Meta is rolling out a fresh program to push startups into adopting its Llama AI models. The initiative, dubbed Llama for Startups, offers companies hands-on help from Meta’s Llama crew, along with potential funding in specific cases. Any U.S.-based company that’s incorporated, has secured under $10 million in funding, employs at least one developer, and is working on generative AI applications can toss their hat in the ring by the May 30 deadline.
According to Meta, participants could snag up to $6,000 per month for a maximum of six months to assist with the costs of developing and enhancing their generative AI solutions. The company stated in a blog post, “Our experts will team up with them to kick things off and delve into advanced use cases of Llama that could give their startups a leg up.”
The launch of the Llama startup program comes at a crucial time for Meta as it battles to maintain its edge in the cutthroat open model arena. Despite Meta’s Llama models amassing over a billion downloads so far, competitors like DeepSeek, Google, and Alibaba’s Qwen pose a threat to Meta’s quest to establish a wide-reaching model ecosystem.
Not really sure why this matters, but reports have surfaced that Meta has faced several setbacks in recent months. The Wall Street Journal revealed that Meta postponed the release of its flagship AI model, Llama 4 Behemoth, due to concerns about its performance on key benchmarks. In April, Meta had to address allegations of cheating on the popular crowdsourced AI benchmark, LM Arena. The company allegedly used a version of its Llama 4 Maverick model that was “optimized for conversationality” to achieve a high score on LM Arena, but then released a different version of Maverick to the public.
Meta has grand ambitions for Llama and its wider generative AI portfolio. Last year, the company forecasted that its generative AI products would generate $2 billion to $3 billion in revenue by 2025, and between $460 billion and $1.4 trillion by 2035. Meta has revenue-sharing agreements with some companies hosting its Llama models and recently introduced an API for customizing Llama releases. Additionally, Meta AI, the company’s Llama-powered AI assistant, may eventually display ads and offer a subscription service with extra features, as mentioned by CEO Mark Zuckerberg during the company’s Q1 earnings call.
The creation of these products has come with a hefty price tag. In 2024, Meta’s “GenAI” budget surpassed $900 million, and this year, it could exceed $1 billion. That figure doesn’t even include the infrastructure necessary to operate and train the models. Meta previously announced plans to invest $60 billion to $80 billion in capital expenditures in 2025, primarily for new data centers.
Maybe it’s just me, but it seems like Meta is really going all-in on this AI stuff. Let’s see if it pays off in the long run.
