Anthropic's 'Economic Scenarios for Transformative AI': extreme case puts 2030 GDP 32% higher — and one in five cognitive workers out of a job

Start with the mildest path, which actually works best as a reference frame. According to the abstract, under modest change, AI adds less than half a percentage point to GDP growth by 2030 and pushes unemployment up by a tenth of a point — in other words, macroeconomically almost invisible. In the accompanying blog post, Anthropic describes this world as one where AI is a small technology, the economy continues on a normal path, and AI makes changes around the margins. Leading with it is useful because it reminds readers that the other end of this debate remains the perfectly serious possibility that AI just may not matter that much.

The middle path, substantial change, sits closest to mainstream institutional forecasts. The paper puts 2030 GDP 8.3 percent above the no-AI path — roughly $36.3 trillion at 2025 prices — with growth over the twelve months to 2030 reaching 5.4 percent, a pace the authors benchmark against the 4.7 percent fastest growth of the 1990s dot-com boom in 1999. The costs land unevenly on cognitive workers: cognitive unemployment rises from 2.9 percent in mid-2026 to 4.5 percent in 2030; cognitive wages end the period 0.3 percent below their no-AI path while wages in all other occupations run 5.9 percent higher, and the labor share of income falls from 60 percent to 56.1 percent. The report notes this scenario is comparable in magnitude to forecasts circulated by Goldman Sachs, McKinsey Global Institute and others around 2023 — roughly 1.5 percentage points of annual labor-productivity growth.

The third path is the stark one. Under extreme change, AI performs nearly half of today's cognitive work by 2030 and annual GDP growth hits 15 percent — compounding to an economy that doubles in size every four and a half years. Yet in the same report the labor share of income falls from 60 percent today to 45 percent, the cognitive wage sits 11.5 percent below its no-AI path, and wages in all other occupations run 34 percent above it. The abstract says only that nearly one in five cognitive workers is unemployed; per OfficeChai's reading, that corresponds to cognitive unemployment rising to 17.9 percent in the extreme case, pushing economy-wide unemployment to 11.9 percent, beyond anything seen in a typical recession. Business Today adds a counterintuitive detail: in this most aggressive scenario, labor's share of GDP is overtaken by capital — workers take 45.2 percent, capital 54.8 percent.

The researchers' own caveats need to be spelled out, or the work reads like a prophecy. The paper states plainly that the scenarios are not predictions and that no probabilities are attached to them — the purpose is to make the consequences of different assumptions comparable. The model also lists what it omits: catastrophic risks, political-economy considerations, business cycles and possible financial-market disruptions are all outside the framework, and because price rigidities and demand feedback are not modeled, it cannot generate the negative loop in which a shock depresses demand and thereby amplifies its own labor-market damage. One more counterintuitive finding is worth remembering: even in the extreme scenario, AI's boost to the pace of innovation is relatively minor, because research remains bottlenecked by physical tasks — labor productivity through that channel rises by well under one percent in all three scenarios. The authors also surveyed 10,980 US adults; the median respondent's answers sit near the substantial scenario (roughly 8 percent higher GDP and 4 percent lower cognitive employment by 2030), meaning public expectations are far less dramatic than the paper's most aggressive line.

There is also a timing detail that may matter most to anyone planning a career: the report notes that almost all of the divergence across the three paths comes after 2027. That means little will be visible in the short run — but it also means the adjustment window is finite, and if the economy does track the substantial or extreme path, the split will concentrate in the next two or three years. Axios and NPR both frame the report the same way: a frontier lab using its own methodology to put the macroeconomic endgame of its own product on the table for inspection. As for whether the conclusions are alarmist, the paper's own posture is restrained: it does not tell you what will happen, only what follows if particular assumptions hold.