Meta floated the idea of becoming an accidental cloud provider if it overbuilds, and Databricks reminded everyone that enterprise AI deals don't die in the lab, they die in the CIO's imagination.

The real shift isn't about models anymore; it's about who controls constrained resources, how safely chaos can be made boring at scale, and whether the next cloud war gets fought with homegrown silicon instead of Nvidia's back catalog.

Hiring internationally is complex.

Learn what an Employer of Record is and how startups use EORs to hire global talent compliantly.

Enterprise AI isn’t getting rejected; chaos is. At TechCrunch Disrupt, Databricks co-founder Arsalan Tavakoli-Shiraji is spelling out the real enterprise AI deal killers: not weak models, but operational instability, governance gaps, and unclear failure modes.

Pilots now fail less in the lab than in the imagination of the CIO. If a system threatens compliance, complicates incident response, or rewires workflows too abruptly, it dies before rollout. The AI stage session at Disrupt is essentially about this: AI is judged on how safely it can be made boring at scale.

For AI startups, that shifts the job from maximizing model impressiveness to minimizing organizational risk. The companies that win the Disrupt registration push era of enterprise AI will be the ones that integrate cleanly, degrade predictably, and make change feel survivable to large organizations.

Smart homes were supposed to be the future.

Hollywood promised us houses that just worked.
Instead, we got disconnected apps, broken automations, voice assistants that fail basic commands, and a graveyard of abandoned smart home gadgets collecting dust in drawers.
So what happened?
In this episode, Caya dives into the hidden history of smart homes — from 1970s protocols like X10 to Zigbee, Z-Wave, Thread, Matter, Alexa, HomeKit, and the modern chaos of incompatible ecosystems.

Meta’s latest AI bet comes with an escape hatch: if it overbuilds data centers, it might just become a cloud provider. In his shareholder Q&A, Zuckerberg said a cloud move is “on the table,” turning massive AI capex into an option to resell excess compute rather than a sunk cost; a quiet but pointed signal to nervous investors.

Meta is already fielding weekly inbound from companies asking it to “stand up an API service” or sell compute at a premium. If Meta does flip the switch, it instantly becomes a fourth U.S. hyperscaler, with the social graph as a built‑in demand engine for AI agents and assistants.

For AI-native startups, a Meta cloud would mean one more deep-pocketed buyer of workloads and one more platform risk: today’s partner, tomorrow’s direct AI competitor.