The number pharma / AI partnerships over the past few years has exploded as awareness, access and PoC milestones have all accelerated. This has been driven both within biopharma (i.e. breakthroughs such as AlphaFold), as well as outside of it (i.e. OpenAI's products, Meta's open-source tooling). Partnership discussions are now focused around the value-add of a particular approach vs. an elucidation of the benefits of AI.

And for pharma, it makes a lot of sense to partner. They can avoid rushing the integration of AI in-house (difficult given the org size, embedded processes & cross-functional nature) and spread their bets across a number of innovative companies.

While the pharma interest is there, the competition for a deal with them is intense. The pace of early-stage life sciences funding from 19-21 led to the creation of many AI-first startups, built from the ground up to balance tech & scientific resources. While these companies often operate at the cutting edge, that alone hasn't warranted a deal, as pharma wants to spend less time ascertaining the tech differentiation between these companies and instead focus on their ability to demonstrate impact via case studies or pilot projects.

This can sometimes create a chicken and egg problem for startups, as they need access to more robust datasets (which pharma has) to demonstrate impact, but they can't get access to these without first demonstrating impact. There's no easy way to round that square, but a combination of internal initiatives, case studies using publicly available data, partnerships with smaller companies that are more flexible with data rights and very persistent BD efforts can all get the ball rolling. Given lead times for some pharma partnerships can be >18 months, half the battle is just staying in the game - it's important to have an internal champion (or two) that can help push the deal along.

However, it's worth noting that despite these options, many of the headline AI / pharma partnerships have repeatedly been with a few companies. This could reflect a scenario where the experience, data quality, talent, funding and prior deals of these companies creates a feedback loop that ensures more of all of these flow back to them.

To get an equivalent deal, most startups must pass through a series of escalating pilot projects, building trust & familiarity. The hope is to transition from answering interesting scientific questions to assisting with key R&D efforts as quickly as possible.

From a macro perspective, the impact of all of these partnerships has been fairly modest - we have not seen any groundbreaking AI-developed assets reinvent SoC. However, it's also early in what will be a long innovation cycle - looking at past incarnations (i.e. mAbs, C>), it typically takes 20+ years before we start to see that type of impact. As Patrick Malone, MD PhD said, "measured expectations is the path forward.