Meta's AI appetite turns it into a top Microsoft Azure customer
Two product launches and a privacy reckoning lead the AI desk today.
Meta is now one of Microsoft's largest AI customers, spending hundreds of millions of dollars a year on Azure to run trillions of AI tokens every week. According to a Bloomberg report carried by Techmeme and reposted across Chinese outlets, Meta's own AI ambitions have made it a major buyer of rival Microsoft's cloud — a striking turn for a company building its own data centers and custom AI silicon. The scale is eye-watering: trillions of tokens consumed weekly implies a voracious appetite for inference that even Meta's in-house capacity can't fully cover.
Why it matters: it shows how concentrated AI compute demand has become, with the biggest consumer-internet companies effectively renting intelligence from one another's clouds. For Microsoft, a marquee AI customer like Meta validates a multi-hundred-billion-dollar Azure build-out; for Meta, it's a reminder that serving frontier-scale models is still a capital-intensive arms race. The disclosure also sharpens the question of whether any single lab can be self-sufficient on compute.
My take: this is less a contradiction than a sign of the times — nobody builds enough GPUs to feed their own models, so even rivals become suppliers.
Slack is launching Slack Code, dedicated channels where teams collaborate with coding agents like Claude and Devin. Available today on every Slack plan, the feature spins up a project-specific channel the moment you tag a coding agent to build a feature, fix a bug, or update a page. Everyone on the team gets visibility into the agent's conversation, can review code diffs, see live previews of the output, and approve the work before it ships; channels auto-archive when the task is done and keep an audit log.
Why it matters: it's the clearest sign yet that "vibe coding" is moving from solo hobbyists into shared workplace infrastructure. Slack is betting that AI agents become teammates you manage in the same place you already manage people — partnering at launch with Claude Code, Devin, Vercel Agent, and GitHub Copilot. The risk is the same as any AI-in-the-loop tool: more code shipping faster, with human review becoming the bottleneck.
Meta's Ray-Ban smart glasses have become a workplace menace, with employees filmed, harassed, and creeped out on the job. A Verge report documents retail and service workers being secretly recorded by customers wearing the glasses, then turned into "prank" content for Instagram and TikTok — one Target worker's ordeal drew hundreds of thousands of views. The problem isn't only customers: one Sprouts grocery employee resigned after her boss wore Meta glasses on the floor, collapsing any expectation of privacy in the break room or bathroom.
Why it matters: as AI wearables go mainstream, the people least able to push back — workers who depend on the job — absorb the surveillance. Meta says its LED indicator and tamper-proofing address the concern, and it has banned hashtags tied to harassing content, but experts note US privacy law offers almost no protection once you're in public, leaving corporate policy as the only real guardrail.
What to watch: whether more employers move to ban recording wearables on the floor the way ICE already has.
Should companies be required to ban recording wearables from the workplace, or is that the employee's call? Tell us in the comments.
