Am I answering a question that nobody has asked? I don't know but I wouldn't be surprised if my boss Bill Anderson has already asked this question long ago...
Eli Lilly just dropped €110 million on a supercomputer with 1,016 NVIDIA Blackwell Ultra GPUs. Add 10 years of running costs, and you're looking at €300-370 million total.
Meanwhile, Bayer's been strategically building cloud-first AI (AWS, Google Cloud, PRINCE platform) while paying down debt. Plot twist: Q3 2025 results look really good. Debt down €3.2B year-over-year, blockbusters launching, margins expanding.
So I asked: Should Bayer upgrade from the company car to our own Ferrari?
(Btw - we have much fewer company cars at Bayer than usually companies of our size have. For business travel in Germany we are obliged to travel by train since Deutsche Bahn is using 100 % green energy to power it's trains.)
Spoiler: The finances say maybe yes. And finding those 35 brilliant minds to run it? Challenge accepted. 😎
The Numbers (Or: Why This Ferrari Idea Isn't Crazy)
Lilly's Ferrari: €110M sticker price, €300-400M over 10 years. Discovers molecules in 5 minutes vs. a year. One of pharma's largest AI supercomputers.
Bayer's Momentum (We're Just Getting Started):
Debt: €32.7B, down €3.2B YoY – crushing it! 🎯
Q3 EBITDA: €1.5B, up 21%
Nubeqa + Kerendia: up 63% YoY, offsetting Xarelto's patent cliff
Lynkuet: Just launched November 2025
Pipeline: Cell therapy, gene therapy – computationally hungry stuff
The Reality: €300M equals 2-3 blockbuster drugs. BUT if it makes us discover molecules 3x faster and saves hundreds of millions in manufacturing? That's not an expense – that's a weapon.
When the mortgage is shrinking €3B+ annually and products are printing money? Maybe it's time to think bigger.
Why This Could Be Brilliant
Here's where the Ferrari metaphor breaks down: A real Ferrari is expensive, prestigious, and mostly about looking good going fast. A supercomputer? That's an expensive machine that can actually shortcut its way to billions in savings and competitive advantage. Here's why:
🚀 Speed = Billions
Molecular design: months → hours
Manufacturing optimization: real-time instead of quarterly
Cell therapy modeling: weeks → days
When racing blockbusters to market, speed isn't just nice – it's worth billions in first-mover advantage
💰 Economics + Manufacturing Gold
On-premise 5-7x cheaper than cloud for 24/7 workloads
Current cloud bill ~€12-15M annually (and growing)
Break-even in 3-4 years, then pure savings
Hidden gem: 1-2% biologics yield improvements = hundreds of millions in profit
The supercomputer could pay for itself in manufacturing optimization alone
🔒 Data Sovereignty
Your hardware, your data, your control
Easier GDPR/regulatory compliance
When you've got billions in IP and litigation provisions, ownership matters
🔬 Future-Proofing
Cell therapy, gene therapy, precision oncology in late-stage trials NOW
This isn't "maybe someday" tech – it's need-to-have for the science we're already doing
Pipeline loaded with computationally intensive work ready to feed a supercomputer
The Challenges (Nothing We Can't Handle)
€300M Is Real Money: Could develop 2-3 drugs instead. Counter: what if it helps us develop them faster?
GPU Refresh Cycles: 2026 GPUs need refreshing by 2029. But we'd own the advantage – trade-off worth it.
Cloud Works: AWS/Google delivering. Counter: at scale, ownership becomes cheaper.
Talent Competition: Germany's HPC pool: ~200 people. Tech offers €250K+. But the best want interesting problems. "Build pharma's largest supercomputer"? Pretty interesting.
German Regulations: Energy mandates, compliance. But Bayer's handled regulations since 1863. We're good at this.
Where to Park The Ferrari? (We've Got Great Options)
Leverkusen: HQ proximity, existing industrial infrastructure, political support (8/10) – safe and smart
Wuppertal: €1.4B R&D modernization happening anyway, perfect integration timing (7.5/10) – scientifically ideal
Berlin: Best talent pool in Germany, PRINCE platform already there, startup ecosystem (7/10) – talent magnet
Honestly? All three work. We've got the infrastructure, the political support, and the sites. The hard part isn't where – it's when. And Q3 2025 results suggest "when" might be sooner than expected.
So... Buy The Ferrari?
My Take: Timing is getting really interesting.
Phase 1 (2026): €30-40M initial, €120-150M over 10 years. Specialized HPC, hire 15-20 people, prove 3x faster OR €10M+ savings.
Phase 2 (2027-2028): Scale when Phase 1 delivers. Additional €150-220M. Total: €270-370M.
Why This Works:
+ Financial momentum: Debt down €3.2B YoY
+ Pipeline ready: Cell/gene therapy needs compute NOW
+ Manufacturing upside: Hundreds of millions in savings
+ Competitive necessity: Can't fall behind Lilly/Novo
Real Question: Not "can we afford it?" (we can). It's "can we afford NOT to?"
Prediction: Plan Q4 2025, decide Q1 2026, deploy Phase 1 end of 2026. Scale fast in 2027.
The Talent Reality
The Mission: 35 specialists at €150-220K competing against tech giants. German HPC pool: ~200 people.
Why We'll Win: Best people follow interesting problems. "Build pharma's largest AI supercomputer to save lives" beats "be Google employee #47,293." Bayer's turnaround story – blockbusters launching, betting on AI – is compelling.
Hot Skills: HPC Architecture (€120-220K), Computational Chemistry + AI (€130-180K). Wet lab → computational = 40-60% salary jump.
Evolution: 2020 = buzzword. 2025 = arms race. 2026 = German pharma needs 100+ specialists. This is the challenge I love.
Final Take
Lilly bought the Ferrari. Novo's shopping. Bayer's evaluating.
After 8 years at Bayer: we're stronger than some people think. Debt dropping €3B+ annually. Blockbusters growing 60%+. Pipeline loaded with compute-intensive science.
Hardware? Straightforward. Talent? That's where it gets fun. Our transformation story is real, our science is cutting-edge. I am sure that "Help build pharma's largest AI supercomputer" attracts exactly the minds we need.
Whether Phase 1 in 2026 (my bet) or conservative path, pharma R&D is an AI arms race. Winners combine right hardware with right minds.
That combination? That's what gets me excited to come to work.
Should Bayer build one of pharma's largest supercomputers?
Please answer in the comments!
If you're an HPC architect, computational chemist, or AI engineer who wants to work on something that matters – let's talk. 🚀
What's your take? Why does it make sense for Bayer to have a super computer?
Or is this nuts and just part of the AI hype?
(I heard people back in the days saying that this whole internet thingy is just a big hype and only a toy for computer nerds :P )
Please let me know in the comments what you are in favour of. Should buy build it's AI Ferrari?
#Pharma #AI #TeamBayer #MoreThanCareer #DSO #DrugDiscovery
Ready to take control of your pharma career?
Book a free getting to know call. I will tell you honestly where you stand, what the market looks like for your profile, and what your next move should be.
Book a Free Getting to Know Call