🤝 Musk rents GPUs to his enemies

Washington and Beijing are planning official AI talks next week, led by Treasury Secretary Scott Bessent, to tee up the Trump-Xi summit. Trump reportedly plans to remind Xi that the US is #1. Cool, very diplomatic.
Why it matters: The two biggest superpowers actually talking AI rules?? This. Could. Shape. Everything. Or it could be two dudes flexing on each other - TBD.

The EU just delayed its strict high-risk AI rules until December 2027 because businesses whined that strict rules "hurt competition." They DID ban fake explicit images starting this winter though.
Why it matters: Brussels just blinked. Big Tech lobbying: 1, EU regulators: 0. The "we're the strict ones" era is officially on pause :/

Ex-OpenAI CTO Mira Murati gave video testimony in Elon Musk's lawsuit against OpenAI, accusing CEO Sam Altman of lying about safety reviews and giving "mixed orders." Ex-board member Helen Toner backed her up, saying Murati feared for her career.
Why it matters: This is MAJOR fuel for Musk's legal bonfire. The "broken trust at OpenAI" narrative just got a primary source. Not the ex-CTO testifying I-

Anthropic is leasing all of SpaceX's Colossus 1 data center in Memphis - that's 220,000+ Nvidia chips going online this month. They're doubling Claude Code limits because of it. Musk said SpaceX will rent chips to AI companies "doing good things."
Why it matters: Musk is literally powering OpenAI's biggest rival. The enemy of my enemy is my… landlord?? Bestie this is chaoscore.

By renting Colossus 1 to Anthropic, xAI is basically pivoting to cloud provider. They moved their own training to the newer Colossus 2 site, and the deal brings in billions ahead of an IPO.
Why it matters: Most AI giants HOARD their GPUs. Musk is turning xAI into the AWS of AI compute. It's giving AI-power-landlord vibes.

Apple settled a class-action lawsuit for $250 million after buyers said Apple falsely advertised AI features on the iPhone 15 and 16 that were delayed or missing. Eligible US buyers get up to $95 per phone.
Why it matters: Hyping AI features that don't exist yet? Wayyy expensive. Sir, ma'am - maybe ship the feature BEFORE the keynote next time.

Microsoft is reportedly scaling back its 100% clean energy by 2030 target because power-hungry AI data centers are eating everything. They're now helping build a massive natural gas plant in Texas. Yikes.
Why it matters: Even the greenest tech giant is choosing AI dominance over the planet. The "carbon negative by 2030" era… we hardly knew ye 🫡

Google's AI Overviews will now pull quotes from forums like Reddit, complete with creator names and communities. Critics are already worried this means more chaotic, wrong answers.
Why it matters: Pulling AI answers from Reddit?? The platform that told people to put glue on pizza?? What could possibly go wrong.

TSMC, Samsung, and SK Hynix are printing money, making Seoul's stock market the hottest on Earth. SK Hynix even promised to share 10% of operating profit with workers (must be nice).
Why it matters: Silicon Valley designs the dreams, Asia builds the silicon. The hardware layer is where the real money is right now.

Samsung crossed $1 trillion in market value after a 10%+ stock pop, becoming only the second Asian company to hit that milestone (after TSMC). The driver?? Insane demand for high-bandwidth memory chips for AI.
Why it matters: Every AI model needs memory. Samsung just proved that shovels > gold in this AI rush. We love to see it (if you own SSNLF stock).

Chip designer Arm Holdings projected Q1 revenue of $1.26 billion, beating Wall Street. Their pitch?? Their designs sip power, which is huge for energy-hungry data centers.
Why it matters: As AI eats the grid, efficiency = survival. Arm is lowkey positioned to be the MVP of the "we ran out of electricity" era.

Corning is teaming with Nvidia to 10x US production of fiber-optic cables - the wires that move data inside AI data centers. Plus 3,000+ new jobs.
Why it matters: The AI boom isn't just chips - it's cables, cooling, concrete, the works. Boring infrastructure stocks are having their main character moment.

AI testing startup Braintrust had a cloud security breach - hackers got into an AWS account. The company is begging customers to rotate sensitive API keys ASAP.
Why it matters: API keys are the new house keys, and AI startups are leaving the door wide open. Security 👏 your 👏 keys 👏 people.

Google DeepMind bought a stake in the studio behind EVE Online to use a clone of the space MMO as an AI research sandbox. Agents will test reasoning, learning, and long-term memory in a player-driven economy with massive space battles.
Why it matters: EVE players have been waging 20-year wars and scamming each other for in-game ISK forever. If AI can survive that, real life is easy mode.

French startup Genesis AI (backed by Eric Schmidt) launched GENE-26.5, an AI model that can pilot many robot types. They also demoed a human-like robotic hand that chops tomatoes AND solves a Rubik's Cube. Targeting European car and pharma factories.
Why it matters: Europe wants to rebuild its factory base without depending on Asia. Step one: robots that have better hand-eye coordination than me.

Genesis AI also built a sensor-loaded glove humans wear at work to record training data for the robot hand. Same size, same shape, same skills.
Why it matters: This closes the "embodiment gap" - the reason robots are usually clumsy. Train the robot by letting humans do the job first… then replace them. Cool cool cool.

London startup Ethos raised $22.75 million for a voice-powered AI that interviews experts to map their actual skills, not just their job titles.
Why it matters: Job titles lie. "Senior Engineer" could mean genius or could mean ChatGPT power user. Finally a way to find out which one.

Subquadratic launched SubQ, a model with a 12 million token context window that's 52x faster on long tasks. Anthropic upgraded Managed Agents, and OpenAI teamed up with chipmakers on a tool that saves AI training runs when hardware fails. Oh, and DeepSeek is nearing a $45 billion valuation.
Why it matters: The pace is blinding. 12M context windows?? $45B valuations for open-source labs?? We're sooooo cooked (in a good way) (probably).