Asger Risom
Wall Street has spent the past two years bidding up anything tied to artificial intelligence.
Wall Street has spent the past two years bidding up anything tied to artificial intelligence.
Now Goldman Sachs is asking whether the market has simply gotten too far in front of the technology’s actual economic payoff — and its answer is cautiously skeptical.
Big gains, modest grounding
In a new assessment published Monday, Goldman analysts framed the debate as “the most important question for the U.S. equity market outlook”: whether investors have correctly valued AI’s long-term earnings potential. The bank’s conclusion is that valuations are elevated but not yet showing the hallmarks of a full-blown bubble.
Even so, the team led by Dominic Wilson and Vickie Chang found that current pricing appears “well ahead of the macro impact.” Their analysis argues the stock market has already assumed a level of eventual economic benefit that sits at the very top of what is realistically achievable.
Goldman uses a macro lens to set what the analysts call “constraints on what is collectively possible,” a shift from the company-by-company focus that usually dominates AI enthusiasm.
A valuation surge at the edge of plausibility
The bank’s estimates place the Present Discounted Value of generative AI–driven capital revenues for the U.S. economy at about $8 trillion in a baseline scenario. The plausible range stretches from $5 trillion to $19 trillion — already large enough, in Goldman’s view, to justify the current wave of AI-related capital spending.
But the market has moved even faster. Since ChatGPT debuted in November 2022, Goldman calculates that companies directly involved in or exposed to the AI ecosystem have collectively gained more than $19 trillion in value. That figure includes major moves in semiconductors, hyperscalers and roughly $1 trillion in valuation among the three largest private model developers.
This total places investors’ implied AI windfall at the upper limit of Goldman’s entire projected benefit range — and more than double the $8 trillion midpoint. In fact, the analysts note that the combined value added to chip designers, model providers and similar firms “already exceed the $8 trillion baseline estimate of increased capital revenues.”
Why markets may be overreaching
Goldman stresses that forward-looking markets always price future gains early, calling this “a feature, not a bug.” But the bank warns that two recurring behavioral errors could be amplifying today’s optimism.
The first is the “fallacy of aggregation,” where investors assume that the extraordinary earnings achieved by a few leaders can be applied across all AI winners simultaneously. The result, Goldman says, is a risk that the “joint value” baked into chipmakers, model builders and cloud giants surpasses what the broader economy can actually deliver.
The second is the “fallacy of extrapolation.” Early profits from breakthrough technologies often fade as competition intensifies, but markets frequently price them as if they will persist indefinitely. Goldman notes that “past innovation-driven booms—like the 1920s and in the 1990s—have led the market to overpay for future profits even though the underlying innovations were real.”
A powerful technology with uneven profits so far
Despite these warnings, Goldman doesn’t question AI’s “underlying productivity promise.” Its models suggest generative AI could lift U.S. productivity by roughly 1.5 percentage points annually for a decade, raising GDP and corporate earnings by around 15% in the long run.
That outlook helps explain why investors continue to treat AI as transformational even as present-day profits remain clustered in hardware. As Goldman puts it, as long as the economy and the AI capex boom “remain on track,” the market is likely to keep its optimism intact.
But the analysis also hints at a pressure point: outside the companies selling chips and compute, actual AI earnings are still thin. If revenue takes longer than expected to show up, the market may discover that its $19 trillion head start was too much, too soon.
Wall Street has spent the past two years bidding up anything tied to artificial intelligence.
Now Goldman Sachs is asking whether the market has simply gotten too far in front of the technology’s actual economic payoff — and its answer is cautiously skeptical.
Big gains, modest grounding
In a new assessment published Monday, Goldman analysts framed the debate as “the most important question for the U.S. equity market outlook”: whether investors have correctly valued AI’s long-term earnings potential. The bank’s conclusion is that valuations are elevated but not yet showing the hallmarks of a full-blown bubble.
Even so, the team led by Dominic Wilson and Vickie Chang found that current pricing appears “well ahead of the macro impact.” Their analysis argues the stock market has already assumed a level of eventual economic benefit that sits at the very top of what is realistically achievable.
Goldman uses a macro lens to set what the analysts call “constraints on what is collectively possible,” a shift from the company-by-company focus that usually dominates AI enthusiasm.
A valuation surge at the edge of plausibility
The bank’s estimates place the Present Discounted Value of generative AI–driven capital revenues for the U.S. economy at about $8 trillion in a baseline scenario. The plausible range stretches from $5 trillion to $19 trillion — already large enough, in Goldman’s view, to justify the current wave of AI-related capital spending.
But the market has moved even faster. Since ChatGPT debuted in November 2022, Goldman calculates that companies directly involved in or exposed to the AI ecosystem have collectively gained more than $19 trillion in value. That figure includes major moves in semiconductors, hyperscalers and roughly $1 trillion in valuation among the three largest private model developers.
This total places investors’ implied AI windfall at the upper limit of Goldman’s entire projected benefit range — and more than double the $8 trillion midpoint. In fact, the analysts note that the combined value added to chip designers, model providers and similar firms “already exceed the $8 trillion baseline estimate of increased capital revenues.”
Why markets may be overreaching
Goldman stresses that forward-looking markets always price future gains early, calling this “a feature, not a bug.” But the bank warns that two recurring behavioral errors could be amplifying today’s optimism.
The first is the “fallacy of aggregation,” where investors assume that the extraordinary earnings achieved by a few leaders can be applied across all AI winners simultaneously. The result, Goldman says, is a risk that the “joint value” baked into chipmakers, model builders and cloud giants surpasses what the broader economy can actually deliver.
The second is the “fallacy of extrapolation.” Early profits from breakthrough technologies often fade as competition intensifies, but markets frequently price them as if they will persist indefinitely. Goldman notes that “past innovation-driven booms—like the 1920s and in the 1990s—have led the market to overpay for future profits even though the underlying innovations were real.”
A powerful technology with uneven profits so far
Despite these warnings, Goldman doesn’t question AI’s “underlying productivity promise.” Its models suggest generative AI could lift U.S. productivity by roughly 1.5 percentage points annually for a decade, raising GDP and corporate earnings by around 15% in the long run.
That outlook helps explain why investors continue to treat AI as transformational even as present-day profits remain clustered in hardware. As Goldman puts it, as long as the economy and the AI capex boom “remain on track,” the market is likely to keep its optimism intact.
But the analysis also hints at a pressure point: outside the companies selling chips and compute, actual AI earnings are still thin. If revenue takes longer than expected to show up, the market may discover that its $19 trillion head start was too much, too soon.
Sources: Goldman Sachs, Fortune
