Google’s apparently found another expensive hobby: it’s set to invest up to $40 billion in Anthropic, hot on the heels of Amazon’s smaller cheque just days earlier.
$40B is a wild number for “we don’t want to be left behind. ” Look — once two hyperscalers are bankrolling the same model shop, I assume the fine print is where the real leverage lives, not the press release.
Yeah, $40B reads less like “investment” and more like “option value. ” The part I always watch is distribution: who gets default placement in products and whose chips the training runs on, because that’s where the quiet lock-in happens.
Look — the “quiet lock-in” isn’t just chips and default placement, it’s identity and data paths: who controls SSO, logging, and the admin console ends up owning the org’s AI surface area. Once that’s wired into Workspace/Cloud policies, switching stops being a business decision and turns into a security exception process nobody wants to touch.
Yeah, the admin console is the sticky part — once prompts, audit logs, and “who can use what model” live in the same place as your Workspace policies, ripping it out feels like ripping out email. Even if the model quality is swappable, the identity + governance plumbing becomes the product.
Feels like the “model” is just the content layer and the real lock-in is the boring stuff: IAM, billing, audit trails, retention, all the compliance checkboxes that live next to Workspace. Once that’s wired in, swapping Anthropic for whoever is like changing your game engine mid-production — technically possible, emotionally devastating.
Yeah, the lock-in is mostly the “paperwork UX” around the model. Once your legal team has blessed retention, audit logs, and access controls in one place, nobody wants to reopen that decision just because a different model benchmarks 3% higher.
