OpenAI researcher Miles Wang in talks to launch AI drug discovery startup valued at $2B

The Great Brain Drain Continues

Another day, another high-ranking OpenAI talent is walking out the door to build something new. This time it is Miles Wang, and he is not looking to build another LLM wrapper or a slightly better coding assistant. He is reportedly in talks to raise funding for a drug discovery startup that investors are already valuing at $2 billion before the first slide deck has probably even cooled off from the printer.

We have seen this pattern before. A researcher spends a few years at the epicenter of the AI boom, gains the kind of internal knowledge that money cannot buy, and then realizes that the real equity value is not in being employee number eighty at a massive firm, but in being founder number one at a specialized one. For builders, this is a signal that the frontier of AI is moving away from general-purpose models and toward deep, vertical integration in fields like life sciences.

Why Biotech is the New Frontier

For the last eighteen months, the tech world has been obsessed with generative AI—writing emails, making art, and summarizing meetings. But if we are being honest, those are low-stakes use cases. If an AI hallucinates a summary of your weekly sync, you lose ten minutes of productivity. If an AI can accurately predict how a protein folds or how a specific molecule interacts with a human cell, you save ten years of research and billions of dollars in clinical trials.

This is why Wang’s move matters. The venture capital world is starting to get weary of the overhead costs associated with training generalist models. Developing a model that competes with GPT-5 requires tens of billions in compute. Developing a model that solves a specific problem in biology requires high-quality data and specialized talent, but it does not necessarily require building a private power plant. It represents a pivot toward utility over novelty.

The Two Billion Dollar Question

A $2 billion valuation for a company that effectively exists in the pre-incorporation or early-seed stage is, frankly, absurd by any traditional metric. It reminds me of the dot-com era where a pedigree was worth more than a product. However, from a founder's perspective, this valuation is not about current revenue—it is about the scarcity of talent. There are only a handful of people on the planet who truly understand the intersection of large-scale transformer architectures and biological data sets.

Investors are betting that the foundational work done at OpenAI can be ported over to the world of pharmaceuticals. They are looking for a 'Bio-GPT' moment. But builders should be cautious here. Raising at such a high valuation puts an enormous target on your back and leaves very little room for the pivots that nearly every startup eventually needs to make. When you start at the top of the mountain, there is nowhere to go but down if your first few experiments fail to yield a breakthrough drug candidate.

Data is the Real Moat

In the world of AI drug discovery, the model is rarely the most important asset. The moat is the data. While Wang likely has incredible insights into how to build efficient training loops and architecture, the success of this new venture will depend on whether he can get access to proprietary biological datasets that others cannot. Traditional big pharma companies are notoriously protective of their research. A startup like this has to either partner with the incumbents or find a way to generate its own synthetic data that actually reflects reality.

If you are building in the AI space right now, you should be looking at how you can apply these complex models to legacy industries that are still using spreadsheets and manual labor. The 'low hanging fruit' of the internet is gone. The next wave of massive companies will be built by people taking AI into the lab, the factory floor, and the pharmaceutical trial.

The Founder Perspective

I have spent a lot of time talking to founders who are frustrated by the dominance of the 'Big Three' AI firms. It can feel like the game is rigged because they have all the GPUs. But Miles Wang's exit shows where the cracks are starting to form. The talent wants to solve real-world problems, not just optimize click-through rates or make a chatbot sound more human. There is a specific type of ambition that is not satisfied by being a small cog in a massive machine.

What Builders Should Take Away

• Vertical Intelligence: The market is rewarding specialized applications over general tools. Find a niche where a 10% improvement in efficiency equals billions in value.
• The Pedigree Premium: Whether we like it or not, where you worked previously still dictates your initial valuation. If you don't have the OpenAI badge, focus on proving your proprietary data access instead.
• The Cost of Ambition: A $2 billion valuation means you are playing a high-stakes game. For most founders, it is better to raise less at a lower valuation to maintain control and flexibility.

We are entering a phase where AI is becoming a tool rather than the product itself. Wang is betting that the most valuable thing he can do with his knowledge is to apply it to the human body. It is a massive gamble, but it is the kind of gamble that actually moves the needle for society. We should stop caring about whether an AI can pass the Bar exam and start caring about whether it can cure a disease. That is where the real value is being built.

The pivot from 'Artificial General Intelligence' to 'Artificial Biological Intelligence' might be the most lucrative transition in the history of tech, but only for those who can navigate the laboratory as well as they navigate the server room.

Wang's move proves that the talent is getting bored with pure software. The next generation of builders will be those who bridge the gap between bits and atoms. If you are waiting for a sign to stop building another SaaS tool and start looking at hard problems, this is it. Just prepare for the reality that the stakes in biology are much higher than they are in software. You can't just 'move fast and break things' when people's lives are on the line.

Read the original at TechCrunch AI →