David Friedberg says Anthropic asked big pharma for their data and nearly everyone said no "There's been an effort by Anthropic to sign up life sciences companies to contribute to a new life sciences focused model. They're approaching these large companies with large proprietary data sets and saying, if... you share your data, we will give you early access, some sort of proprietary value. Sign this NDA and you can participate with us." "I think nearly everyone I've spoken with has woken up to the fact that they are trying to commoditize everyone's business. If all of the tens of billions of dollars you have invested in experiments and product development, and you've generated all of this proprietary data along the way, that data is a true asset of your organization. It's an asset that you've spent billions of dollars developing." "And by handing it over to a model company to then combine with other people's data, you are commoditizing the one core differentiation that you have. And so everyone is largely saying no." "I think what everyone's realizing is they're better off developing their own weights and their own models using either an open source basis or there might be some intermediary business model that evolves."show more

Obviously the right decision by big pharma. Anthropic has nothing to offer them and no proprietary data and a history (aka Figma) of screwing over “partners”. This is exactly why I don’t think Anthropic is a threat in pharma.

Why hand over proprietary data you’ve spent billions creating just to help Anthropic or any frontier lab build a better general model that could then compete against you? Pharma’s “no” makes perfect sense, and the same logic applies across industries: finance, energy, manufacturing, defense, anywhere differentiated data is the real moat. Sharing it commoditizes your edge for vague “early access” promises that rarely materialize exclusively. This is exactly why open-source models running locally (or on-prem) win: full control, no data leakage, continuous fine-tuning on your own assets, and no vendor lock-in. The future isn’t feeding everything into a few centralized black boxes, it’s sovereign AI that stays yours.

Dario asking pharma for their data is very cute

They just have to buy out companies to get the data. You can’t just steal it.

I think most of these companies will just buy their own gear and make their own models. If you have a good model like GLM 5.2 you can model train the next one your self, you don't need that much compute. They might only need one rack of GH200 servers to serve their needs. Like most of their employees will be in the pharma lab, they will only need a small IT team. You could set it up in the basement of most buildings, the fire system and safety will be around $5m to build out the basement, maybe maximum $10m Then a server rack of inference is about $15m maybe have two and then all the network gear maybe another $5m So for a big pharma company they could build a system that would fulfill their needs for under $50m in my estimate depending on how large their needs are. This way they will be fully self reliant and will not have any ongoing API bills. Also if they need software they can likely build it them selves.

It’s worse. Big Pharma R&D is often subsidized with govt funding. So they’re essentially stealing from We The People to then attempt to sell it back to us at absurd costs that only the elite will access

Commoditizing everbody's business may be true, but this Chinese governance would not make this an option!

The disease of capitalism is how profit often outweighs progress. Big pharma protects its own drugs instead of pooling discoveries, slowing breakthroughs that could accelerate cures and benefit humanity.

They don't give them their data, it's okay. They will take it anyway. Why did Lilly spend a billion dollars on GPUs?

Proprietary data is becoming the real moat in AI. The companies that hand it over risk turning their most expensive advantage into someone else’s platform.

I'm betting smaller players with solid fundamental science research (in all fields) will become buyout targets for frontier models in the near future.

Elites fighting over the power of AI. They all want to use it in a way that advantages them. Separate R&D from production. R&D excels the fastest when information is shared. Shared R&D benefits everyone. Product companies pay for access to R&D and compete by productizing, marketing and distributing. R&D becomes a utility, not a competitive edge. The competition remains in how R&D is used. Probably won't happen this way because elites would rather fight over the full prize. This is going to be a knife fight in a closet, benefiting no one until some faction wins and is likely to get worse from there. The danger is not AI, it's what elites will do to control it and what the winners do with it once they do.

This is the next step

Easy to train specialized protein #ai yourself with @open_fold or reach out for contract help with fine tuning!

Friedberg is spot on. This was my thesis in 2022: once local compute can efficiently process proprietary data at scale, companies stop commoditizing their biggest asset by feeding it to centralized models. They build their own instead.

The idea that a pharma company can catch up with Anthropic is silly. First prisoner to defect and join Anthropic wins.

pharma treating their data as the real moat instead of trading it away makes total sense once you frame it that way

Nothing to do with AI finding cures for things that would otherwise make pharmaceutical companies lots of money while they sell drugs to manage conditions.

The pharma refusals make total sense when you read the dual-role problem clearly: Anthropic is now both selling tools to pharma and running its own preclinical pipeline, so handing over your compound data is handing it to a competitor. Private-infrastructure deployment is the mitigation they're offering, but data officers apparently aren't buying it.

The last thing Big Pharma can afford is AI getting ahold of their fraudulent data. ChatGPT... What really happened in 2020?

Pharma is going to be like the newspapers vs. Google. GFL

well if anthropic starts building models that can synthesize new molecules and proteins they can then patent them and get that revenue for themselves. ML molecular bio will save millions of lives in the future but the legal outcomes are not looking good.

@chamath @DavidSacks @jason The problem is that these companies don't know how to best leverage these highly informative data sets. And in the end, that hurts patients who need those advances that are lying latent at these companies.

Yeah sure. Name them then.

What's with Chamath's smokestack background? Big warmist?

have to invite dracula in

open weights is freedom and sovereignty the "intermediary business model" is ai companies that help non-ai companies do private post training with their proprietary data

no possibility that AI would find major problems in the 'data' that big pharma has collected. It's all about protecting research.

Good to see greed prevailing

Not surprised

Nasty work

Data is the only moat left in a world where compute and talent are for rent. No pharma exec is giving that up for "early access."

asked pharma... truly insane in the blind way

@novonordisk