The news: Novo is teaming up with Anthropic for AI-driven drug research. Novo will test R&D workflows through Claude Science, Anthropic’s AI workbench for researchers, and use Anthropic’s frontier models to strengthen the software development needed to scale AI across the company. Together, the companies aim to accelerate drug discovery and development.

Why it matters: AI drug discovery has attracted significant investment from both Big Pharma and venture capitalists backing companies built around AI-powered drug research. Drugmakers hope the technology can analyze vast, complex biological datasets to identify drug targets and disease mechanisms that researchers might miss.

Rather than build the computational expertise, many large pharma companies are buying it through partnerships. Novo alone has deals with OpenAI and Amazon Web Services, and other recent tie-ups include Merck with Google Cloud, Eli Lilly with Insilico, and Nvidia with several drugmakers. The money is also flowing to standalone platforms: Isomorphic Labs raised more than $2 billion earlier this year in one of the sector’s largest private financings.

The investment has not bought proof, though. AI can accelerate early research, but it has not cleared one of pharma’s biggest R&D hurdles: demonstrating efficacy in large clinical trials, where failure rates can exceed 90%. A drug discovered with AI has no better chance of approval or reaching the market.

AI’s technical wins have also not yet become better medicine. It has shown strength at predicting protein structures and identifying potential molecule-target interactions, but it has not proven those advances consistently translate into safer, more effective drugs, according to academic, pharma, and biotech experts writing a paper recently published in Nature Reviews Drug Discovery. They called AI’s clinically relevant impact “disappointingly limited,” arguing that models are often judged on benchmark performance rather than whether they improve real-world decisions. Messy biological data compounds the problem: when datasets are inconsistent, incomplete, or lack critical context, AI’s ability to predict human outcomes narrows.

Implications for pharma: Thin results so far reflect missing evidence, not proof that AI cannot deliver. That distinction should shape how pharma paces its spending: AI can accelerate specific research tasks, but evidence that it improves clinical success rates will take years to emerge as drug candidates progress through trials. Set expectations around measurable gains at each development stage, instead of promising near-term breakthroughs or scoring programs solely by the number of targets and molecules identified.

This content is part of EMARKETER’s subscription Briefings, where we pair daily updates with data and analysis from forecasts and research reports. Our Briefings prepare you to start your day informed, to provide critical insights in an important meeting, and to understand the context of what’s happening in your industry. Not a subscriber? Click here to get a demo of our full platform and coverage.