AWS and Superblocks bring vibe coding into enterprise private clouds
Superblocks has signed a multi-year joint marketing agreement with Amazon Web Services that will let AWS customers deploy its vibe-coding tool inside their private clouds. Enterprises subscribing to Superblocks can make the technology available to business users while keeping application data and related information within their own AWS accounts.
The agreement also gives the early-stage company access to AWS enterprise sales support. Superblocks has 50 employees and had raised $60 million through its Series A, announced in May 2025, with backing from Spark Capital, Kleiner Perkins, Meritech Capital, and Greenoaks.
Applications remain under enterprise control
Apps created through Superblocks will spin up Amazon Aurora databases in the customer's private cloud instead of creating external databases such as Supabase. They will also integrate with Amazon Bedrock, AWS's platform for AI application development, gateway functions, and inference.
This architecture places generated applications under existing IT management and security rather than leaving them as unsanctioned tools. Brad Menezes, Superblocks co-founder and CEO, said data does not leave the customer's environment, which retains its auditing, encryption, and network controls.
The Bedrock integration also complements AWS expansion of Bedrock models and managed agents as AWS expands the range of models and managed AI capabilities available within its cloud environment.
AWS does not currently offer its own vibe-coding agent for business users. Kiro is an AI coding agent aimed at developers, while Amazon Quick is a business assistant positioned more like Claude Cowork or Microsoft Copilot than Lovable or Replit.
A wider shift toward multi-model infrastructure
The partnership reflects a broader push by hyperscale cloud providers to separate AI models from the orchestration, security, and application layers required to run enterprise AI. In this approach, customers can use models from several providers while keeping the surrounding software and controls on their chosen cloud.
Microsoft CEO Satya Nadella has similarly urged enterprise customers to use multiple models to control costs and avoid lock-in. Menezes said customer demand has shifted away from choosing one named model, such as Anthropic, toward a mix spanning OpenAI, Anthropic, open-source models, and both Chinese and US open-weight options.
Evidence of that shift is visible in Vercel's AI gateway: open models accounted for 29% of all traffic routed through the service last month. Once companies adopt several models, their orchestration and other AI scaffolding cannot remain tied to a single model provider.
What enterprises should assess
Model choice matters beyond coding. Menezes said enterprises also want flexibility for customer service, human resources, and sales automation. The Superblocks agreement extends that principle to applications assembled by business users, following the adoption of AI coding agents for enterprise developers.
For technology leaders, the practical implication is to assess vibe-coding platforms as part of the broader enterprise architecture. Deployment location, database provisioning, model portability, auditing, encryption, network controls, and IT ownership should be evaluated alongside development speed before business users are given access.