Across the life sciences sector, funding is being tied more closely to operational goals. Risk and legal teams are setting boundaries around data use, model oversight and accountability, while finance teams are unlocking budgets where ROI can be clearly demonstrated. Against this backdrop, investment intentions are rising. Partnership models are the preferred route to scale, and most organizations plan to source capabilities close to home.
"In order to make AI systems flexible and more fit-for-purpose, the spend has to be increased,” says the CFO of a US healthcare provider. "However, we are still unsure about the value that can be derived from AI functions and how long it will take to see promised returns.”
Research from the Boston Consulting Group indicates that returns are strongest when companies concentrate their resources. Across sectors, organizations that focus on a small number of high-impact use cases (about 3.5 on average) report about 2.1× higher ROI than peers pursuing a broader range (6.1 use cases). The implication for life sciences is clear: Higher AI budgets are more likely to deliver value where scope is focused, data access is pre-cleared and outcomes are tracked against well-defined business goals.
Growth strategies
Joint ventures and strategic partnerships top the list of planned expansion models. Some 62 percent of respondents plan to use partnerships in the next two years, and 30 percent cite this as their most important investment route. The logic is pragmatic: Partnerships offer faster access to trained models, specialized tooling and scarce AI talent—while allowing both parties to evaluate technical compatibility, data interoperability and governance fit before making further long-term commitments.
A vice president of a life sciences multinational explains the importance of partnerships over outright acquisitions, saying: "We've always been agnostic about where innovation comes from, whether internally or externally. That said, there's a definite trend toward more collaborations and licensing in the life sciences sector, especially to derisk. Acquisitions are costly and complex—you have to integrate systems and people, which isn't always straightforward. Licensing or partnerships allow us to set milestones and assess progress along the way. In some cases, that also includes an option to acquire later.”
In September 2025, Eli Lilly launched TuneLab, opening up its AI/ML discovery models, trained on more than US$1 billion worth of internal R&D data, to external biotechs, with initial partners including AI-enabled drug discovery and development company insitro. In parallel in the fall of 2025, Lilly announced collaborations aimed at advancing AI-assisted drug discovery, including a collaboration with insitro to build novel ML models to advance small-molecule discovery, a collaboration with Insilico Medicines to generate and design candidate compounds using Insilico's Pharma. AI platform, and a collaboration with NVIDIA to build an AI supercomputer to expand the scope of designing and testing potential compounds across multiple therapeutic indications.
Such alliances are an increasingly common sight. The same month, Novartis and Monte Rosa Therapeutics struck a licensing deal worth up to US$5.7 billion in immune-mediated diseases.
Monte Rosa's AI-enabled QuEEN platform will be used to develop selective protein degraders, while Novartis leads clinical development and commercialization—clear evidence that AI-powered design is already reshaping early-stage drug development. This type of partnership model allows companies to combine their expertise and resources. A joint venture between a pharma or healthcare company and an AI company can often deliver stronger and faster results by combining the expertise of both parties.
Life sciences companies working in association with a third party can lead to out-of-the-box thinking, which may be stifled when building internally. On the other hand, tech-native companies that also operate in healthcare often prefer to build in-house rather than partner, because they already have strong data-science capabilities and can rely on those internally.
Elsewhere, third-party buy-ins and venture capital investments are more common. In animal health and healthcare providers, 48 percent and 54 percent, respectively, plan to pursue buy-ins, while 52 percent and 44 percent are looking at VC investments. Buy-ins suit use cases where a plug-and-play tool can be dropped into existing workflows with limited modification.
Venture investments, on the other hand, offer exposure to emerging tools and partnerships without immediate operational commitments. These are often structured with commercial options or first-look rights to deepen engagement if performance meets expectations.
Regional investment
Most organizations expect the bulk of their AI investment over the next two years to remain regional. Local sourcing minimizes complications around data transfer, employment law and compliance—and is often better aligned with language, regulatory expectations and time zones.
That said, some firms are looking further afield. Asia-Pacific–based respondents are the most internationally focused: 27 percent expect their primary AI investment to go to North America or EMEA. This contrasts sharply with EMEA–based respondents, only six percent of whom expect their biggest AI investment to go outside the region, and only into North America. These findings reflect the gravitational pull of US-based AI vendors, startups and service providers, which are widely viewed as market leaders.
The US, in particular, is a natural target. Private investment in AI reached an estimated US$109 billion in 2024, by far the highest globally, while North America accounts for 49.3 percent of the global AI-in-healthcare market. The vendor ecosystem spans foundational model providers, life sciences-specific platforms, data engineering specialists and sector-aligned consulting firms.
Regulatory enablers are also stronger than in many other jurisdictions. The FDA maintains a public list of AI/ML-enabled medical devices and has authorized more than 1,200 to-date, 235 of them in 2024 alone, the most ever in a single year. The agency has also published frameworks around algorithm change control and Good Machine Learning Practice (GMLP), helping reduce ambiguity around compliance and review standards. This makes the US especially attractive for device makers and digital health companies seeking a clearer pathway to regulatory approval.
[View source.]
