Superblocks inks AWS deal to trap vibe-coded apps inside private clouds

Vibe coding just got a corporate leash. Superblocks, a 50-person startup with $60 million in funding, has struck a multiyear joint marketing deal with AWS that lets enterprises keep AI-generated apps locked inside their own private clouds. No data exfiltration to external models. No rogue Supabase instances spinning up outside IT’s sightline.

The mechanics are straightforward: when a business user on AWS prompts their way to a new internal tool, Superblocks provisions an Amazon Aurora database inside the company’s virtual private cloud and routes everything through Amazon Bedrock. “The big thing about that is data never leaves,” CEO Brad Menezes told TechCrunch. “It’s their AWS account and basically secure with all of the auditing, all of the encryption, all of the network controls.” AWS will also co-sell the product through its Marketplace, the same channel it uses for thousands of other partners.

But the deal signals something larger than one startup’s distribution win. It’s a shot across the bow of frontier model providers. Menezes claims the enterprise mood has flipped decisively toward multi-model strategies. “Any enterprise that is betting on a single model provider, that executive will be fired,” he said, pointing to the sudden embrace of Chinese open-weight models. Microsoft’s Satya Nadella has been making a parallel argument, warning customers that AI labs cannot be trusted with orchestration layers because they might study a company’s data and later compete with it.

The numbers back up the shift. Open models accounted for 29% of all traffic routed through Vercel’s AI gateway last month, a popular enterprise tool for managing multi-model usage. AWS doesn’t have its own vibe-coding agent for business users yet — Kiro targets developers, and Quick is more of a Copilot-style assistant — so this partnership plugs a gap while reinforcing the hyperscaler pitch that the scaffolding around AI models should live on their turf, not in the labs’ walled gardens. The quiet implication: the cloud providers see the application layer, not just compute, as the real lock-in battleground.