Sponsored by
Hello readers,
Welcome to the AI For All newsletter! Today, we’re talking about the AI supercomputer that aspires to invent the drugs of tomorrow, AI risk visibility as the future of cybersecurity, and more!
AI in Action: Running Billions of Experiments You'll Never See
Last month, Eli Lilly inaugurated LillyPod, what it's calling the most powerful AI factory wholly owned by a pharmaceutical company. Built on an NVIDIA DGX SuperPOD with 1,016 Blackwell Ultra GPUs, the Indianapolis-based system delivers more than 9,000 petaflops of AI performance — and was assembled in just four months. The ambition is straightforward: break the physical constraints of the traditional wet lab, where even the most productive research teams can typically evaluate roughly 2,000 molecular ideas per target per year. In the computational dry lab LillyPod enables, scientists can simulate billions of molecular hypotheses in parallel before committing to a single physical experiment.
The system is designed to support large-scale training of protein diffusion models, small-molecule graph neural networks, and genomics foundation models — with Lilly's genomics team alone working from 700 terabytes of data. Lilly also plans to make select models available through TuneLab, its AI platform for biotech partners, which will offer both proprietary Lilly models and NVIDIA's open BioNeMo foundation models under a federated learning architecture — meaning partner companies can tap in without their data ever leaving their own environment.
To be sure, the gap between computational scale and clinical outcomes is wide, and has humbled the industry before. AI-assisted drug discovery has generated enormous excitement for over a decade, but most AI-designed drug candidates are still in early-to-mid-stage trials, and none have yet crossed the finish line to FDA approval. Simulating billions of molecules is only valuable if the models predicting their behavior are accurate — and model fidelity in complex biological systems remains an active research challenge. The 10-year drug development timeline Lilly hopes to cut in half is shaped by clinical trial logistics, regulatory review, and human biology as much as by discovery speed.
But if the models perform, the bottleneck in medicine may be moving, slowly but meaningfully, away from the lab bench.
🔥 Rapid Fire
Analysis: The AI industry is lying to you
Data center capacity additions halved in Q4 2025
This refers to announced capacity, not capacity brought online
Announced capacity reached 241 GW at the end of 2025
Only 33% of this capacity is under “active development”
This does not necessarily mean construction has started
67% of announced capacity is “not real” and is a mixture of:
Hopeful permits
Speculative land deals
Projects that assume not-yet-built power sources
AI data centers are financed using debt and are not profitable
The sale of GPUs is outpacing the building of data centers by years
Hundreds of billions of dollars of GPUs are mostly sitting idle
The question is being raised — why keep buying NVIDIA GPUs?
Supermicro’s co-founder was arrested for selling NVIDIA GPUs to China
Wally Liaw and others sold hundreds of millions of dollars of GPUs
Wally Liaw resigned in a 2018 fraud scandal yet was rehired
Supermicro is a reseller of GPUs used by CoreWeave and Crusoe
Supermicro is inexplicably not named in the indictment
The question is being raised — how did Jensen Huang not know?
Big Tech is mistaking token burn for quality, productive work
One of Meta’s internal AI tools caused a major security breach
Amazon’s Kiro and Q LLMs caused the following:
Incorrect delivery times
120,000 lost orders
1.6 million website errors
A 99% drop in orders across North America
This resulted in another 6.3 million lost orders
Multiple outages
Analysis: The Hater’s Guide to Adobe
OpenAI shuts down Sora, Disney cancels $1 billion equity investment
Sora was losing as much as $15 million per day
OpenAI makes Fidji Simo “CEO of AGI Deployment” and preps another model
OpenAI makes Greg Brockman the “Head of Mermaid Relations”
OpenAI surpasses “$100 million annualized revenue” from ads pilot
This means OpenAI made only $8.3 million in seven weeks
OpenAI’s first advertisers can’t prove ChatGPT ads work
Advertisers say platform is low-tech and low on data
SoftBank tests its own borrowing limits with $30B OpenAI investment
SoftBank needs another loan ($40B) to cover the investment
Concerns are mounting over costs and uncertain returns
AI slop is turning up in the opinion pages of major news publications
AI hallucinations lead to $10,000 fine for Oregon lawyer
TurboQuant enables compression of LLMs and vector search engines
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.
Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:
7 high-impact AI strategies to accelerate your marketing performance
Practical use cases for content creation, lead gen, and personalization
Expert insights into how top marketers are using AI today
A framework to evaluate and implement AI tools efficiently
Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.
📖 What We’re Reading
Attackers can now use AI to crack passwords, identify weak access points, and execute credential theft attacks faster than before. When adversaries enhance AI with quantum computing, the time required to break encryption and access sensitive systems drops significantly.
At the same time, these tools provide defenders with new advantages. Machine learning (ML) models can process millions of network traffic records to pinpoint anomalies that previously required hours of manual analysis. AI-based pattern recognition helps identify abnormal behavior early in the attack lifecycle.
