The AI bull market reaches a crossroads! Wall Street sees a rare major debate between bulls and bears

Artificial intelligence trading has entered its third year, and Wall Street is now witnessing a rare public split in its judgment of this rally’s nature.

Top macroeconomists at BCA Research have recently engaged in an internal bull-bear debate over the AI frenzy, with both sides taking sharply opposing views. The bears argue that the US stock market is currently in an earnings bubble, with valuations priced for the most optimistic scenarios, leaving almost no room for positive surprises.

The bulls, on the other hand, insist that the compute power market is under-supplied, the rate of return on capital expenditures is severely underestimated by the market, and the breadth of profit growth is far greater than generally recognized. The core of this debate is not whether AI is real, but how much longer the frenzy can last and what price the market will eventually pay when it ends.

On Monday, US stocks broadly closed lower. The S&P 500 Index fell 0.2%, the Dow Jones Industrial Average dropped 0.6%, and the Nasdaq Composite Index was little changed.

Bears: Earnings bubble has formed, valuations surpass 2000 internet peak

BCA Research Chief Economist Peter Berezin and Chief Macro and Emerging Markets Strategist Arthur Budaghyan clearly expressed their bearish view during last week’s live bull-bear debate. They argue that the current US market is in the midst of an earnings bubble—profit figures may look strong, but the sustainability of profit margins is far less solid than they appear, and valuations have already fully priced in the most ideal outcomes.

From a valuation perspective, almost every metric shows that stock prices are expensive. Adjusting the S&P 500’s profit margins back to 2019 levels, its current forward P/E ratio is about 27x, higher than the historical peak of roughly 26.5x in March 2000. Even after excluding technology and financial stocks, valuations in the remaining sectors are still not cheap—according to data from BCA Research and FactSet, this part of the market is currently trading at 26x rolling P/E, while profit growth over the past two to three years is just about 3%.

On the issue of AI capital expenditures, Berezin and Budaghyan point out that large technology hyperscalers are collectively investing hundreds of billions of dollars in data centers, specialized chips, power grids, and energy upgrades. While these massive capex investments boost reported earnings, they erode real cash flow. "Every chip sold is booked as revenue and profit on the seller’s reports, while the hyperscaler buying the chips records it as capital expenditure, not operating cost. As a result, reported earnings go up, but cash flow does not improve in tandem," the two said.

Berezin further noted, "AI may ultimately follow a trajectory similar to that of electricity—it boosts efficiency for companies, but if everyone gains access, it may not necessarily make companies more profitable."

Bulls: Compute power outstrips supply, hyperscaler valuations at decade lows

Juan Correa, Head of Portfolio Construction, and Noah Weisberger, Head of Equity Research at BCA Research, take the exact opposite stance. They believe that the compute power market is still in a state of scarcity, not surplus; the market underestimates, rather than overestimates, the return on capital expenditure; and profit growth is much broader than just the few companies typically associated with AI.

On the demand side, BCA Research data shows that hyperscalers’ cloud business backlogs—orders received but not yet fulfilled—have expanded by about $750 billion over the past two quarters, indicating that industry demand remains ahead of supply capacity.

In terms of valuation, Correa and Weisberger argue that while market participants are generally concerned about high valuations, hyperscalers are actually trading at their lowest valuation levels in nearly a decade, even as their rate of return continues to improve.

From broader profit signals, the latest generation of GPU rental prices have continued to rise throughout the year, cloud contract value per megawatt has roughly doubled compared to a year ago, and the ratio of EBIT to headcount has surged; meanwhile, profits outside of core cloud businesses are also growing rapidly, including AI-driven advertising boosts at Meta and a strong rebound in Google’s search business. "The market will be amazed by the actual profit scale produced by these capital expenditures over the next two quarters," they said.

Main divergence: not "if," but "when" and "how it ends"

It’s noteworthy that this bull-bear debate is not about the authenticity of AI technology itself—both sides agree that AI is fundamentally reshaping corporate productivity, capital allocation, and the labor market. This premise is undisputed.

The real disagreement lies in the timing: To what extent can the AI spending frenzy continue to outrun skepticism, and when the market eventually faces a reckoning, how severe will the impact be?

Interestingly, even the bulls admit that this rally will eventually end destructively. "The capex trade will ultimately end, and likely in a disruptive way," said Correa and Weisberger. "But for now, economic and fundamental tailwinds will dominate. For the bulls, the biggest concern is valuation compression—not economic recession or a profit collapse."

This means that, whether investors realize it or not, they are already involved in this debate—and the outcome will be determined by the market over the coming quarters.