The fight moves to hardware
The biggest story of the day is a courtroom collision between two giants. Apple has sued OpenAI, alleging the ChatGPT maker systematically stole its hardware trade secrets as OpenAI pushes beyond software into building physical devices. The timing is no accident: the whole industry is pivoting toward machines that act in the real world. A cluster of reports now frames "physical AI" — intelligence embedded in robots and machinery — as the next great industrial investment bet. Apple's suit is a warning shot in a race where the prize is no longer just chatbots, but the devices they live inside.
Tesla made that shift tangible. After a decade of promises, it finally launched a robotaxi service on public streets in Dallas and Houston — turning a long-repeated pledge into actual driverless cars carrying passengers.
AI's real bottleneck is electricity
Strip away the software hype and a harder truth emerges: AI runs on power, and the bill is coming due. Gas plants are rising to feed AI's surging electricity appetite, reshaping how energy gets made — increasingly by burning more natural gas. The same logic drives India, where the data center industry is becoming one of the country's defining tech bets of the decade, hinging less on software than on securing enough electricity to run it.
Against that backdrop, a counter-current is fascinating. A proof-of-concept called Colibrì reportedly ran a 1.5-terabyte, frontier-class AI model on an ordinary computer with just 25GB of RAM — work that normally demands a data center. If efficiency breakthroughs like this hold up, they could ease the very power crunch that's redrawing the energy map.
AI leaves the phone behind
Intelligence is migrating off your phone and onto your body. In South Korea, reports describe runners wearing wrist devices that deliver AI features without a phone in their pocket — phone-free edge AI. Even Realities, meanwhile, is betting that smart glasses win by ditching the camera entirely: no lens means no one wonders if they're being recorded, and the company is pitching the devices for work rather than capturing the world around you. Both moves target the same trust and convenience problems that have dogged wearables for years.
Chips, geopolitics, and hidden sensors
The chip world served up friction on several fronts. A report says AI models from OpenAI and Google are reaching Chinese technology firms that the Pentagon has blacklisted — a reminder of how hard export lines are to hold. The Trump administration, separately, is under fire over an arrangement to make AI chips with the United Arab Emirates, which critics have branded a "corrupt deal."
Closer to the hardware, NVIDIA quietly stripped regular users of the ability to read the "hotspot" temperature — the hottest single point — on its new RTX 50 series GPUs built on the Blackwell architecture. Modders found the sensor anyway, reopening a window the company had tried to close.
Preparing for the quantum break
Finally, the long game. As quantum machines advance toward the day they could crack today's encryption, a parallel scramble is underway to lock down tomorrow's security — with firms like SEALSQ moving early to get ahead of the threat.
The throughline across all of it: AI's center of gravity is shifting from clever code to physical reality — the devices it inhabits, the power plants that feed it, and the legal and geopolitical walls going up around it.