Chai Discovery’s AI model draws 20 pharma partners
Last June, a 30-year-old founder named Josh Meier sat in San Francisco watching his phone light up. It was 2 a.m. His cofounder Jack Dent was seeing the same thing on his screen. Messages poured in from strangers. They came from scientists at pharmaceutical companies. They came from researchers in Europe. They came from people who said they could not sleep because of what they had just seen.
What they had seen was a computer model that could design antibodies. The model was called Chai-2. It was built by a company that had existed for only 15 months. The company was called Chai Discovery. Before that night, almost nobody outside of a small circle of AI researchers knew it existed. After that night, nearly 20 pharmaceutical companies wanted to talk.
“It was like we dropped a bomb on the field,” Jack Dent says now.
The Human Story: Two Friends Who Started Early
Josh Meier grew up in Teaneck, New Jersey. His family was full of doctors. He watched them work long hours. But when he was 8 years old, he discovered something else. He learned to program. He wrote his first lines of code. The computer did what he told it to do. That feeling never left him.
He attended Harvard. For a while, he thought about becoming a physician. But he also loved chemistry and computer science. He majored in both.
“I love programming because you can scale your impact,” he says. A doctor helps one patient at a time. A program can help millions.
On his first day of classes at Harvard, he met Jack Dent. Jack was from London. He had been building apps since he was a teenager. At 14, he was selling apps for 99 cents each. He thought he was set for life. He was not set for life. But he learned how to build things that people wanted to use.
They both graduated in 2018 with bachelor’s and master’s degrees. Then they went their separate ways. Josh worked at OpenAI, Meta’s generative biology group, and a company called Absci. Jack went to Stripe and became one of the company’s top engineers.
Every few months, they met. Sometimes at a Portuguese restaurant in San Francisco. Sometimes at an ice cream spot in New York. They compared notes. They watched AI explode around them.
The Scientific Story: What Was Happening in the Background
In 2021, DeepMind released AlphaFold publicly. It could predict the shape of proteins—the machines that run our bodies. For 50 years, scientists struggled to determine protein shapes, a process that took months or years per protein. AlphaFold could do it in minutes, and the team behind it — Demis Hassabis, John Jumper, and protein design pioneer David Baker — shared the 2024 Nobel Prize in Chemistry.
By 2024, something had shifted. AI models were getting better every month. But the protein discovery field was lagging behind.
“We had this sense that everything in AI was about to start working big time,” Dent says. “And the protein discovery field was lagging by a few years.”
They launched Chai Discovery in March 2024. They brought two more cofounders. Matt McPartlon had also worked at Absci. Jacques Boitreaud came from a French AI drug discovery company called Aqemia. Together, they believed they could build something better.
“Humans are just very bad at drug discovery,” Meier says. “It’s honestly miraculous we can make drugs at all with the tools available today.”
Drug discovery is one of the great promises of AI. Today, a single drug typically costs well over $1 billion and takes more than 10 years. The dream is for AI to let drug hunters find potential therapies faster, with more precision, and to come up with treatments for diseases that had been considered undruggable.
The First Model That Changed Everything
They launched their first model, Chai-1, in just a few months. It could predict the shape of proteins. But the team recognized it was insufficient for real-world drug discovery.
“The models were smoke and mirrors for a long time,” Dent says. “We knew we had to make things literally 100 times better for it to be valuable for real drug discovery programs.”
They made it free. Any pharmaceutical company could test it.
Then came Chai-2. This was the model that dropped the bomb. This was the model that could design antibodies.
Antibodies are Y-shaped proteins that our immune system uses to fight invaders. They are also one of the most important types of modern drugs. Some of the best-selling medicines in the world are antibodies. They treat cancer, autoimmune diseases, and infections.
Designing antibodies by hand is incredibly difficult. It is like trying to find a single key that fits a lock you cannot see. The lock is a protein on a disease cell. The key is an antibody that binds to it. There are billions of possible keys. Most of them do not work. Finding the right one takes years.
Chai-2 could find promising keys in days.
The Bet That Paid Off
Most AI drug discovery companies develop their own drugs and build their own pipelines, hoping for a blockbuster. Chai took a different approach. They decided to sell access to their technology.
“When we started, people told us the only way to make money is to make your own assets and become a drug company,” Dent says. “That’s the dogma we had to challenge.”
Drug companies spend hundreds of millions of dollars for one promising molecule. If you can build a software engine that can spin up scores of potential therapies quickly, that engine is extremely valuable.
Mikael Dolsten understands this. He retired from Pfizer as president of worldwide R&D and now sits on Chai’s board of directors.
“I think it was a brilliant insight,” Dolsten says. “If you want to be the trusted one that traditional industries feel comfortable teaming up with, you cannot at the same time try to have your own little shop.”
What Chai-3 Could Do
Earlier this year, Chai quietly deployed the next iteration of its antibody design model. They called it Chai-3.
“That got the Pfizer team really excited,” Dent says.
The model works by predicting how proteins interact. Not just what they look like, but how they fit together. This is a much harder problem. Two proteins can look like they should fit, but they do not. The model has to understand the physics, chemistry, and biology of the interaction.
In January 2025, the startup announced a deal with Eli Lilly. The drug behemoth, best known for its weight-loss drugs and with a market cap of $1 trillion, agreed to design multiple novel therapeutics with Chai’s AI model.
Then came the Pfizer deal. The company is now in talks with more than 15 additional pharmaceutical companies — which, together with Lilly and Pfizer, brings the total to nearly 20 prospective partners.
The Money Behind the Science
Investors have taken notice. Chai has raised more than $225 million from investors including OpenAI, General Catalyst, Menlo Ventures, and Oak HC/FT. Now the company is in talks to raise an additional $400 million at a valuation of $3.4 billion.
Annie Lamont is a managing partner of Oak HC/FT and a Forbes Midas List investor.
“I think it’s become very clear that they are winning the war,” she says. “They are winning the commercialization war, and they are winning the model and product war.”
The numbers back her up. Investors poured $11.4 billion into AI drug discovery companies globally in 2025. That is more than double the $5.6 billion of the previous year.
“There has been all this hope and expectation around AI in drug discovery,” Dent says. “People have become a little jaded because it was hard to put your finger on anything tangible. But we are in a completely different universe than we were a year ago.”
The Timeline to Reality
Diogo Rau, chief information and digital officer at Eli Lilly, told Forbes in March 2025 that, given regulatory timetables, it would be “mid-2030s, if not late-2030s” before any of its AI-developed medicines are on the market.
“It’s a big bet on the future,” he said.
This is the reality of drug development: even with the best AI, it takes years to test a drug for safety, efficacy, and regulatory approval. But the AI changes what happens at the front end. It changes how quickly you find the right molecule, how many candidates you can test, and the probability that any given candidate will succeed.
Through its collaboration with Eli Lilly, Chai is working on accelerating the development of biologic drugs — therapies derived from natural sources such as proteins or cells. Biologics are more complex and harder to design, but they can be more effective.
The Open Questions: What Comes Next
The models are getting better. The partnerships are growing. The money is flowing. But there are still open questions. How good can the models get? How many diseases can they help treat? How fast can they bring drugs to market?
The company is already working on the next iteration. They are improving the model, adding new capabilities, and expanding into new areas.
“We want to raise the bar of medicines that are created,” Meier says.