Goldman Sachs' credit team has raised its forecast for full-year US investment-grade bond issuance by $200 billion to a record $2.3 trillion, with AI-related issuers making up a quarter of all supply.
Oppenheimer isn't the only senior Wall Street analyst holding this view. Five days before his report, Torsten Slok, chief economist at Apollo Global Management, published his own take, arguing that the past "savings glut" has now turned into a "savings shortage." Slok said the two-decade era of ultra-low interest rates was rooted in excess savings chasing scarce opportunities. "Today the situation has changed," he wrote. "There are now more projects than there is capital... When projects are abundant and capital is scarce, capital has to compete for projects, and the way it competes is by demanding a higher return. The return that clears the market corresponds to a higher yield."
In August, Oppenheimer only briefly referenced historical parallels without fully fleshing them out. He cited several precedents: the banking sector in the 2008 crisis, the internet bubble of the late 1990s, and Japan's bubble in the late 1980s. In each case, it was earnings that collapsed first — not valuations. He pointed out that in the lead-up to the 2008 financial crisis, banks were the largest sector in the S&P 500, but bank stocks never traded at extreme valuations like tech did in 1999 or Japanese stocks did in the 1980s. Instead, bank earnings were supported by rapidly expanding leverage — and the underlying asset for that leverage, US real estate, was where the valuation bubble truly was. When the property market crashed and dragged the economy into recession, bank earnings plunged, even though the stocks themselves weren't obviously overvalued beforehand.
He then applies the same framework to the tech sector and lays out three reasons why the current situation is different. First, tech profits remain "very strong" and balance sheets are "broadly healthy," in sharp contrast to the credit-driven fragility that eventually broke bank earnings. Second, the S&P 500's overall interest coverage ratio sits at the 99th percentile of the past 20 years, with the median stock at the 68th percentile — proof, he argues, that most companies haven't over-leveraged the way banks did. Third, AI compute demand "continues to accelerate and is supply-constrained," and isn't built on top of an asset that's already in a bubble. He cites examples: Microsoft has announced plans to triple its data center compute capacity within six years, and Nvidia, at Goldman Sachs' Communacopia technology conference, reiterated its forecast that AI's addressable market could reach $3 trillion to $4 trillion by 2030.