From OpenAI’s offices to a deal with Eli Lilly — how Chai Discovery became one of the flashiest names in AI drug development

Drug discovery, the art of identifying new molecules to develop pharmaceuticals, is a notoriously time-consuming and difficult process. Traditional techniques, like high-throughput screening, offer an expensive scattershot approach — one that is not often successful. However, a new breed of biotech companies are leveraging AI and advanced data technologies in an attempt to accelerate and streamline the process.

Chai Discovery, an AI startup founded in 2024, is one such company. In a little over 12 months, its young co-founders have managed to raise hundreds of millions of dollars and rally the backing of some of Silicon Valley’s most influential investors, making it one of the flashiest firms in a growing industry. In December, the company completed its Series B, bringing in an additional $130 million and a valuation of $1.3 billion.

Last Friday, Chai also announced a partnership with Eli Lilly, a deal in which the pharmaceutical giant will use the startup’s software to help develop new medicines. Chai’s algorithm, called Chai-2, is designed to develop antibodies — the proteins necessary to fight illnesses. The startup has said it hopes to serve as a kind of “computer-aided design suite” for molecules.

It’s a critical moment for Chai’s particular field. The startup’s deal was announced shortly before Eli Lilly said it would also collaborate with Nvidia on a $1 billion partnership to create an AI drug discovery lab in San Francisco. This “co-innovation lab,” as it’s being called, will combine big data, compute resources, and scientific expertise, all in an attempt to accelerate the speed of new medicine development.

The industry isn’t without its detractors. Some industry veterans seem to feel that — given how difficult traditional drug development is — these new technologies are unlikely to have a major impact. However, for every naysayer, there seem to be just as many believers.

Elena Viboch, managing director at General Catalyst — one of Chai’s major backers — told TechCrunch that her firm is confident that companies that adopt the startup’s services will see results. “We believe the biopharma companies that move the most quickly to partner with companies like Chai will be the first to get molecules into the clinic, and will make medicines that matter,” Viboch said. “In practice that means partnering in 2026 and by the end of 2027 seeing first-in-class medicines enter into clinical trials.”

Aliza Apple, the head of Lilly’s TuneLab program — which uses AI and machine learning to advance drug discovery — also expressed confidence in Chai’s product. “By combining Chai’s generative design models with Lilly’s deep biologics expertise and proprietary data, we intend to push the frontier of how AI can design better molecules from the outset, with the ultimate goal to help accelerate the development of innovative medicines for patients,” she said.

Chai may have been founded less than two years ago, but the startup’s origins began around six years ago, amid conversations between its co-founders and OpenAI CEO Sam Altman. One of those founders, Josh Meier, previously worked for OpenAI in 2018 on its research and engineering team. After he left the company, Altman messaged Meier’s old college friend, Jack Dent, to ask about a potential business opportunity. Meier and Dent had originally met in computer science classes at Harvard but, at the time, Dent was a Stripe engineer (another company Altman was an early backer of). Altman asked him if he thought Meier would be open to collaborating on a proteomics startup — that is, a company focused on the study of proteins.

Altman “messaged me to say that everyone at OpenAI thought highly of him and asked if I thought he’d be open to working with them on a proteomics spinout,” Dent said. Dent told Altman “of course,” but there was just one hitch: Meier didn’t feel like the technology was quite “there” yet. The AI tech behind such firms — which leverage powerful algorithms — was still a growing field and far from where it needed to be.

Meier was also pretty dead set on joining Facebook’s research and engineering team, which is what he would go on to do. At Facebook, Meier helped to develop ESM1, the first transformer protein-language model—an important precursor to the work Chai is currently doing. After Meier’s time at Facebook, he would spend three years at Absci, another AI biotech firm based around drug creation.

By 2024, Meier and Dent finally felt prepared to tackle the proteomics company they had originally discussed with Altman. “Josh and I reached back out to Sam and told him we should pick up that conversation where we left off—and that we were starting Chai together,” Dent said.

OpenAI ended up becoming one of Chai’s first seed investors. Meier and Dent actually founded Chai — along with their co-founders, Matthew McPartlon and Jacques Boitreaud — while working out of the AI giant’s offices in San Francisco’s Mission neighborhood. “They were kind enough to give us some office space,” Dent revealed.

Now, a little over a year later, as Chai basks in the glow of its newfound partnership with Eli Lilly, Dent says that the key to the company’s fast growth has been assembling a team of hugely talented people. “We really just put our heads down and pushed the frontier of what these models are capable of,” said Dent. “Every line of code in our codebase is homegrown. We’re not taking LLMs off the shelf that are in the open source [ecosystem] and fine-tuning them. These are highly custom architectures.”

General Catalyst’s Viboch told TechCrunch that she felt Chai was ready to hit the ground running. “There are no fundamental barriers to deployment of these models in drug discovery,” she said. “Companies will still need to take drug candidates through testing and clinical trials, but we believe there’ll be significant advantages to those who adopt these technologies — not just in compressing discovery timelines, but also in unlocking classes of medicines that have historically been difficult to develop.”