Good morning. Thanksgiving is almost here. Hope yours is filled with good company and food nobody has to lie about enjoying. We’ve definitely come a long way since Tofurkey was the weirdest thing on the menu. Now scientists are using CRISPR to tweak the genes of a fungus to make it taste just like turkey. In a few years, Thanksgiving might mean less roasting and more harvesting. Pass the gene-edited gravy, please.
🍗 We’re taking off for Thanksgiving. Enjoy the holiday and see you Monday!
Google needs 1000x more AI servers by 2030. To handle exploding AI demand, Google’s head of AI infrastructure says they’ll have to double serving capacity every 6 months. Analysts call this “stage 2” of the AI boom, where the hardest problems aren’t compute, they’re physical constraints: power supply, cooling, and space. Google’s betting on their power efficient Ironwood AI chips to help ease the capacity crunch, but at this pace, they’ll soon run out of planet.
Gemini crushed benchmarks last week, then Claude said, “hold my beer.” Anthropic says its latest AI model is better for coding, agents, and office work. Claude Opus 4.5 is the first model to score over 80% accuracy on SWE-Bench Verified, beating recently released rivals Google’s Gemini 3 and OpenAI’s GPT-5.1 in coding. It’s built for enterprise users, managing spreadsheets, presentations, and complex work tasks. Welcome to the singularity, where frontier AI becomes outdated within days.
The self-driving wars are down to Google vs Tesla. Robotaxis are predicted to become a trillion-dollar market, and both companies are positioned to win big. Google’s Waymo already has hundreds of thousands of weekly riders in 5 US cities, with over 100 million autonomous miles logged. Tesla barely has any robotaxis on the road, yet claims it can instantly convert millions of customer cars into a massive fleet with software updates. Whoever wins could own the future of driving, if they can convince everyone it’s safe to let go of the wheel.
Guide to AI innovation
Looking to adopt and scale agentic AI solutions?
This report from Glean and Gartner combines innovative insights with practical steps: how to tame agent sprawl, set guardrails, combine LLMs, and ensure reliable performance at scale.
Zero in on opportunities and solve critical challenges. Get the report.
Antitrust laws? Apparently they don’t apply to AI. Lawyers are sounding alarms over “Stargate,” the $500B AI supercomputer backed by OpenAI, Nvidia, Oracle, and SoftBank. One Yale researcher calls the project a textbook violation of antitrust rules on the books for 135 years. These rules exist to stop companies from forming cartels that raise prices and squash innovation. US regulators are silent, as AI dominance is more important than enforcing old laws. Competition is great – until it interferes with profits and national security.
A German court ruled ChatGPT 4 violates copyright. OpenAI was accused of illegally storing and reproducing popular song lyrics. Even inaccurate or incomplete lyrics generated by ChatGPT were counted as infringement by the court. Unlike Stability AI’s recent UK win over Getty Images, the German ruling makes OpenAI directly responsible, not its users. Now OpenAI must disclose exactly how much copyrighted material it used and how much money it made. Big copyright holders might cash in, but smaller creators shouldn’t hold their breath.
Trump just handed AI startups the keys to federal supercomputers. His new executive order, the Genesis Mission, gives private companies direct access to the DOE’s National Labs. These labs hold some of the world’s most powerful supercomputers – and sensitive datasets like nuclear research, alien autopsies, and biotech studies. The idea is to fuel rapid scientific innovation with AI. Let’s hope someone remembers to lock down the national security secrets.
fun stats
🔎 900+. Amazon’s hidden data centers in more than 50 countries, including colocation facilities.
💰$60 billion. Nvidia’s current cash pile. It’s grown 40% this year, giving Nvidia massive ammunition for acquisitions or vendor financing.
🎬 $23.4 million. Q3 revenue CuriosityStream got from licensing its video library for AI training – over 50% of what it made from subscriptions in 2024.