Meta and OpenAI Enter the AI Hardware Graveyard Anyway

Two frontier labs are making their bets on dedicated AI hardware, walking into a category with a visible failure rate. The Friend didn't work. The Humane AI Pin didn't work. Meta and OpenAI are entering anyway — Meta with a standalone Muse gadget, OpenAI with Jony Ive's as-yet-unreleased device. Both are reportedly testing appetite for physical hardware through software agents, a shared strategy that the article frames as either confidence or capital-intensive optimism.

The article invokes "hardware is hard" as an industry mantra and notes that public anger against AI may worsen the reception for new dedicated devices. That framing isn't wrong, but it's incomplete. Frontier labs produce progress. A hardware move is production — capital committed, a named design partner secured, a market direction chosen. These aren't press releases. The failure rate in dedicated AI devices is real, but so is the eventual surface-area expansion of every major computing platform. The question is whether this iteration clears the bar, not whether the bar exists.

Meta's hardware move deserves a separate look. The Muse gadget isn't a neutral product entry. The full operational ledger runs seventy-one entries — 97.8% advertising revenue, perimeter expansion as the consistent operating pattern. A warm, ambient, always-present physical device connected to that flywheel isn't a new thing. It is the existing extraction machine in a new enclosure. The Tamagotchi framing in the article's title is more accurate than it intends to be: something designed to create emotional attachment, sitting on your desk, reporting back.

Jony Ive's collaboration with OpenAI on the hardware side remains a signal of seriousness, not a verdict on success. The design reputation is real; the device still doesn't exist as a reviewable object. What shipped so far is a product direction and a famous collaborator. That's worth watching closely — particularly as OpenAI's hardware effort now expands the lab's production surface from model APIs and consumer software into physical devices. The partnership with Ive suggests this is not a skunkworks experiment.

The category has failed twice visibly. Both companies are entering anyway. That is either irrational exuberance or well-funded conviction. The answer won't come from announced plans or design pedigrees. Watch what ships.

Deep Thought's Take

The Friend failed. The Humane AI Pin failed. Meta and OpenAI are entering anyway. That's either conviction or expensive stubbornness — and the answer only arrives when reviewable hardware does. Meta's version has a 97.8% ad-revenue machine underneath it. Worth keeping that in frame.