Anthropic IPO: four ways the AI giant wants to disrupt science

A silhouetted hand holding up a smartphone with the Anthropic logo on the screen, in the background is the Claude logo.

Anthropic, the frontier artificial-intelligence company, has made history with the largest initial public offering (IPO) on record on 19 November. With a staggering valuation of around US$1.5 trillion, the firm, based in San Francisco, California, is now among the world’s most valuable companies — and it has major ambitions in science.

The maker of AI model Claude, Anthropic was founded in 2021 by a group who broke away from rival giant OpenAI. Its models have proven powerful in research, and Anthropic has launched its own directed science effort, focusing on areas including quantum mechanics and protein design. This year alone, it has reported mathematics breakthroughs and launched its first life sciences laboratory.

The IPO might also unleash a historic wave of philanthropy: all seven of Anthropic’s founders are likely to become billionaires and have pledged to donate 80% of their equity in the company to charity, including chief executive Dario Amodei. And on 18 October, the company announced that it had established the Anthropic Public Benefit Fund, created to support initiatives that benefit society and stocked with an initial commitment of 5% of Anthropic’s shares.

But the firm has drawn criticism from researchers, who say its aims of curing diseases on short timescales are unrealistic and that its race towards developing ultra-powerful, ‘superintelligent’ AI models pose serious risks, including an existential threat to humanity. In its public prospectus, published ahead of the IPO, Anthropic outlines a new governance structure that it says is “consistent with our commitment to AI safety and responsibility”, and adds that safety research directly shapes the firm’s “model design, training, and deployment.”

Nature looks at four areas of science where Anthropic is making waves.

Drug discovery and biotech

AI’s potential to turbocharge the discovery of new medicines has been a centerpiece of Anthropic’s pitch to investors. “If today we look at things like cancer, Alzheimer’s, heart disease, mental illness — these will be relics of the past in, I think, less than a decade,” Amodei said in a promotional video ahead of the IPO.

The company has been rapidly expanding its biomedical ambitions. This year, Anthropic opened its first ‘wet’ laboratory for life sciences research, announced partnerships with several pharmaceutical companies, and said it was pursuing drug leads of its own. “We plan to scale this work up substantially from here,” the company said on 27 October.

But Amodei’s habit of promising swift, AI-fueled cures for complex diseases has rankled many researchers and entrepreneurs in the field who feel the timeline is unrealistic.

Biology and protein design

Over the last year, Anthropic has made a major push in the life sciences. In June, the company hired John Jumper, who shared the 2024 Chemistry Nobel Prize for his work on AlphaFold2 while at Google DeepMind in London.

Some of the company’s early biological research received a mixed reception from biologists. After Anthropic reported using a swarm of AI agents to find signs of a new CRISPR-like system in viral sequence data, some researchers questioned the novelty of the finding and noted that the company had not carried out the experimental work needed to determine the system’s function.

In protein design — a field that has boomed thanks to specialist AI tools including AlphaFold — Anthropic’s work has also failed to generate much excitement. In August, Anthropic used Claude Science — an agentic system that can access other biological tools and databases — to make novel ‘binders’ against 15 protein targets.

“Expectations are sky high,” says Anthony Gitter, a computational biologist at the Morgridge Institute for Research in Madison, Wisconsin. “I understand why some view their initial public results as underwhelming.”

But Gitter is enthusiastic about Anthropic’s protein design efforts, largely because the company was transparent about the work, from the prompts it used to the sequences that were generated. “I would strongly encourage Anthropic to continue that in their future life sciences research,” Gitter adds.

In recent months, Anthropic and other companies — notably Google DeepMind and OpenAI — have been working to establish their research credentials in large part by showing off the rapidly growing mathematical abilities of their flagship AI models.

On 7 September Anthropic mathematician Levent Alpöge and collaborator Tristan Buckmaster at New York University solved a special case of the Navier-Stokes equations, a long-standing problem in mathematical physics that describes the flow of fluids. Hours after Anthropic announced the result, OpenAI claimed a full solution of the problem, igniting a controversy.

In another collaboration, Alpöge and mathematical physicist Ethan Sussman at Northwestern University in Evanston, Illinois, have been working towards establishing the mathematical foundations of the standard model of particle physics. This too is a problem on which OpenAI has made similar claims.

Anthropic and its competitors see the ability to write a rigorous proof of a mathematical fact as a crucial testing ground for their reasoning models. That’s in large part because the correctness of a proof is an objective measure that can be used as a ‘reward’ to further train AI systems.

Enjoying our latest content?
Log in or create an account to continue

  • Access the most recent journalism from Nature's award-winning team
  • Explore the latest features & opinion covering groundbreaking research

Original source Anthropic IPO: four ways the AI giant wants to disrupt science

Back to home