Anthropic’s CEO warns that half of all entry-level white-collar jobs could vanish within one to five years, potentially driving U.S. unemployment to between 10% and 20%, a range that at its upper end would exceed any level seen since World War II.
Goldman Sachs estimates that 300 million jobs globally are exposed to AI-driven automation, a figure that is difficult to ignore, according to Goldman Sachs Research.
J.P. Morgan Private Bank tackled that question in a sweeping analysis published on April 21, 2026, and the verdict from strategists Jacob Manoukian and Justin Biemann is far more measured than the doom-and-gloom narrative suggests. The report identifies three specific forces that could keep AI from steamrolling the entire labor market at once.
J.P. Morgan identifies three constraints holding back AI job displacement
Manoukian and Biemann, the bank’s U.S. head of investment strategy and its global investment strategist, contend that three limiting factors will slow AI’s march through the workforce:
- Capability constraints of current AI models: Manoukian and Biemann argue that existing systems still face limitations in reasoning, accuracy, and the ability to handle complex real-world tasks.
- Massive physical infrastructure required to scale the technology: The two strategists note that expanding AI adoption depends on building costly data centers, securing advanced chips, and meeting rising energy demands.
- Regulatory, organizational, and sociopolitical resistance: They also emphasize that governments, companies, and the public are likely to impose rules, adapt cautiously, and push back against rapid disruption.
This framing matters because the most prominent AI leaders are painting a far darker picture of what lies ahead for workers
AI models are improving fast, but the reliability gap with human workers remains large
The bank acknowledges that frontier AI models can complete tasks that take human experts approximately 12 hours in less than half the time, a massive leap from just months earlier. The catch is that reliability drops sharply when the success threshold rises even modestly, the report argued, using data from AI benchmarking firm METR.
At an 80% success rate, current models can handle work that takes a human expert just over one hour, compared to nearly 12 hours at a 50% threshold, according to METR benchmarking data cited by J.P. Morgan.
A study by Stanford economist Erik Brynjolfsson, MIT’s Danielle Li, and Lindsey Raymond found that AI assistance boosted customer support worker productivity by 14% on average, with the largest gains accruing to novice and lower-skilled workers, according to the Quarterly Journal of Economics.
The largest gains are going to novice workers, who saw a 34% improvement in solving issues per hour, according to the National Bureau of Economic Research. The market pricing itself tells the story: a white-collar worker earning $75,000 per year costs roughly $50 per hour, while a GPU chip can be rented for $2.50 per hour, a 20-to-1 cost advantage.
Milan_Jovic/Getty Images
Data centers and power grids impose a physical ceiling on AI’s labor market impact
Replacing 10 million workers at the technology’s current performance ratio would require nearly 50 million specialized GPU chips, roughly double the U.S. pipeline of planned capacity through 2028.
Research firm Epoch AI estimates that only 24 frontier data centers are operational, under construction, or planned across the country, housing approximately 25 million chips by the end of 2028, J.P. Morgan noted.
Microsoft’s Fairwater data center campus in Mount Pleasant, Wisconsin, Phase 1 of which went live in early 2026, represents a $3.3 billion initial investment, with Microsoft’s total Wisconsin commitment reaching more than $7 billion across two facilities, according to Microsoft.
Epoch AI estimates the campus will reach a peak electrical load of 3.3 GW in late 2027, exceeding Los Angeles’s average of 2.5 GW. Grid interconnection wait times currently range from three to five years, and TSMC, which produces over 90% of leading-edge chips, has only recently begun expanding capacity to meet surging demand.
Regulators and corporate caution will slow AI deployment across major industries
Financial services regulators, including the Federal Reserve and the Office of the Comptroller of the Currency, have issued guidelines on the use of AI models in decision-making, according to J.P. Morgan Private Bank.
“The big story in 2026 in labor will be AI…If we see some job losses pulled forward, that sets the stage for potential underperformance relative to our forecast, and that may lead the Federal Reserve to cut rates.” said Joseph Briggs, Co-lead of global economics, Goldman Sachs Research.
McKinsey research suggests that even routine enterprise software takes 18 to 36 months to move from pilot to scaled deployment. The political dimension is heating up as well, with Missouri Republican Senator Josh Hawley positioning himself as an explicitly anti-AI presidential candidate for 2028, Axios reported in January 2026.
Yann LeCun, former Meta chief AI scientist and now Executive Chairman of AMI Labs, weighed in on April 18, 2026 in a post on X, urging the public to listen to economists rather than AI company executives when evaluating labor market risks.
Current labor data offer reassurance, but younger workers face growing pressure
The Yale Budget Lab published an updated analysis in April 2026 tracking AI’s impact on U.S. employment since ChatGPT’s launch and found no substantial acceleration in labor market disruption attributable to the technology. The U.S. unemployment rate stood at 4.3% in March 2026, with the economy adding 178,000 nonfarm payroll jobs that month, the Bureau of Labor Statistics reported.
Anthropic’s research team released a study in March 2026 reaching a similar conclusion, finding limited evidence that AI has meaningfully affected employment to date when measured across broad occupational categories, according to Anthropic.
More Wall Street:
- JPMorgan resets S&P 500 price target for the rest of 2026
- Vanguard challenges the S&P 500 as a one-stop strategy
- Goldman Sachs resets Broadcom stock forecast
The picture is more complicated for younger workers, however, and Brynjolfsson’s research team found that employment among workers aged 22 to 25 in the most AI-exposed occupations fell by 13% relative to late 2022.
The Dallas Federal Reserve published research in February 2026 showing that wages are climbing in AI-exposed occupations that value tacit knowledge, gained through experience, while entry-level workers whose jobs rely on textbook knowledge face mounting competition from automated systems.
J.P. Morgan sees investment opportunity where others see only disruption
The infrastructure bottleneck is not just a brake on worker disruption; the bank frames it as an investment thesis for firms supplying the physical backbone of the AI buildout. The bank pointed to chipmaker Nvidia, which showed approximately 73% year-over-year revenue growth and 68% operating margins, as well as Broadcom.
Goldman Sachs Research’s Briggs estimates that, in his base-case scenario, with AI adoption spreading over a decade, the unemployment rate would rise by just 0.6 percentage points.
For workers wondering whether AI is coming for their role, the answer from J.P. Morgan’s research is not if, but when, and the bank is betting the timeline will be far more gradual than the loudest voices in Silicon Valley predict.
Related: J.P. Morgan shares tips to protect your portfolio amid Middle East tensions