Pharma Bets Big on AI Platforms with Flurry of New Year Deals

2026 kicked off with a stream of AI platform deals across pharma, signaling a cultural shift away from single-asset bets and toward investment in AI infrastructure for broad discovery.  

The collaborations between AI start-ups, Chai Discovery, Noetik, and Boltz, and pharma giants, Eli Lilly, GSK, and Pfizer, respectively, will implement AI platforms across diverse applications, including biologics design, cancer clinical outcome prediction, and small molecule drug discovery.  

“If 2025 was the year of breakthrough research, we believe 2026 will become the year of deployment,” said Jack Dent, co-founder at Chai. 

Launched in 2024, the AI-driven biologics company aims to develop medicines against targets that have previously been undruggable. Chai closed out 2025 with a $130 million Series B, reaching a $1.3 billion valuation just 18 months after launch. 

Chai’s core technology centers on Chai-2, a de novo antibody design model capable of generating full-length antibodies with therapeutic attributes, thereby reducing reliance on labor-intensive and time-consuming experimental screens. 

In a preprint posted on bioRxiv, Chai-2 was reported to achieve double-digit success rates, a more than 100-fold improvement over previous computational methods. Chai’s platform has hit a range of hard targets, including peptide MHCs, immune complexes implicated in cancer, and G protein-coupled receptors (GPCRs). The latter are signaling proteins that represent approximately one-third of current drug targets yet are notoriously hard to target because their accessible regions barely protrude from the cell membrane. 

Early adopters 

Traditional tech-pharma collaborations focus on a small number of drug targets, such as one to three programs over multiple years. The pharma company presents a challenge while providing funding and clinical development expertise. The tech partner works to discover a molecule. 

In contrast, Chai’s latest deal with Lilly will deploy Chai’s technology to design novel biologics for multiple targets. Chai will also develop an exclusive AI model for Lilly that is trained on the pharma giant’s proprietary data and tailored to Lilly’s discovery workflows. 

According to Dent, these platform shifts come around very rarely. Adopters of these tools will have a competitive advantage across speed, cost, and achieving first-in-class molecules for difficult targets. Notably, Lilly representatives entered Chai offices within days of seeing Chai-2 research results in an early collaboration. 

“We chose Lilly as our launch partner because they are the largest pharma company in the industry. The potential impact is huge and the team there is exceptional,” Dent told GEN Edge. “It was interesting to see that the biggest player in the space is also the fastest.” 

Aliza Apple, PhD, vice president of Lilly Catalyze360 AI and global head of Lilly TuneLab, reiterates that Lilly’s AI and drug discovery philosophy centers on being early adopters of promising tools. Models need to be fueled with rich data and undergo rigorous testing to design better molecules from the outset.  

“We want to lean in early to the tools that look truly differentiated and put Lilly’s weight behind them, not just rely on what we’ve already built,” Apple told GEN Edge. Instead of outsourcing molecular design, the collaboration with Chai brings generative AI capabilities in-house to empower Lilly’s scientists. 

The deal with Chai is just one of Lilly’s numerous AI-based initiatives. For example, at last week’s J.P. Morgan Healthcare Conference, Lilly announced a $1 billion partnership with Nvidia to build a co-innovation AI lab to address key challenges in drug discovery. The lab will be staffed by combined teams from the pharma giant and the computational powerhouse. 

Get a license 

When Noetik was founded three years ago, the San Francisco-based company made a bold bet on building biological foundation models trained on human data, even as much of the industry remained fixated on molecular design, recalled Ron Alfa, MD, PhD, Noetik’s CEO. 

While AI sectors, such as protein modeling, are driven by troves of large public datasets, such as the protein data bank (PDB), large-scale translational data remains an empty chest.  

Noetik tackles this gap by generating multimodal data from primary human tissue samples with intact in vivo context. These data train the company’s foundation models which aim to predict clinical outcomes in cancer patients. 

Earlier this month, Noetik announced a five-year licensing partnership with GSK, which gives the pharma giant access to Noetik’s non-small cell lung cancer and colorectal cancer foundation models. The deal includes a $50-million upfront payment and will follow a subscription-based framework. 

Alfa describes the GSK partnership as one of the “first true foundation model licensing deals in biotech.” The unique partnership bridges the gap between SaaS-style software licensing and the rigorous requirements of biopharma surrounding IP and data security.  

“For years, the sector has looked for a way to commercialize AI as infrastructure rather than the standard R&D collaborations,” Alfa told GEN Edge. “Now, we have a template.” 

Kim Branson, SVP global head of artificial intelligence and machine learning at GSK, told GEN Edge that even with limited precedent for licensing AI models, “our ambitions made this the right move.”  

Noetik’s models can be fine-tuned with GSK datasets and patient outcomes to improve target discovery, patient segmentation, and response prediction. “This complements our internal platforms and organoid capability as we discover, develop and deliver potential new cancer medicines to patients,” Branson said. 

Faster for patients 

The AI-big pharma deals don’t end there. 

Boltz, the new AI research and product company launched earlier this month with $28 million, has already solidified a multi-year collaboration with Pfizer to build exclusive models that improve target selection for structure prediction, small-molecule affinity, and biologics design. Boltz scientists will also work closely with Pfizer’s discovery teams to advance custom models and workflows for a number of target programs to enhance preclinical decision making. 

Boltz is a public benefit corporation (PBC) advancing open science for AI-based drug discovery. The start-up is co-founded by a trio of MIT researchers—Gabriele Corso, PhD, Jeremy Wohlwend, PhD, and Saro Passaro, known as the developers of the widely adopted Boltz series of models. 

Additionally, Isomorphic Labs announced a deal with Johnson & Johnson this week, marking its third pharma partnership. The Google DeepMind spinout previously unveiled collaborations with Eli Lilly and Novartis last year. 

Under the new multi-target research partnership, Isomorphic Labs will be responsible for in silico predictions and design, while J&J will run experimental tests to advance programs through development. In contrast to Chai, Noetik, and Boltz, the AlphaFold 3 developer has traditionally kept models in-house. 

Reflecting on this pattern of deals, Elliot Hershberg, PhD, partner at Amplify, says two things are clear. First, pharma leaders view AI as a critical area of investment for the future of their business. Second, some portion of this innovation will be achieved by external dealmaking.  

“While these tools are powerful, they are also increasingly difficult to build and maintain internally,” Hershberg told GEN Edge.  

When scanning the AI competitive landscape, the field anticipates a future where models will commoditize, presenting a challenge to companies seeking to differentiate based on platforms. Yet, Hershberg tempers that we’re not close yet.  

“Despite real competition, only a handful of groups have been able to approach state-of-the-art results,” he said. AI companies continue to distinguish themselves by pursuing proprietary data, out-of-distribution problems, or building workflows and infrastructure needed to make the tools broadly accessible to scientists. 

Simon Barnett, partner and head of research at Dimension, adds that many of today’s protein language and structure models have saturated the same benchmarks, giving the false impression that technology has hit a wall.  

“There is so much we can’t do with these tools yet, but that I believe will become tractable,” Barnett told GEN Edge. Among the remaining challenges for these models include accurately predicting in vivo immunogenicity, generating complex modalities such as bispecifics, and extending hits to more difficult targets, including T-cell receptors. 

Barnett also foresees more large-scale partnership deals to come between the current wave of AI drug discovery companies and big pharma, noting that a combination of companies providing data, models, infrastructure, evaluation services, and more, will power the innovation engine for future R&D. 

“Pharma is replete with cash and eager to experiment, whether through traditional M&A to bolster their pipelines or more ambitious model and data-licensing deals like what Lilly is doing,” Barnett continued. 

Nick Myerberg, partner at Braidwell, describes a larger trend that places less fixation on model novelty and more focus on how scientific decisions actually get made. 

“The real value is not just in building a better model,” Myerberg wrote on LinkedIn, “but in creating the systems that will aim to make discovery faster, smarter, and more reliable for patients.” 

All told, cultural transformation is slow and rebuilding infrastructure takes time. Yet pharma’s accelerating investment in AI reflects a growing conviction that these technologies can redefine drug discovery. This proposition will ultimately be validated, or not, in the clinic.