Anthropic, the San Francisco artificial-intelligence company behind Claude, made history on November 19 with the largest initial public offering (IPO) ever. The company is now valued at approximately US$1.5 trillion and is pursuing ambitious research programmes in areas including drug discovery, protein design, quantum mechanics and mathematics. Its rapid expansion has also prompted debate over the feasibility and risks of its scientific goals, as well as the company’s responsibility to fund research and philanthropy.
Shares of artificial-intelligence company Anthropic began trading on the Nasdaq Global Select Market on Thursday in the largest initial public offering in history. Credit: Idrees Mohammed/AFP via Getty
Anthropic was founded in 2021 by a group that had separated from rival AI company OpenAI. Its Claude model has already been used in research, and Anthropic has launched scientific initiatives focused on subjects such as quantum mechanics and protein design. This year, the company reported results in mathematics and established its first life-science research institute.
The IPO could also lead to a significant wave of philanthropy. All seven of Anthropic’s founders are likely to become billionaires, including chief executive Dario Amodei, who has pledged to donate 80% of his company stock to charity. On October 18, Anthropic announced the creation of the Anthropic Public Benefit Fund, which will support initiatives intended to benefit society. The fund has an initial commitment of 5% of Anthropic stock.
At the same time, researchers have criticized the company’s claims that AI could help cure diseases in the near term. They argue that such timelines are unrealistic and warn that the race to build increasingly powerful, “super-intelligent” AI systems could create serious risks, including threats to humanity’s survival. In a public prospectus issued before the IPO, Anthropic described a governance structure that it said was “consistent with our commitment to AI safety and accountability”. The company added that safety research would directly influence its models’ “design, training, and deployment”.
Drug discovery and biotechnology
AI’s potential to accelerate the development of new medicines is central to Anthropic’s pitch to investors. “If you look at cancer, Alzheimer’s disease, heart disease, mental illness today, I think in less than 10 years these will be relics of the past,” Amodei said in a promotional video released before the IPO.
Anthropic is expanding rapidly into biomedical research. This year, it opened its first “wet” laboratory for life-science research, announced partnerships with several pharmaceutical companies and said it was pursuing its own drug candidates. “We plan to significantly expand this initiative in the future,” the company said on October 27.
Amodei’s predictions of rapid, AI-enabled treatments for complex diseases have concerned many researchers and entrepreneurs in the field, who believe that the proposed timelines are unrealistic.
Biology and protein design
Anthropic has also made substantial advances in life sciences over the past year. In June, the company hired John Jumper, who shared the 2024 Nobel Prize in Chemistry for his work on AlphaFold2 while he was at Google DeepMind in London.
Some of Anthropic’s early biological research received mixed reactions from biologists. After the company reported using a swarm of AI agents to identify signs of a new CRISPR-like system in viral sequence data, researchers questioned how novel the finding was. They also noted that Anthropic had not carried out the experimental work needed to establish what the system could do.
The company’s work in protein design has attracted a similarly cautious response. In August, Anthropic used Claude Science, an agent system with access to biological tools and databases, to create novel “binding agents” against 15 protein targets.
“Expectations are very high,” says Anthony Gitter, a computational biologist at the Morgridge Institute in Madison, Wisconsin. “I can understand why some might view the first published results as underwhelming.”
Even so, Gitter is enthusiastic about Anthropic’s protein-design programme, largely because the company was transparent about its methods, including the prompts it used and the sequences it generated. “I strongly encourage Anthropic to continue this in their future life science research,” he added.
Mathematics
In recent months, Anthropic and other companies, particularly Google DeepMind and OpenAI, have sought to demonstrate the mathematical capabilities of their leading AI models and establish their scientific credentials.
On September 7, Anthropic mathematician Levent Arposis and Tristan Buckmaster of New York University solved a special case of the Navier–Stokes equations, a long-standing problem in mathematical physics that describes the movement of fluids. Hours after Anthropic published its results, OpenAI sparked controversy by claiming to have produced a complete solution to the problem.
In another collaboration, Alpöge and Ethan Sussman, a mathematical physicist at Northwestern University in Evanston, Illinois, have been working to establish the mathematical foundations of the Standard Model of particle physics. OpenAI has made similar claims in this area.
Anthropic and its competitors view the ability to produce rigorous mathematical proofs as an important test of AI reasoning. A key advantage is that the correctness of a proof can be assessed objectively, allowing it to serve as a “reward” for further training of an AI system.
Source: www.nature.com












