Anthropic Survey Finds 91% of Staff Expect AI Could Make a Nobel-Level Discovery by 2030

Anthropic says 91% of its surveyed staff expect AI to be capable of making a Nobel-level scientific discovery by 2030, compared with 39% of U.S. adults. The confidence gap comes as the company approaches an expected November IPO, where investors will weigh AI's promise against the costs and uncertainties of scaling it. Anthropic's survey\n\nThe report compares an August poll of 1,334 Anthropic staff with a Morning Consult survey of 10,980 U.S. adults. The public sample was weighted to represent the adult population; the staff poll was unweighted. Respondents were asked when AI could perform eight tasks at the level of a skilled professional, including making a Nobel-level discovery. The result measures expectations about capability, not a verified probability, a company forecast, or a discovery already made.\n\nThe divide extends to work. Anthropic staff expect AI to be capable of performing an average of 75% of their current work hours by 2030, compared with 28% among employed U.S. respondents. The report also points to internal use: after engineers adopted coding agents, the number of lines of code they added tripled in the following month; sales representatives closed more deals after adopting AI agents. Those are measures from one company and do not establish that similar gains will occur elsewhere. In the public survey, respondents identified customer preference for human service, privacy and security, reliability, and legal or liability concerns as obstacles to workplace adoption. The survey findings\n\nInvestor sentiment is more mixed. Some fund managers who attended Anthropic's investor day found Dario Amodei's presentation uplifting, while some investors discussed a valuation around $1.5 trillion, below earlier expectations of roughly $2 trillion, and were disappointed by slower sequential growth. Investor-day reactions and valuation discussions show the tension facing the offering. The IPO is expected to be an early test of public-market appetite for AI developers investing heavily to scale while still losing money. Market context\n\nThe report also cautions that technical ability does not ensure adoption: how widely AI is used will depend on workplace choices and constraints as well as model progress. For Anthropic, the public-market test will be whether its employees' confidence can be matched by evidence that AI capabilities translate into reliable, broadly adopted products and sustainable returns.