Hello, dear readers!
This is our weekly brief on remarkable AI topics, so you can follow the revolution without having to live inside a model leaderboard.
Today’s focus — Meta’s possible second career as an AI landlord. The company is reportedly considering leasing Anthropic up to $10 billion worth of computing capacity over two years, turning infrastructure built for its own AI ambitions into a potential business. If the fiercest competitors start renting each other the machinery, is the AI race becoming less like a duel and more like an industrial estate?
Also in this week’s edition:
The EU wants Google to give rival AI assistants deeper access to Android — because the smartest assistant is not much use if somebody else owns the phone.
Netflix says generative AI has already been used in roughly 300 titles, suggesting the AI movie revolution may arrive one ordinary-looking crowd scene at a time.
Meta May Become Anthropic’s Landlord
Meta and Anthropic are reportedly discussing a deal under which Anthropic could lease up to $10 billion of Meta’s computing capacity over two years. The talks are still at an early stage and may produce nothing, but Anthropic would reportedly pay monthly and either company could leave the agreement early.
There is something wonderfully circular about this. Meta spent years building enormous AI infrastructure to compete with companies such as Anthropic. It may now discover that one perfectly good way to profit from the AI race is to rent that infrastructure to the competition.
This does not mean Meta is abandoning its own models or that the companies have suddenly become friends. It reflects a simpler reality: powerful AI requires a great deal of expensive hardware, and there are not many companies capable of supplying it at the necessary scale. Competitors may fight over products, talent, and users while quietly buying electricity and server time from one another downstairs.
The potential deal would also take Meta into unfamiliar territory. Its business still revolves mainly around advertising, and Reuters notes that it does not yet have an established operation for selling computing capacity. Mark Zuckerberg has nevertheless said that entering cloud computing is “on the table,” while Meta is reportedly developing a business around spare capacity and model hosting.
The AI race is often presented as a contest to build one decisive, all-conquering model. The Meta–Anthropic talks point toward something messier. Some companies will build models, some will build products, and some will sell the industrial machinery underneath — with the largest players occasionally doing all three.
Meta wanted to build the best AI. It may also have accidentally built a cloud company.
Europe Opens Google’s Side Door
The European Commission has ordered Google to give competing AI services access to 11 important Android functions. Rival assistants should eventually be able to respond to voice commands, understand information on the screen, act across apps, use on-device models, and complete tasks such as sending messages or ordering food.
The same decision requires Google to share anonymized data used to improve its search services with qualifying search engines and AI competitors. The search-data measures are expected to apply from January 2027, while the new Android access is due with the next Android iteration in July 2027. Google argues that the requirements could weaken privacy and security protections; the Commission says access will remain subject to safeguards and user consent.
The larger point is about distribution. An AI assistant does not become useful merely by producing clever answers in a chat window. It needs access to the button, microphone, apps, personal context, and operating system. Europe is effectively telling Google that owning the building does not mean Gemini should be the only assistant with keys to every room.
The AI Movie Is Already Here
Netflix says generative AI has already been used in roughly 300 titles on its platform, mostly during post-production. The tools helped productions create or enhance crowds, historical battles, establishing shots, and other complicated visual sequences.
One documentary, The American Experiment, contains 17 minutes of AI-enhanced footage. Netflix co-CEO Ted Sarandos said those scenes were produced twice as quickly and at half the cost of previous alternatives — and that some of the shots might otherwise have been dropped entirely because they were too expensive.
This is probably how generative AI enters film and television at scale: not through one proudly advertised, entirely synthetic blockbuster, but through hundreds of small production decisions. An extra crowd here, a historical battlefield there, and a shot the accountant would previously have crossed out.
The AI revolution in cinema may not look like an AI movie. It may look like a producer deciding that the battle scene can stay.
Thanks for reading AIport. Until next Monday — by then, AI will almost certainly have found another industry to enter through the service door.
