TechCrunchAI & LLMs

Reid Hoffman Exits Microsoft Board to Go Full-Time on AI Drug Discovery Startup Manus

After a decade on Microsoft’s board—a tenure that included the $26.2 billion LinkedIn acquisition, the first $1 billion OpenAI investment in 2019, and the $650 million Inflection AI acqui-hire—Reid Hoffman is stepping down to focus on his latest venture. The announcement came Thursday, bringing to a close a board seat that began when Microsoft bought his company in 2016.

Hoffman’s new priority is Manus, an AI-driven drug discovery startup that raised over $50 million across seed rounds last year. He holds the title of co-founder and board chairman, but the CEO role belongs to Dr. Siddhartha Mukherjee—a physician, biologist, and Pulitzer Prize-winning author. Hoffman is an investor alongside General Catalyst.

On a recent episode of his podcast “Possible,” Hoffman told Microsoft CEO Satya Nadella he’s ready to go “founder mode.” He cited rapid progress at Manus, particularly around what he calls “Move 37” AI—systems that surpass human creativity in chemistry, especially for cancer treatment. The phrase references a famous AlphaGo move that defied conventional strategy.

From an engineering perspective, Manus represents a shift in how AI models are applied to combinatorial chemistry. Rather than optimizing for general language tasks, the startup is building models that generate novel molecular structures with therapeutic potential. The architecture likely involves graph neural networks for molecular representation, reinforcement learning for property optimization, and large-scale distributed training on proprietary datasets. The key trade-off: generalizability versus domain-specific accuracy. Early funding suggests investors are betting that narrow, high-fidelity models can outperform broad foundation models in drug discovery pipelines.

Hoffman’s departure from Microsoft’s board—where he also served on OpenAI’s board until 2023 due to conflict-of-interest concerns—signals a deeper commitment to operational execution over strategic oversight. For data and ML engineers, the move underscores a growing trend: domain experts trading board seats for hands-on model building.

Source: TechCrunch

← Back to News