The AI drug discovery industry has already gathered
momentum with AI start-ups having signed more than 200
deals with 50+ pharma companies over the last few years, and these are just the
disclosed deals. Few top companies like InSilico Medicine and Cyclica claim to
have over 100 collaborations each with Academia and Industries. With billions
of dollars pouring-in, it is likely to gain further impetus with the industry
approaching maturity in the next few years from its formative stage.
A conundrum that has been bothering these groundbreaking
start-ups is the business model.
AI companies have been shuffling with their partnership models,
having to display high flexibility to tend to the specific requirements of the
partners. The roles could range from utilizing AI to develop internal pipelines
as a biotech or providing AI as software or AI-driven services like a CRO.
In this model, AI companies’ model is analogous to that of a
typical biotech, either repurposing old drugs in new indications or designing
new drugs and fill their pipelines. Such companies usually aim to utilize AI
and create assets with lower costs and faster development timelines. These
assets could be then partnered or licensed out to pharma companies having
clinical development capabilities to generate revenue.
Such AI-driven companies would face the same challenges just as
a regular biotech pharma, needing a strong internal team with robust
therapeutic knowledge, having an experimental R&D infrastructure or
capabilities across the spectrum to outsource work, commercial or regulatory
These biotech’s, while powered by AI for better success, will
still have to compete for attention by the pharma. These companies will have to
attract an investor community with an appetite for high-risk opportunities and
long incubation periods. A few AI-powered biotechs like BERG or AI therapeutics
are well funded and have a few assets in clinical development.
In this model, AI companies sell their software or services to
pharma companies to build revenue. These companies develop platforms to work
with client’s data and aid with their programs. The aim is to create the best
computation tools which are usually therapeutic domain agnostic and can be
leveraged in a wide variety of applications. The companies like Atomwise,
PathAI, or Trails.ai are some good examples of this segment.
A challenge with such companies is the revenue generation model.
While software subscriptions are usually straightforward, payments in the
services segment are often structured in milestones to de-risk the investments
and bound with the ability to deliver what they claim, be it identifying new
targets or find new drug candidates. This could be a significant barrier
considering the development cycles of drugs can range from 4-6 years even after
entering clinical trials.
Another challenge is ‘AI is as good as the data, and it’s the
pharma companies that own the data.’ While even having developed the best
algorithms, these AI companies strive to prove their worth by collaborating or
providing free-of-cost services to academic labs or pharma companies as they
seek to develop and validate their platforms. Only after the proof of concept
has been generated, these companies transition to revenue-expecting
Learning from early collaborations, many service provision
companies quickly discovered the enormous market value of assets and started
working on a few internal programs or in joint ventures with other small
biotech’s. Internal programs can also help in generating enough validation data
to further attract pharma companies.
Likely forced by investors who did not want to amalgamate a
relatively low-risk service model to a high-risk biotech model, Spin-offs
became a sophisticated way to divest the offerings.
BioXcel therapeutics was among the first ones to spin-off its AI
platform in a service provider-only company, InveniAI in 2017. In 2019,
Atomwise spun-off X-37, to develop small molecule therapies for endodermal
cancers. More recently, InSilico Medicine spun-off its internal programs in
aging biomarker and deep aging clock as Deep Longevity to remain a service-only
one step at a time towards a robust business strategy
As the industry moves to maturity, the competition is going to
be fierce between upcoming start-ups and the big-pharma companies, who are the
major consumers of these technologies. In the coming years, majority of the ‘big-pharma’ players are likely to either acquire
such platforms or build internal capabilities, considering the
overall skepticism to share their data.
Both, biotech and service provision (with or without spin offs),
models have robust potential when executed correctly. However, succeeding
simultaneously in both provisions is going to be difficult as resources are
going to be limited. Investors are also likely to specialize to prioritize one
business model over the other, after carefully understanding the risks that
accompany each side.
More important question is that considering the advantages and
improvements that AI platforms offer, what is the fraction of the pie that they
deserve? How can the pharma companies quantify the value including the time and
cost advantage to compensate the AI companies? Or, what is the reasonable
proportion of IP that an AI-company deserve while developing a program in a joint
There are no correct answers to these questions yet.
However, the takeaway is to stay agile early in the journey,
experiment early, and move in a direction where the executive leadership
observes key strengths and experience.