Amazon Web Services has signed a multi-year joint marketing agreement with Superblocks that lets the vibe-coding startup embed its tool directly inside AWS customers' private clouds, keeping data, databases, and inference calls from ever leaving the customer's account. Apps built through Superblocks will spin up Amazon Aurora databases rather than external Supabase instances, and route model calls through Amazon Bedrock rather than to frontier labs directly. AWS will also help sell Superblocks to enterprises through its Marketplace channel, giving a 50-person startup with $60M raised a distribution surface it could not otherwise reach.
Superblocks closed its Series A in May 2025, backed by Spark Capital, Kleiner Perkins, Meritech Capital, and Greenoaks. The company sits in a crowded vibe-coding market alongside Lovable and Replit, but the AWS deal reframes the category: rather than a consumer-flavored app builder pointed at external SaaS databases, Superblocks becomes a governed, IT-sanctioned tool that produces applications automatically enrolled under enterprise auditing, encryption, and network controls.
That governance pitch is the whole product argument. Business-user apps built with vibe coding have historically been rogue applications from a CIO's perspective — spun up outside IT, sending data to whichever model or database the tool defaulted to. Superblocks CEO Brad Menezes told TechCrunch the AWS integration inverts that default.
Key facts
- 01AWS and Superblocks signed a multi-year joint marketing agreement to embed vibe coding inside AWS customers' private clouds.
- 02Apps built via Superblocks will spin up Amazon Aurora databases and route inference through Amazon Bedrock, keeping data inside the customer's AWS account.
- 03Superblocks has 50 employees and $60M raised total, backed by Spark Capital, Kleiner Perkins, Meritech Capital, and Greenoaks since its May 2025 Series A.
- 04Open models accounted for 29% of traffic through Vercel's AI gateway last month, signaling a rapid shift away from single-vendor stacks.
- 05AWS has Kiro for developers and Quick for business users, but no in-house vibe-coding agent to rival Lovable or Replit.
The deeper story is what the deal reveals about how hyperscalers now want the AI stack sliced. AWS, along with Microsoft and Google, is pushing enterprise customers to separate frontier models from the scaffolding around them — the agent harnesses, orchestration layers, security tooling, and app builders — and to buy that scaffolding from the cloud provider rather than from the model lab. The pitch is neutrality: the cloud runs whichever model the customer picks today and whichever one they pick next quarter.
Microsoft CEO Satya Nadella has been making the same argument publicly, telling enterprise customers to adopt multiple models to avoid lock-in and warning that AI labs may use enterprise data to study — and eventually compete with — the businesses they serve. AWS is making the same bet with a different piece: rather than build its own vibe-coding agent for business users, it is partnering with Superblocks and running the workload on Aurora and Bedrock. AWS does have Kiro, an AI coding agent for developers, and Quick, a Copilot-style assistant for business users, but neither competes directly with Lovable or Replit.
The multi-model shift among CIOs has moved faster than most vendors expected. Menezes said the change is recent and sharp.
Open models accounted for 29% of all traffic routed through Vercel's AI gateway last month, a figure that would have been unthinkable a year ago when Anthropic's Claude was the reflexive enterprise pick for coding workloads. Menezes argues that a multi-model strategy spanning OpenAI, Anthropic, and open-source models — increasingly including Chinese open-weight releases alongside emerging US open-source options — is now table stakes for a CIO. The demand extends beyond coding into customer service, HR, and sales automation.
Menezes went further, predicting consequences for executives who resist the shift.
“any enterprise that is betting on a single model provider, that executive will be fired.”— Brad Menezes, Superblocks co-founder and CEO
There are reasons to treat that framing with some skepticism. Superblocks has a direct commercial interest in a multi-model world — its whole product design assumes customers do not want to be tied to a single lab, and the AWS deal is worth more if that assumption holds. Enterprises that have standardized on Claude for coding, or on GPT-class models for reasoning tasks, may take longer to rewire their stacks than Menezes' 60-day framing suggests. And Bedrock itself is not model-neutral in practice — it favors the models AWS has commercial arrangements with.
Still, the direction is clear. Vibe coding for business users is following AI coding agents for developers into the private cloud, wrapped in IT governance and sold through hyperscaler channels. That is a very different distribution model from the consumer-viral path Lovable and Replit took, and it is the path enterprises with compliance requirements will actually walk down.
For the AI market, the Superblocks deal is a small line item — a Series A startup getting a marketing agreement with a cloud giant. For the shape of the enterprise AI stack, it is a data point in a much larger repositioning. The hyperscalers have decided the durable margin sits in the scaffolding, not the model, and they are systematically arming startups that reinforce that thesis. Model providers that assumed they could sell agents, orchestration, and application layers on top of their own APIs are about to find those categories occupied by AWS, Microsoft, and Google partners with distribution the labs cannot match.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
