AI Drug Discovery Companies: The Leading Private Startups in 2026

Developing a new medicine usually takes more than a decade and costs well over a billion dollars once failures are counted. Most of that time and money goes into finding molecules that work, proving they are safe and running clinical trials. Artificial intelligence promises to compress the early part of this journey by predicting how molecules behave and designing new ones on a computer. Since AlphaFold showed that AI could predict protein structures with remarkable accuracy, investors have poured billions into AI drug discovery companies. This guide covers the leading private companies, how they differ, how they make money and the risks that remain.

Every company below links to a full profile with its sourced valuation, funding history and investors. Several of the most prominent AI drug discovery companies have not disclosed a valuation, so this list is ordered by a mix of disclosed valuation and funding scale rather than a strict ranking. We focus on private companies; public AI drug discovery companies such as Recursion are mentioned for context only.

The leading companies

Why AI drug discovery attracts so much capital

Drug discovery has long been a game of searching enormous spaces. There are more possible drug-like molecules than there are atoms in the solar system, and biology is complex enough that even well-understood targets can surprise researchers. Traditional discovery relies on screening libraries of compounds, making small chemical changes and running many rounds of laboratory experiments. Each cycle takes weeks or months. AI offers a way to explore this space faster, by predicting which molecules are likely to bind a target, how they will behave in the body and which ones are worth making. Even modest improvements in the success rate of early research can be worth billions, because most of the cost of a successful drug is the cost of all the failures along the way.

The breakthrough that changed investor sentiment was AlphaFold, developed by Google DeepMind, which predicted protein structures at near-experimental accuracy. Its creators shared the 2024 Nobel Prize in Chemistry with David Baker, whose lab pioneered computational protein design. These advances suggested that AI could do more than speed up existing workflows; it could design entirely new molecules, such as antibodies and enzymes, that do not exist in nature.

Isomorphic Labs: Alphabet's drug discovery bet

Isomorphic Labs was spun out of DeepMind in 2021 under Demis Hassabis to apply AlphaFold-style models to drug design. It has signed partnerships with major pharmaceutical companies, including Eli Lilly and Novartis, and raised large outside rounds, including a US$600 million round led by Thrive Capital. Our profile covers later funding and reported valuation talks. Isomorphic's advantage is its connection to some of the world's most advanced AI research and its access to Alphabet's computing resources. Its challenge, like everyone's, is turning better predictions into approved medicines.

Xaira Therapeutics: a billion-dollar launch

Xaira Therapeutics launched in 2024 with more than US$1 billion from ARCH Venture Partners, Foresite and others, one of the largest starting war chests in biotech history. It is led by Marc Tessier-Lavigne, the former president of Stanford University, and draws on protein design methods from David Baker's lab. Xaira aims to build both AI models and an in-house drug development pipeline, which requires far more capital than a software-only approach but captures more value if drugs succeed.

Lila Sciences: autonomous science

Lila Sciences, created by Flagship Pioneering, is pursuing what it calls scientific superintelligence: AI systems that design experiments and run them in automated laboratories, learning from every result. It raised a US$235 million Series A at about US$1.23 billion in September 2025 and an extension including Nvidia that valued it at more than US$1.3 billion. Lila's ambitions extend beyond drugs into chemistry and materials, but life sciences are a central focus.

Chai Discovery: open models and antibody design

Chai Discovery, backed early by OpenAI, released Chai-1, an open model for predicting molecular structures, and Chai-2, a model for designing antibodies from scratch. It raised a US$130 million Series B co-led by Oak HC/FT and General Catalyst at US$1.3 billion in December 2025. Chai's open research built credibility with scientists, and its antibody design results attracted pharmaceutical interest.

Periodic Labs and BenchSci

Periodic Labs, founded by former OpenAI and Google DeepMind researchers, raised a US$300 million seed round to build AI scientists and autonomous labs, initially focused on materials such as superconductors rather than drugs. It shows how the same closed-loop approach can be applied across sciences. BenchSci, based in Toronto, takes a different approach: it uses AI to help pharmaceutical scientists plan experiments by analysing published research and reagent data, selling software to large drug companies rather than developing drugs itself.

Three business models

AI drug discovery companies tend to follow one of three models, or a combination. The first is the platform partnership model: the company licenses its AI or performs discovery work for pharmaceutical partners in exchange for upfront payments, research funding, milestone payments as drugs progress, and royalties if they reach the market. This model brings revenue early and validates the technology, but the biggest rewards go to the partner. The second is the pipeline model: the company uses its AI to discover its own drugs and takes them into clinical trials, keeping more of the potential value but requiring much more capital and accepting clinical risk. The third is the software model: selling tools to pharmaceutical researchers on a subscription basis, which produces steadier revenue but smaller upside. Isomorphic and Chai mainly pursue partnerships today, Xaira emphasises its own pipeline, and BenchSci sells software.

Many companies blend these approaches, using partnerships to fund operations while building a few wholly owned programmes. Investors pay close attention to the terms of partnerships, because large headline numbers often consist mostly of milestones that will only be paid if drugs succeed.

What AI can and cannot do yet

AI is strongest at problems with abundant, high-quality data and clear answers, such as predicting protein structures or binding between molecules. It is weaker at predicting how a drug will behave in the whole human body, including toxicity, metabolism and effectiveness in diverse patients, because data on these outcomes is scarce and noisy. As a result, AI can speed up the discovery and optimisation of drug candidates, but clinical trials remain long, expensive and uncertain. Several AI-discovered drugs have entered trials, and some have failed for the same reasons conventional drugs fail. The industry is still waiting for clear evidence that AI-designed drugs succeed in trials at meaningfully higher rates.

This is why many newer companies emphasise generating their own data through automated laboratories. Better data on how molecules behave in cells and tissues could improve models in areas where public data is lacking.

The role of big tech and big pharma

Technology companies are deeply involved. Alphabet owns Isomorphic and DeepMind continues biological research. Nvidia invests in many AI biotech startups and provides computing platforms and models for biology. OpenAI invested in Chai. Pharmaceutical companies, meanwhile, have built internal AI teams and signed numerous partnerships, often working with several AI companies at once. This creates a competitive market for partnerships but also gives startups many potential customers.

How to evaluate an AI drug discovery company

Investors and partners typically ask several questions. Has the company shown laboratory validation of its predictions across many targets, not just selected examples? Does it have proprietary data, or does it rely on public data that competitors can also use? Are its partnerships substantive, with meaningful upfront payments, or mainly options and milestones? How many programmes has it advanced, and at what speed and cost compared with industry norms? What is its plan to fund clinical development, and how much capital will it need? And does it have the right mix of AI researchers, chemists, biologists and drug developers to turn models into medicines?

Risks

The main risk is that faster discovery does not translate into more approved drugs. If AI-designed molecules fail in trials at similar rates to conventional ones, the economic advantage is smaller than hoped. Capital intensity is another risk, especially for companies building pipelines, as clinical trials can cost hundreds of millions of dollars. Competition is intense, with open models eroding advantages in some areas. Valuations for young companies can be high relative to their progress, and funding conditions for biotech can change quickly. Finally, regulatory and intellectual property questions about AI-generated molecules are still evolving.

What to watch

Watch for clinical trial results from AI-discovered drugs, new pharmaceutical partnerships with large upfront payments, open model releases that change competitive dynamics, and fundraising by companies building their own pipelines. Our guides to healthcare AI startups, new AI labs and health tech unicorns provide related context.

Key takeaways

AI drug discovery has moved from promise to large-scale investment, led by Isomorphic Labs and Xaira, with newer companies such as Lila Sciences and Chai Discovery reaching unicorn valuations quickly. The technology clearly speeds up parts of early discovery, especially protein and antibody design. Whether it improves the odds of clinical success remains the central question, and the answers will come from trials over the next several years.