AWS is helping vibe-coding startup Superblocks, and the implications are big

Superblocks, the vibe-coding startup, has signed a multi-year joint marketing agreement with Amazon Web Services that lets its tool live directly inside the private clouds of AWS customers. In practice, that means a business user at an AWS-hosted enterprise can spin up an app through Superblocks without any data or code ever leaving the company's own account. The apps create Amazon Aurora databases rather than the Supabase instances so common in vibe-coding circles, and they plug into Amazon Bedrock for the AI model plumbing. The result is that these apps land squarely under IT's existing security and management setup instead of floating around as rogue shadow-IT projects.

Superblocks co-founder and CEO Brad Menezes frames it simply: the tool comes to the customer's data, inside their own cloud, rather than the other way around. He points out that everything stays wrapped in the account's existing auditing, encryption and network controls. AWS, for its part, will also help sell Superblocks to enterprise customers the way it does for other Marketplace partners, telling TechCrunch it backs partners where it sees real customer demand and alignment with how people want to build.

What makes this notable is what AWS doesn't have yet. There's no in-house vibe-coding agent aimed at business users. Kiro exists, but it's built for developers, not the vibe-coding crowd. Amazon's Quick assistant is closer to something like Copilot than to a Lovable or Replit. So this partnership hands Superblocks, a company with 50 employees and $60 million raised as of a Series A in May 2025 from Spark Capital, Kleiner Perkins, Meritech Capital and Greenoaks, a meaningful distribution boost through a much bigger partner.

But the real story sits a level above this one deal. Cloud giants are pushing enterprises to treat AI models as interchangeable parts, separate from the orchestration, security and app-building layers built on top of them, and they want those layers bought from the cloud providers themselves rather than the frontier AI labs. Microsoft's Satya Nadella has been making the same pitch lately, urging customers toward multi-model setups to cut costs and dodge lock-in, while also warning that AI labs handling orchestration duties might use customer data to study and eventually compete with them.

Enterprises appear to be moving that way regardless of the warnings. Menezes describes a sharp shift in mindset: just weeks ago, customers wanted one specific model, and that model was usually Anthropic's. Now he says a multi-model approach spanning OpenAI, Anthropic, and open-weight models, increasingly Chinese ones with U.S. open source starting to catch up, has become close to mandatory for CIOs, not just for coding but for customer service, HR and sales automation too. Vercel's AI gateway data backs that up somewhat, with open models accounting for 29% of traffic routed through it last month. Menezes goes as far as predicting that any executive betting the company on a single model provider will eventually get fired for it.

So the pattern extends further than one startup's AWS deal. First came AI coding agents built for professional developers running inside enterprise clouds. Now vibe coding for ordinary business users is following the same path into locked-down, private infrastructure. AWS calls it an emerging category with real momentum, which is corporate speak for: they plan to be everywhere this trend goes next.