Chai Discovery, a San Francisco startup building artificial-intelligence models that predict and design how molecules interact, has raised a $400 million Series C round at a $3.8 billion valuation. The financing, disclosed on July 13, was led by Index Ventures with participation from Kleiner Perkins, Sequoia Capital and Dimension, alongside a long list of new and returning backers that includes Bain Capital Ventures, Battery Ventures, Baillie Gifford, Sapphire Ventures, Thrive Capital, Menlo Ventures, General Catalyst and OpenAI.
It is a striking figure for a company founded only in 2024, and it caps an unusually fast fundraising cadence: this is Chai's third round in under a year, lifting total capital raised past $600 million. The company started with a $30 million seed and the launch of its first model, followed by a $70 million Series A in the summer of 2025 and a $130 million Series B in December. Each raise has arrived alongside a new model and, increasingly, signed contracts with large pharmaceutical companies.
The pitch is deceptively simple. Rather than building a general-purpose chatbot, Chai is training foundation models aimed squarely at biology—systems that can predict the three-dimensional structure of proteins and other molecules and then generate new molecules to order. Its Chai-1 model, released in 2024, drew attention for matching or beating Google DeepMind's AlphaFold on certain structure-prediction tasks. Chai-2, unveiled in 2025, pushed into generative territory, and a newer Chai-3 has continued to sharpen the numbers.
Why It Matters
The round is less interesting as a single deal than as a data point about where AI money is now flowing. For two years the largest checks went almost entirely to horizontal model builders—the labs training ever-larger general systems. Chai's raise is part of a visible rotation toward vertical or applied foundation models: companies that take the same architectural ideas and point them at one hard, valuable domain where proprietary data and measurable results create a moat.
Drug discovery is a natural early proving ground. The economics of pharmaceutical research are brutal, with most candidates failing and each approved medicine carrying the cost of the many that did not. If an AI model can raise the hit rate at the earliest design stage, the value created is enormous and, crucially, quantifiable. That is a very different proposition from consumer AI, where usefulness is real but hard to price. Investors are effectively betting that domain-specific models with a scientific feedback loop can compound faster than general ones.
There is also a competitive subtext. AlphaFold made protein-structure prediction a public benchmark and, through DeepMind's Isomorphic Labs, a commercial ambition. A well-funded independent challenger changes the shape of that market, and the willingness of tier-one venture firms to underwrite it at nearly $4 billion suggests they see room for more than one winner in AI-for-science.
The Reaction
What separated this round from a typical AI funding announcement was the emphasis on customers rather than benchmarks. Partners at Index Ventures, Kleiner Perkins and Sequoia Capital each pointed to real-world use of Chai's models by established drugmakers as the reason for writing checks. Over the course of the year the startup has signed agreements with Eli Lilly, Pfizer and Novartis—names that carry weight precisely because they are conservative buyers with their own substantial research operations.
Chief executive Joshua Meier framed the milestone as a shift in the field's maturity, saying AI drug discovery had moved "from promise to deployment." The distinction matters. A great deal of AI-for-biology work has lived in demos and leaderboard scores; contracts with companies that could build the technology themselves are a harder signal. Chai has also tried to back its marketing with published preprints, including results reporting a double-digit success rate in designing antibodies from scratch—an outcome the company argues is orders of magnitude better than earlier computational methods and good enough to skip some laboratory screening steps.
What Comes Next
Chai has said relatively little about how it will spend the money, limiting public comment to a desire to "further accelerate progress." In practice, a raise of this size buys three things: compute to train larger and more specialized models, scientific and engineering talent in a fiercely competitive market, and the runway to move from designing molecules on a screen toward validating them in the lab and, eventually, the clinic.
The open question is how far a software company can travel down that path before it must either become a drug developer itself or remain a tools-and-models supplier to the industry. The current strategy—licensing models and partnering with pharmaceutical companies—keeps Chai asset-light and lets its customers carry the enormous cost and risk of clinical trials. But the biggest returns in biotech accrue to whoever owns the resulting medicines, and a $3.8 billion valuation implies expectations that will eventually test that boundary. Rivals from DeepMind's Isomorphic Labs to a growing field of AI-native biotech startups are wrestling with the same choice.
Closing Thoughts
Stripped of the specifics, Chai's Series C is a small window into how the AI industry is changing. The first phase of the boom rewarded scale and generality; the emerging phase is beginning to reward depth—models that know one domain extremely well and can prove their value in outcomes rather than conversation. Whether biology turns out to be the template for that shift or simply its earliest example, the pattern is worth watching.
For now, the more grounded takeaway is that capital, scientific talent and industrial demand are converging on the idea that AI can help design better molecules. That convergence does not guarantee new medicines—the history of computational drug discovery is littered with premature optimism—but it does mean the next few years will offer an unusually public test of whether these models deliver in the one arena where the results are measured in real-world success and failure rather than benchmarks.
한글 요약
샌프란시스코의 AI 스타트업 차이 디스커버리(Chai Discovery)가 7월 13일 4억 달러 규모의 시리즈 C 투자를 유치하며 기업가치 38억 달러를 인정받았다. 인덱스 벤처스가 주도하고 클라이너 퍼킨스, 세쿼이아 캐피털, 디멘션 등이 참여했다. 2024년 설립된 이 회사는 1년도 안 되는 기간에 세 번째 라운드를 마감하며 누적 조달액이 6억 달러를 넘어섰다. 분자 간 상호작용을 예측·설계하는 AI 파운데이션 모델을 개발하며, 대표 모델 차이-1은 구글 딥마인드의 알파폴드와 견줄 만한 구조 예측 성능으로 주목받았다.
이번 투자의 의미는 단일 딜을 넘어 AI 자금의 흐름 변화를 보여준다는 데 있다. 그동안 가장 큰 투자는 범용 대형 모델을 만드는 곳에 집중됐지만, 이제는 하나의 어려운 분야에 특화한 '수직형 파운데이션 모델'로 무게중심이 옮겨가고 있다. 신약 개발은 성공률을 조금만 높여도 가치가 크고 측정 가능하다는 점에서 이런 특화 모델의 초기 시험대가 되고 있다. 실제로 인덱스·클라이너·세쿼이아는 일라이 릴리, 화이자, 노바티스 등 대형 제약사가 차이의 모델을 실사용한다는 점을 투자 근거로 들었다.
조슈아 마이어 CEO는 AI 신약 개발이 "약속에서 실전 배치로" 넘어갔다고 표현했다. 다만 소프트웨어 회사가 분자 설계를 넘어 실험실 검증과 임상까지 얼마나 나아갈 수 있을지는 열린 질문이다. 현재의 모델 라이선싱·제휴 전략은 자산을 가볍게 유지시키지만, 바이오텍의 가장 큰 수익은 결국 신약을 소유한 쪽에 돌아간다. 앞으로 몇 년은 이런 특화 AI 모델이 벤치마크 점수가 아니라 실제 성패로 가치를 증명할 수 있는지를 가늠하는 공개 시험 무대가 될 전망이다.
참고: FierceBiotech · TechCrunch · BusinessWire