When Benjamine Liu was a young computational biologist working on his doctorate at the University of Oxford, he had some ideas for novel drugs that could treat Alzheimer’s. He was so excited by their promise that he tried shopping them around to a few pharmaceutical companies. Not a single one was interested.

“They said, ‘We have more drugs than we can afford to develop,’” Liu tells Forbes. “A discovered drug isn’t worth that much.”

That rejection led him to a realization: The biggest problem in drug R&D wasn’t in the sexy part of searching for new discoveries. Instead, it was the long, grueling, expensive process of clinical development, where most potential drugs fail.

In fact, while there was a nearly twofold increase in drug candidates in the last decade, the number of drugs approved by the FDA, at around 50 per year, hasn’t changed much, he says now. And that, Liu figured, is something that AI is uniquely suited to help.

It’s a contrarian take at a time when AI is being touted as a godsend for faster and cheaper drug discovery that will usher in a golden age of new therapies. But that’s a long game. And Liu is convinced the bigger opportunity is in fixing clinical trials. “What we think the world has wrong is that drug discovery is the bottleneck, and it has not been the bottleneck for a long time,” he says.

In 2016, Liu, now 36, teamed up with Linhao Zhang, 34, a computer scientist who had worked on the engineering team at Oscar Health, to start Formation Bio to help pharma companies do their clinical trials better and faster. Originally called Trialspark, it was set up as a service business: drug companies and biotechs would hire Liu and Zhang to run their trials for them. But Liu is now chasing a bigger goal: Buying a portfolio of 10 early-stage drug candidates, many of which failed or stalled out in early-stage clinical trials, and then using AI to help get them back on track.

Today, New York City-based Formation Bio has some of the world’s top investors betting on Liu’s vision, including Andreessen Horowitz, Sequoia, Thrive Capital, Kleiner Perkins’ chairman John Doerr and OpenAI’s Sam Altman. It has raised a total of $615 million at a $1.8 billion valuation. Forbes estimates that Liu’s stake is worth more than $150 million, while Zhang’s is worth above $100 million.

“It has the potential—the potential—of being one of these enormously significant companies in an industry whose apple cart has not been upset all that much by a young company started in the last 10 or 15 years,” says Michael Moritz, the billionaire venture capitalist and former Sequoia chairman. Way back in 2016, Liu cold-emailed Moritz about his idea of creating a next-generation pharma company with the help of AI. Moritz, who Liu says responded “within like 28 minutes,” personally wrote the first $2.25 million check in the business.

There are a lot of would-be drugs for sale these days. That’s because many biotech firms have been left without the cash to develop them as the spigot of funding has dried up. Big Pharma’s priorities are constantly shifting, leading it to cut back on spending for some drugs in favor of others. Meanwhile, China has become a drug innovation powerhouse, with more therapies available to license.

Formation Bio is targeting drugs that have yet to go through phase 2 trials. For these early-stage therapies, there’s more risk, because only 30% succeed at this stage. But there’s also more possible payoff. Liu figures that Formation can work its magic with AI. With the technology they’ve finetuned over the past decade, he says, they can do those trials up to 50% faster, a big advantage when every day of delay can be worth millions.

“Drugs aren’t worth much until post-phase 2, and there are a lot of post-phase 1 drugs worth betting on,” Liu says. “You can play in phase 3, but no one’s trying to give away a great phase 3 drug.”

"What we think the world has wrong is that drug discovery is the bottleneck, and it has not been the bottleneck for a long time."

To increase the odds of success, Liu has assembled a team of top drug pickers, led by Mikael Dolsten, who retired from Pfizer as president of worldwide R&D, to scour the world for the best candidates. The human drug pickers will get AI tools to help them make better bets on which drugs to buy. Formation’s software can pick up early trial data in non-English publications, for example, to try and find overlooked drugs in China. It can quickly rank and rerank various options based on their highest medical impact or biggest commercial opportunity. Then, Formation Bio’s team will use AI to deal with administrative tasks, like patient recruitment and regulatory filings, as well as the thorny issue of matching its drug candidates to the right patients.

So far, the company has purchased or licensed five potential drugs, including a treatment for knee osteoarthritis and a therapy for ulcerative colitis, both of which are currently in clinical trials. While none of its drugs are on the market yet, it licensed one, for chronic hand eczema, last June to French pharma giant Sanofi for some $630 million (at current exchange rates), plus future royalties.

Liu’s strategy of rescuing forgotten drugs echoes what billionaire biotech entrepreneur Vivek Ramaswamy did with Roivant Sciences, which today has a market cap of $21 billion and a broad pipeline of drugs. “It’s not like we’re the first ones to do it. We’ve studied PureTech, Roivant, BridgeBio—there are all these different iterations,” Liu says. But Roivant was founded before AI swept the industry. Formation’s differentiator, he argues, is both in how it’s set up clinical trials in house and how it’s using AI to improve the chance of success.

“I think drug discovery is going to be commoditized with China and AI,” Liu says. “If we don’t solve this drug development bottleneck, we’re actually not going to end up with more medicines for patients.”

Liu immigrated with his family from Taiwan when he was 2, and grew up in Thousand Oaks, California. After graduating from Yale, he earned his master’s from the University of Cambridge in computational biology, then became a Rhodes Scholar. For his Oxford thesis, he used AI and big data to develop diagnostics and therapeutics for Parkinson’s and Alzheimer’s disease.

Even before Liu finished his dissertation, he was working on the idea that would become Formation Bio. “There is no fakeness in what Ben wants to do,” cofounder Zhang says. “He’s not trying to start a startup because his friends are doing it or because it’s cool or because he wants to make a brand of himself.”

While drug discovery makes splashy headlines, drug development is “the big unsolved challenge for the industry,” says Paul Hudson, former CEO of Sanofi, which made an investment in Formation Bio and set up collaborations with it during his tenure. The time and cost of trials is a huge part of that, and is especially important because of the ticking time bomb of drug patents. “There’s a lot of dead time in a clinical trial,” explains Pratap Khedkar, CEO of life sciences consulting firm ZS of the industry generally. “How do I cut half the time and half the costs and more importantly increase the probability that it actually crosses?”

"It has the potential—the potential—of being one of these enormously significant companies in an industry whose apple cart has not been upset all that much by a young company started in the last 10 or 15 years."

During the early days of the business, when they were still focused on running clinical trials for others, Liu and Zhang were in a hurry to buy their own drugs. But their investors warned that it was too much too soon. Instead, they worked with former Novartis CEO Joe Jimenez to set up a new venture firm called Aditum Bio in 2019 as something of a test. Like Formation does now, Aditum bought drugs and set them up under separate companies, in which Formation was a minority investor. To date, they had one big win, for a drug that grew lean muscle mass: Eli Lilly bought the Aditum company that owned it for up to $1.9 billion in 2023. “We kind of joke, if we started this strategy fully on balance sheet, it would have been an even bigger win,” Liu says.

Five years after starting the company, in 2021, Liu and Zhang finally felt they had enough experience to build their own portfolio of drugs. “I’m a very Bayesian human and a very probability driven human,” Liu says. “When we think about the company strategy, we try to be really conservative, assume industry averages, and make sure we have a competitive advantage around cheaper, faster trials.”

In January 2022, Formation Bio licensed a possible therapy to promote cartilage growth in the knees of patients with osteoarthritis, which more than 230 million people suffer from globally. Originally developed by Darmstadt, Germany-based Merck KGaA, it had already gone through three clinical trials in more than 800 patients. Formation Bio paid an undisclosed upfront amount for the drug; it holds a majority stake in the new company set up to house it, while Merck KGaA retains a minority piece.

The clinical trial data showed that injections of the drug do increase cartilage growth, but it remained unclear if they reduced pain overall more than a placebo or prevented or delayed the need for a knee replacement. To try to answer that question, Formation built an AI-based prediction model trained on 23,000 patients and 48,000 MRIs to determine who might be at risk of a knee replacement and whether thicker cartilage helped protect the joint. Liu says its analysis found that patients who took the drug and increased their cartilage growth had a lower five-year risk of needing a knee replacement. Without AI, drug picker Dolsten says, “I probably would have abandoned it.” The drug is now in late-stage clinical trials.

No matter how good its process is, not all of the drugs it buys will succeed. Liu is well aware of the failure rates, so his goal is to play the odds, not win every time. “We hope we hit three out of 10, but even if we hit one out of 10 we can still double our money,” he says.

More from Forbes

ForbesThis Serial Entrepreneur Wants The FDA To Approve His AI DoctorForbesRoche Bought Thousands Of Nvidia AI Chips To Speed Up Drug DevelopmentBy Amy FeldmanForbesHow Lilly Used AI To Crank Up Production Of Its Popular GLP-1sBy Amy FeldmanForbesThis Startup Wants To Use Mini Robots To Treat Alzheimer'sBy Amy FeldmanForbesGene Editing Has Struggled To Go Commercial. This Nobel Laureate Has A $1 Billion Plan To Fix That.By Amy Feldman