In September 2024, The New York Times profiled Goldman Sachs research head Jim Covello as “Wall Street’s leading AI skeptic.” He warned that Silicon Valley was spending hundreds of billions of dollars chasing a technology whose commercial value remained speculative. The core thesis was blunt: the economics did not justify the hype.
Covello’s skepticism was not subtle. He questioned whether generative AI solved expensive enough problems to justify the scale of infrastructure investment being demanded by the market. He argued that many AI products lacked a clear path to monetization. He framed the entire AI buildout as potentially resembling prior technology bubbles fueled by speculative enthusiasm rather than sustainable returns.
At the time, this skepticism distinguished him from a financial industry eager to embrace AI evangelism. The New York Times article portrayed him as one of the few senior Wall Street voices willing to publicly challenge the prevailing narrative.
And yet, less than two years later, Goldman Sachs research notes and media coverage now describe an aggressively AI-oriented investment posture that favors hyperscalers and AI infrastructure plays.
The shift is striking.
Not because views cannot evolve.
But because the underlying logic appears to have changed without any meaningful acknowledgment of the contradiction.
The Original Skepticism
Covello’s earlier critique rested on several central arguments:
- AI infrastructure spending had exploded far faster than demonstrated commercial demand.
- Companies were investing extraordinary amounts of capital without evidence of corresponding revenue generation.
- Generative AI lacked sufficiently expensive real-world problems to justify trillion-dollar investment cycles.
- Historical technology bubbles often involved massive infrastructure buildouts preceding disappointing returns.
The argument was not merely cautionary. It was existential.
The implication was that markets were irrationally pricing future AI value without adequate proof that customers would ultimately pay enough to justify the cost structure.
This skepticism resonated precisely because it addressed a real weakness in the AI narrative: monetization.
Even today, many generative AI companies continue operating at enormous losses while relying heavily on investor subsidies, venture capital, or hyperscaler spending.
The skepticism itself was not unreasonable.
What is remarkable is how quickly the framing appears to have shifted from “the economics do not make sense” to “buy the infrastructure anyway.”
The New Position
The messaging increasingly resembles a traditional Wall Street rotation thesis:
- Infrastructure providers may outperform.
- Hyperscalers remain central beneficiaries.
- AI spending continues driving capital allocation decisions.
- Investors should focus on where the next wave of monetization emerges.
That is not skepticism.
That is participation.
And participation is not inherently problematic.
The problem is the absence of any serious reconciliation between the earlier critique and the current positioning.
If the core economics of generative AI were fundamentally unsound, then why would the infrastructure layer now represent an attractive investment thesis?
If the industry lacked viable monetization, why would hyperscaler dominance become bullish rather than alarming?
If excessive spending was irrational in 2024, what specifically changed?
Where is the inflection point?
Where is the evidence that the commercial assumptions previously criticized have now been validated?
Because without that explanation, the shift begins to look less like disciplined analysis and more like strategic repositioning around prevailing market sentiment.
The Value of Contrarianism
Another possibility is that the original skepticism was never intended to function purely as a long-term investment thesis.
On Wall Street, highly contrarian positions often generate disproportionate attention precisely because they cut against prevailing market enthusiasm. In a financial environment saturated with bullish AI commentary, becoming one of the industry’s most visible skeptics carried its own strategic value: visibility, differentiation, and media relevance.
That does not necessarily invalidate the original analysis. But it does raise a fair question about whether the forcefulness of the earlier position reflected durable conviction or the professional advantages that accompany being the loudest dissenter during a speculative cycle.
And once the market made clear that AI infrastructure spending would continue at scale regardless of unresolved monetization concerns, the incentives surrounding that skepticism may have shifted as well.
The Incentive Structure Problem
This contradiction also reveals a broader issue inside financial analysis culture.
Wall Street research often presents itself as objective and rigorously data-driven. In reality, institutional incentives frequently reward proximity to momentum.
During periods of technological enthusiasm, skepticism attracts attention.
But once capital markets commit to a dominant narrative, research frequently shifts from questioning the premise to optimizing exposure within it.
That appears to be exactly what happened here.
The debate quietly moved:
From:
“Is this entire AI spending cycle economically rational?”
To:
“How should investors position themselves inside the AI spending cycle?”
Those are fundamentally different questions.
The first evaluates whether the emperor has clothes.
The second assumes the emperor exists and merely debates which tailor profits most.
The Convenient Reframing of AI
There is also a deeper rhetorical maneuver occurring throughout much of Wall Street’s AI coverage.
When AI hype appears excessive, analysts emphasize discipline, monetization, and skepticism.
When AI stocks rally, the conversation shifts toward infrastructure demand, data center growth, enterprise adoption curves, or secondary beneficiaries.
The goalposts move constantly.
AI does not need to fully justify itself today because the promise of future adoption becomes the new justification.
And when adoption metrics remain unclear, infrastructure spending itself becomes evidence of opportunity.
The cycle becomes self-referential.
Companies spend because investors expect AI growth.
Investors expect AI growth because companies are spending.
Meanwhile, analysts position themselves to appear simultaneously cautious and opportunistic.
If AI disappoints, they can point to earlier skepticism.
If AI succeeds, they can point to later bullish positioning.
The result can resemble institutional hedging presented as analytical evolution.
What Actually Changed?
That remains the unanswered question.
Because the underlying concerns Covello previously raised have not disappeared.
The AI industry still faces:
- extraordinary infrastructure costs,
- uncertain long-term monetization,
- unclear productivity replacement timelines,
- significant energy and compute demands,
- growing commoditization pressure,
- and unresolved questions surrounding profitability.
In many respects, those concerns have intensified.
Yet the market conversation increasingly treats continued AI spending itself as validation.
That is not necessarily analysis.
Sometimes it is simply momentum wearing the costume of inevitability.
Criticism Is Not Disparagement
Criticizing public investment theses, market analysis, or economic reasoning is not the same thing as making personal attacks.
Jim Covello is a public-facing financial analyst whose professional role involves publishing market opinions designed to influence investors.
Those opinions are fair subjects for scrutiny.
And scrutiny matters most precisely when narratives shift without clear acknowledgment.
Analysts routinely critique companies, sectors, executives, business models, and macroeconomic assumptions.
Public analysts do not become immune from criticism simply because they work for large institutions.
The relevant issue is not whether someone changed their view.
Serious analysts should evolve when evidence changes.
The issue is whether the change was transparently explained, logically reconciled, and intellectually consistent.
That is the standard financial commentators themselves routinely apply to everyone else.
Conclusion
Perhaps Jim Covello’s original skepticism was overstated.
Perhaps the newer infrastructure thesis is correct.
Perhaps both positions contain elements of truth.
But what remains difficult to ignore is the speed with which Wall Street skepticism transformed into strategic participation once the market made clear that AI capital spending would continue regardless of unresolved concerns.
The contradiction matters because it reflects something larger than one analyst.
It reflects the tendency of institutional finance to oscillate between caution and enthusiasm without fully accounting for the inconsistency between the two.
On Wall Street, conviction often matters less than staying aligned with the direction of capital.