The Age of Generative Biology is Here
Biotech Frenzy Online? | My Ex-NASA Engineer Dad Says "Cope"...
The AI Pivot :
Biotech Was Always the Final Boss
A while back I made a post about the coming LLM deployment to drug discovery. While I’m still never wrong, todays AI models won’t be able to create these next-gen drugs on a whim.
Biotech is the ultimate test: You can’t bullshit biology. I reckon the LLMs of today will be an accelerant, but for the science of tomorrow? Brute-forcing raw compute simply isn’t enough.
AI as an “Accelerator”?
The tech-bros will try either way and make headlines. And whether you like it or not, from now on — AI is here to stay. In fact we’re just past the starting line.
But let’s be honest for a second…
AI models remain locked behind screens, patched to model updates and has yet to fully interact with the laws of physics. So what would actually qualify as a breakthrough?
Currently it takes 8-10 years and at least ~$2B to research a drug to market with an abysmal 6.1% success rate from the FDA. Without FDA approval a drug can’t enter phase I.
The dream is turning the standard 10-year research cycle cut into 18 months with near perfect success rates… That would silence AI critics instantly, RIP Gary Marcus.
That’s unlikely. But if AI stands to change the game, it would need to cut the speed at least in half or increase success rates significantly.
Filling the Gaps?
In any case: I still think the dream is a decade off minimum. But that won’t stop the tech-bros from attracting enough VCs to put things into motion. And eventually yield tangible results…
OpenAI’s “Rosalind” sounds cool but to actually make the ‘next level’ difference? It needs more data, automated labs working 24/7 and honestly further research overall.
This is coming, and when it does? An undeniable breakthrough will have been achieved. So what players are set to win this game?
Tempus AI: Endless Data Flywheel
It remains highly uncertain but I’d say the only real application right now suited for AI’s current capabilities is Tempus AI.
They’ve gathered an insane data MOAT through real-world screenings and diagnostics which they then leverage every petabyte of to improve individual patient outcomes.
The healthcare system is known for its waste but with Tempus a doctor can order tests like tumor sequencing and get reports in days rather than months. And for every test sent into Tempus? Adds to its AI dataset.
But that’s not all, data isn’t equal.
AstraZeneca has tons of it, but because of sloppy categorization + privacy laws? It’s locked. Now Big Pharma are willing to pay millions for datasets to design drugs + trials that work on humans.
The Potential of TEM 0.00%↑?
That’s where the magic lies, every pharma giant thinks their “in-house” data is enough. It isn't.
Proprietary trial data only gets you so far. You actually need the messy, real-world patient data at scale. And that's what Tempus has which no-one else can realistically compete with.
So when OpenAI's Rosalind or whatever model Big Pharma deploys next needs to train on actual human outcomes? They'll come knocking on Tempus's door.
TEM 0.00%↑ has grown 114% since it’s IPO and consensus only expects ↑36% this year, I think that’s a miscalculation… But either way:
Market cap is $10B with $1.27B in sales which means investors are paying $8 for each dollar in sales. Pretty cheap because the potential here is real and still evolving.
Might honestly be a 10x before 2030.
On the Drug-Discovery side?
Keep reading with a 7-day free trial
Subscribe to Stock Invader Substack to keep reading this post and get 7 days of free access to the full post archives.