Anthropic has built a "wet lab" in the San Francisco Bay Area, extending its AI ambitions into the field of drug discovery. According to two sources familiar with the matter, as public fears about AI continue to spread, the startup has moved its biological research beyond "in silico" simulations and onto real‑world laboratory workbenches.
Eric Kauderer-Abrams, head of life sciences at Anthropic, confirmed the existence of a wet lab in an interview this Tuesday (September 15). "We believe that, when it comes to biological research, the ultimate test has been—and will continue to be for some time—work conducted in real‑world laboratories," he said. He added that the company carries out part of its work in its own facilities and collaborates with external partners, a model typical of most biotech firms. A spokesperson later clarified that the lab is not dedicated exclusively to drug discovery.
From "Computer Simulation" to a Real-World Test Bench
Building a wet lab is just a small step in a much larger ambition. According to a source familiar with the matter, Anthropic aims to tackle diseases that it believes the pharmaceutical industry has been overlooking—and to do so before employees and the public lose confidence in whether AI is worth risking jobs, or even human lives. Yet drug development inherently carries no guarantee of success; the vast majority of candidate compounds fail to pass clinical trials for safety and efficacy.
For CEO Dario Amodei, this matter carries a personal dimension as well. In a recent article, he wrote that a disease took his father's life, yet just a few years later, a cure for that very condition emerged. Kordel‑Abrams echoed the same sentiment: "Our mission is to develop powerful AI in ways that benefit the world. The greatest opportunities we see lie in the life sciences, and that drives everything we do."
From selling tools to getting involved ourselves
Anthropic's intention to build an in-house drug pipeline was made public as early as June. At an event in San Francisco, Kordel‑Abrams stated on stage that Anthropic would conduct preclinical research in areas that traditional pharmaceutical companies deem financially unattractive.
To achieve this goal, the company has laid a series of groundwork: it launched a research platform called Claude Science, appointed Novartis CEO Vas Narasimhan to its Board of Directors, and acquired the startup Coefficient Bio for approximately $400 million in stock. Anthropic confirmed the acquisition but declined to comment on the deal's price.
Today, the company is bolstering its in-house laboratory capabilities to prioritize speed and first‑hand biological expertise, while continuing to outsource more efficient processes. "There are tasks we can handle ourselves much more quickly," says Kauder‑Abrams, adding that the goal is "to operate at the largest possible scale." Hiring activity underscores this strategy: one LinkedIn post seeks a Head of Procurement and Operations, another recruits biochemistry experts specializing in areas such as protein and nucleic acid characterization, and a third announcement states, "Our aim is to accelerate advances in the life sciences by an order of magnitude." He notes that life sciences have become one of Anthropic's biggest investment priorities.
Living with the fear of a "human extinction" catastrophe
Like most biotech companies, Anthropic is embracing physical automation. According to the aforementioned insiders, the company aims to advance Claude's ability to command robotic units and carry out scientific experiments with minimal human intervention. However, a spokesperson stated that Anthropic believes human oversight and involvement are essential for safety.
"When it comes to using AI to automate laboratory workflows, we're still in a very early stage," said Caudle‑Abrams. "And this field has the potential to deliver substantial speedups across a wide range of processes."
This race to the finish line comes at an exceptionally unusual moment for the company. Over the past two weeks, some of its researchers have warned that AI could spell humanity's extinction; Anthropic has also claimed to have uncovered instances in which its systems might be repurposed to develop biological weapons, a claim that has drawn considerable skepticism. Meanwhile, on September 12, Amodei issued a statement calling for a slowdown in development, arguing that a future AI—more powerful than the OpenAI agent that breached Hugging Face this July—could inflict catastrophic harm. At the same time, Anthropic is preparing for an IPO that could value the company at as much as $2 trillion.
"In our life sciences work, we are constantly striving to strike a balance," said Caudle‑Abrams. "We aim to refine our models in ways that we believe will enhance and accelerate drug discovery, while deploying them with care and prudence to mitigate risks."
This tension is far from mere rhetoric. On September 10, Anthropic released a threat intelligence report stating that between December 2025 and August 2026, five instances of research activities "potentially supporting the development of biological weapons" were identified. The company also emphasized that "information capable of being used to develop biological weapons can likewise be employed to create vaccines or therapies for certain diseases."
Targeting "undruggable" targets
He believes that AI can accelerate the development of therapies for diseases previously deemed "undruggable," such as speeding up the discovery of bispecific and trispecific antibodies—complex molecules that can engage multiple sites on a target protein or cell and simultaneously hit several targets.
However, it remains unclear which specific diseases Anthropic is targeting or what progress has been made. Even after a molecule is identified, it typically requires years of clinical trials before it can reach the market—a challenge that Anthropic has yet to tackle. Its competitor, Isomorphic Labs under Alphabet, has already invested heavily in this area and has pushed its clinical milestone from late 2025 to late 2026. According to an analysis by the tech media outlet ai2.work, the company has raised roughly $2.7 billion in funding, yet as of mid-2026, not a single patient had received its experimental treatment.
Robots, Hardware Standards, and Pharmaceutical Clients
In the automation domain, Anthropic launched a research preview of its "Model Hardware Standard" (MHS) on August 27. According to the company's official blog, this standard connects devices such as microscopes, liquid handlers, and robotic arms to AI agents, reducing hardware integration time from weeks or months to "hours or even minutes." Early testing partners include Genentech (which integrated liquid handlers, robotic arms, and plate readers for protein assays), Carnegie Mellon University (with an integration time of about 8 hours and experiment run times roughly three times faster), and the quantum computing company QuEra (achieving a laser-locking autonomous recovery rate of 99.3%); vendors such as Amazon Web Services, Danaher, and QIAGEN have already joined in supporting the initiative.
Anthropic is providing these kinds of results, along with broader AI services, to leading pharmaceutical companies, including Genentech under Roche, Bristol-Myers Squibb (BMY.US), and Novo-Nordisk A/S (NVO.US). This partnership has just deepened further: on September 16, Novo-Nordisk announced a collaboration with Anthropic to leverage its cutting-edge models and Claude Science in support of scientific reasoning currently under development. Amodi said, "As AI capabilities continue to improve, it may be possible to compress biological and medical breakthroughs that once took a century into just a decade."
An inescapable issue of trust
Meanwhile, Anthropic is grappling with a trust issue, according to the two sources familiar with the matter. Its clients may worry that Anthropic could glean insights from their competing drug pipelines, even though the AI startup keeps client data isolated from its own view.
Kodell‑Abrams stated that, for competitive reasons, Anthropic has drawn a line: it is not conducting clinical trials at this time and is focusing on needs that the industry is not addressing. "We are not competing with pharmaceutical and biotech companies whose business model revolves around bringing drugs to market," he said.
Track Background
This strategic positioning is set against the backdrop of a rapidly rising sector. According to the industry report "AI‑Driven Drug Discovery Enters the Asset‑Monetization Era," since 2019, approximately US$60 billion has been invested in AI‑powered drug discovery, and around 175 AI‑led early‑stage R&D programs have advanced to human clinical trials—yet to date, not a single candidate has received regulatory approval for market launch.
Converging validation signals are emerging on the industry front: According to reports, Ren Feng, Co-CEO and Chief Scientific Officer of Insilico Medicine (03696), stated that in early September, the company's investigational drug Rentosertib completed first‑patient dosing in Phase III clinical trials at Peking Union Medical College Hospital and Shanghai Pulmonary Hospital—marking the world's first candidate drug whose target identification and molecular design were both driven by AI, targeting idiopathic pulmonary fibrosis, with top-line data expected in 2029. The synthesis of this molecule involved 78 compounds, taking 1.5 years and $2.6 million, whereas a conventional approach would have required 4.5 years and tens of millions to hundreds of millions of dollars. Meanwhile, policy support intensified on the same day: on September 18, ten ministries, including the Ministry of Industry and Information Technology, issued the "15th Five-Year Plan for the Development of the Pharmaceutical Industry," which, for the first time, designated AI‑driven drug discovery as a key priority within the sector's five-year plan.
Anthropic's own capital market activity is also accelerating. According to reports, the company plans to raise up to $100 billion through an IPO, targeting a valuation of approximately $2 trillion, while NVIDIA is in talks to invest up to $10 billion as a cornerstone investor. The company secretly filed its S-1 document in June and has chosen Nasdaq. Anthropic told investors that its second-quarter revenue exceeded $11.5 billion, with an annualized revenue run rate surpassing $65 billion.
Institutional Perspective
Domestic securities firms have come to view the entry of tech giants as a key variable in AI‑driven drug discovery. In its pharmaceutical industry weekly report on September 14, Dongwu Securities noted that AI‑based drug development is entering a "strategic entry phase," with a dry–wet closed loop serving as the core logic of AI‑driven drug discovery (AIDD); wet‑lab experiments remain indispensable. Upstream stages—including gene synthesis, protein expression and purification, in vitro validation, and the use of model organisms—will systematically benefit from the surge in AI‑generated candidate molecules. Meanwhile, Bank of China Securities frames AIDD as a closed loop of "design–build–test–learn," emphasizing that experimental validation is a critical step for unlocking long-term value.
Judgments from overseas are far more cautious. Frank von Delft, a professor of structural biology at the University of Oxford, notes that AI models "are still nowhere near rendering experiments unnecessary"—candidate drugs will continue to require costly real-world trials to confirm efficacy and safety. Matthew Todd, a professor at University College London, estimates that any returns from Anthropic's proprietary pipeline are at least a decade away.