Meta and AWS team up to back 30 Llama startups

Running open-source AI models sounds great until you get your first cloud computing bill. Meta gives away Llama weights for free, but actually training, fine-tuning, and serving those models on cloud clusters costs serious cash.

Meta found the perfect partner to help cushion that blow: Amazon Web Services (AWS). The two tech giants announced a joint program to provide resources and backing to 30 promising startups in the United States that are actively building on Llama.

Six months of compute and mentorship

The initiative builds on the Llama Startup Program that Meta first unveiled back in May. Teaming up with AWS turns that initiative into tangible computing power.

Selected founders will get access to dedicated technical mentorship, AWS compute resources, and strategic go-to-market support over a six-month virtual program. Startups will also need to enroll in AWS Activate to unlock promotional cloud credits.

AWS VP and Global Head of Startups and Venture Capital Jon Jones noted that models like Llama are fueling applications prioritizing flexibility, transparency, and developer control. The goal is to help founders turn ambitious ideas into shipping products much faster.

The eligibility criteria are pretty specific: startups must be based in the U.S., be under ten years old, have raised outside capital, and already have an idea ready to execute with Llama. Applications close on August 8, 2025, and AWS plans to notify the 30 chosen startups by August 29, 2025.

Why it matters

We think this partnership is a win-win for both tech giants, and a playbook worth paying attention to.

For Meta, open-source adoption cannot just live on local laptops and research papers. To truly challenge proprietary giants like OpenAI and Google, Llama needs to be the production backbone for real-world companies. Partnering with AWS gives Meta a direct line into the cloud where modern startups already build.

For AWS, it is an essential counterweight against Microsoft's tight grip on OpenAI and Google Cloud's push with Gemini. By becoming the de facto enterprise home for Llama builders, AWS secures high-margin cloud consumption from the next wave of AI winners.

If you are an eligible U.S. founder building an AI application right now, this is free compute and direct engineering access from the people who actually built the weights and the cloud. It is an absolute no-brainer to apply before the August 8 deadline.

The only real catch is scale. Capping the cohort at just 30 U.S. companies leaves thousands of international developers out in the cold, still footing their own GPU bills.