AI M&A Wave Surges: Recursion and Exscientia Merger Signals New Landscape of Industry Consolidation

Recursion Pharmaceuticals reached a deal worth nearly $700 million with AI biotech company Exscientia last year. According to an EY report, this merger isthe largest AI acquisition in the life sciences sector to date, and "the surge in AI collaborations and acquisitions over the past five years demonstrates the opportunities this technology brings to life sciences companies."

Recursion's Chief Financial Officer and President of its UK division, Ben Taylor, said that successfully integrating these companies into more traditional drug development processes requires creative thinking and a full commitment to future transformation. He compared the current moment to the phase of "biologics growth and maturation" about a decade ago.

"At that time, various strategies emerged in the pharmaceutical industry—some acquired assets through large-scale purchases, some started small and grew gradually, and others did not acquire at all and relied entirely on internal R&D," Taylor said. "The same thing is happening now because most companies have not yet gained substantial validation on AI platforms, and this will become the trigger point for many pharmaceutical companies to start thinking deeply."

Although Exscientia has struggled in recent years due tothe loss of a key partnership with Bayerandleadership changes, Recursion has been building its AI-driven drug discovery pipeline and securing high-profile collaborations. However, Taylor pointed out that the two companies share many similarities, and the combined entity now has 10 programs in development, with potential milestone payments of hundreds of millions of dollars from large pharmaceutical partners such as Sanofi and Roche.

"You cannot solve all the different aspects of creating new molecules with a single algorithm—you need to integrate them."

Ben Taylor
Chief Financial Officer, Recursion Pharmaceuticals

Despite the promising technology of the new Recursion and the support of the world's leading chipmaker NVIDIA, its pipeline remains in early clinical stages. Its most advanced program targets the rare disease cerebral cavernous malformations and has crossed the Phase II clinical threshold, while oncology and other rare disease programs are at earlier stages or in preclinical development.

Although this kind of mega-scale AI merger is still a new concept, Taylor believes that leaders in the life sciences sector will see more such deals in the future, incorporating new technologies and changing the way drugs are made.

Here, Taylor discusses AI's disruptive impact on biopharma, what companies should focus on when exploring collaborations and acquisitions, and the challenges of collecting and using big data to drive further industry transformation.

This interview has been edited for length and style.

PHARMAVOICE: AI is clearly becoming a disruptor in the life sciences—why is this happening now?

BEN TAYLOR:The right way to think about AI is as a tool that is far better than what we have had before, providing capabilities we did not previously have. I compare it to when Excel suddenly appeared—doing things on a spreadsheet was much easier than using a calculator. We are seeing a similar leap in computational power, moving from the traditional methods used over the past 20 to 30 years to approaches that can perform efficient multi-parameter functions.

In biology and chemistry, there are no simple problems—everything is a complex multi-parameter problem. For example, if NVIDIA could not provide its processors to the world, you could not even perform the analysis, but beyond that, the architecture of analysis has fundamentally changed, enabling us to do deep learning on big data problems. We are looking for a needle in a haystack, with a large number of algorithms to optimize—for instance, figuring out where this atom should be placed in this molecule. You cannot solve all the different aspects of creating new molecules with a single algorithm—you need to integrate them.

Speaking of integration, from your perspective, why did the Recursion and Exscientia deal make sense?

The two companies were founded around the same time about 12 years ago with very similar missions, namely drug discovery. At that time, there were new technologies, but they were not predictive, so even with some of the smartest people in the world, we still could not solve the very basic, fundamental problem of a failure rate exceeding 95%. Why is that? Because we did not have predictive methods to understand what would happen in late-stage clinical trials. You actually have to predict how molecules should work, and both companies had that vision—Recursion focused on biology, Exscientia on chemistry. Observing the integration process has been fascinating because it has been much easier than imagined. Now chemists have all these amazing biological tools, biologists have these amazing chemical tools, and the two come together so powerfully.

What should drug developers focus on when looking for AI partners or acquisition targets?

Focus on use cases and validation. We are in an industry where many people have ideas about how to do both of these things, and those ideas all sound like they are heading toward the same goal or using the same language. So the only way to truly differentiate them is to define what you actually produce through your platform or technology. Interestingly, I have been deeply involved in this field for years, and when I read a press release, I think, 'Oh, it sounds like they are doing exactly what we do.' Then I dig deeper and find, 'Oh, they are doing something completely different from us.' The only way to truly grasp this is to look at what they actually produce.

Conversely, what should AI companies focus on when collaborating with the life sciences sector?

The first is commitment, because it is always easy to find partners or even funding around new, exciting technologies. In fact, I think many partners have a portion of their budget specifically dedicated to 'go spend this money on new, crazy ideas.' Truly impactful partners—like our collaborations with Roche and Sanofi, each with multiple projects—are as invested as we are on their side, and that investment comes from the top of the organization and runs throughout. When their relationships do not progress fast enough, it is almost always due to a lack of that commitment and the determination to get it done no matter what. You have to put in the time. If you do not buy into the change in process, it is like plugging an analog cable into the middle of fiber optics—either way, you are limited by the lowest quality link. So if a large pharmaceutical company says they will run everything the traditional way, then insert AI for three months, then go back to traditional methods, you lose most of the benefits of the novel platform. In our relationships with Sanofi, Roche, and all our partners, we have been able to break down many silos and move into an AI-first environment.

If you consider the pharmaceutical decision-making process, it takes a lot to move it forward. There are many different sign-offs, committees, panels—all sorts of things. I can always tell which companies are truly committed to AI because you can see the smoothness of decision-making, and that engagement comes from all levels within the organization. Without that, crossing milestones can be very difficult because it always requires another level of approval.

If you consider data as the currency in the life sciences sector, are there limits to the data that AI systems can utilize?

I do not think there are limits. But most data is worthless. Much of the data collected historically is often not in the right format, or it is very specific and not very useful. We use public datasets, or in collaborations we use their data and dig as deep as we can, but honestly, that only gets us to a very rough starting point. What is more useful is creating fit-for-purpose datasets to answer questions—it is designed directly to address the specific question you are asking. Since we are in proprietary drug R&D, if you really want to come up with new concepts, you have to start with something that can answer very specific questions.

So you also have to create your own data for this, right? That is the primary focus. We had a discussion with a pharmaceutical partner who asked us to look at their old data and help them build models. Our conclusion was that it would be more efficient and likely more predictive to actually recreate all the data they had been generating (on a smaller scale). You can achieve your goal without trying to reconstruct 30 years of data, right? Because all their data is in different formats and based on very specific projects, you would need an army working day and night to organize it into a form that can actually be used. So this is a point that I think many people overlook.