Amazon says that it’ll commit up to $230 million to startups building generative AI-powered applications.

The investment, roughly $80 million of which will fund Amazon’s second annual AWS Generative AI Accelerator program, aims to position AWS as an attractive cloud infrastructure choice for startups developing generative AI models to power their products, apps and services. Much of the new tranche — including the entire portion set aside for the accelerator program — comes in the form of compute credits for AWS infrastructure, meaning that it can’t be transferred to other cloud service providers like Google Cloud and Microsoft Azure.

To sweeten the pot, Amazon is pledging that startups in this year’s Generative AI Accelerator cohort will gain access to experts and tech from Nvidia, the program’s presenting partner, and will be invited to join the Nvidia Inception program, which provides companies opportunities to connect with potential investors and additional consulting resources.

The Generative AI Accelerator program has also grown substantially. Last year’s cohort — ten startups — received only up to $300,000 in AWS compute credits, amounting to around a combined $3 million investment.

“With this new effort, we will help startups launch and scale world-class businesses, providing the building blocks they need to unleash new AI applications that will impact all facets of how the world learns, connects, and does business,” Matt Wood, VP of AI products at AWS, said in a statement.

Amazon’s growing spending on generative AI tech, which includes efforts like its $100 million AWS Generative AI Innovation Center and its Project Olympus generative model, come as the company looks to catch up to tech giant rivals in the blooming — and increasingly competitive — generative AI space. While Amazon claims that its various generative AI businesses have reached “multiple billions” in run rate, the company is widely perceived as having missed the boat on generative AI.

According to reporting from The Information, AWS originally planned to unveil its own generative AI model akin to OpenAI’s ChatGPT code-named Bedrock — which eventually became Amazon’s Bedrock model hosting service — at its annual conference in November 2022. But major bugs forced the organization to postpone the launch.

Fortune’s Sharon Goldman this week reported that Amazon’s Alexa division has been beset with challenges as well, thanks to technical setbacks and political infighting. Nine months after a splashy press demo of a “next-gen” Alexa, which our intrepid hardware reporter Brian Heater covered, the new Alexa is reportedly far from ready for prime time — the result of insufficient training data, inadequate access to training hardware and other roadblocks.

Amazon also passed on early opportunities to back two leading AI startups, Cohere and Anthropic. The company later tried to invest in Cohere, but was rejected — and had to settle for a co-investment (albeit a large one, totaling $4 billion) in Anthropic with chief rival Google.