As AI burns cash faster and monetizing open-source models becomes difficult, Meta seeks funding from Amazon and Microsoft

The AI arms race is heating up, and Meta is also facing financial pressure, seeking external support for the development of its large model Llama.

According to technology media The Information on the 17th, over the past year, Meta has sought funding support from several tech companies, including Microsoft and Amazon, to share the training costs of its flagship large language model Llama.

In return, Meta has proposed various cooperation plans, such as allowing funding supporters to participate in the development decisions of Llama's future features. It is currently unclear whether these proposals have made substantial progress. The report noted that when Meta initially proposed this cooperation concept called the "Llama Alliance," the market response was relatively lukewarm.

The report also mentioned that although Meta has had the most in-depth discussions with Amazon and Microsoft, insiders revealed that Meta has also negotiated with companies such as Databricks, IBM, and Oracle, and even with at least one representative from a Middle Eastern investor. Discussions about the "Llama Alliance" have continued into early this year.

"Llama Alliance": Casting a Wide Net for Support

Analysis indicates that Meta's move aims to strengthen relationships with major cloud service providers like Amazon and Microsoft and explore more possibilities for the commercialization of Llama. To promote the commercial application of its model, Meta is actively working on several initiatives, including an internal project called "Llama X," which aims to develop application programming interfaces (APIs) for enterprises.

The analysis pointed out that the key challenge Meta faces is that Llama, as an open-source software, can be used for free by anyone, which makes its commercialization path fraught with difficulties. Meta's primary profit model is advertising, and it has almost no experience in commercial software sales.

Some companies approached by Meta are cautious about investing in a model that will ultimately be open-sourced, as this amounts to paying for technology that can be obtained for free in the end. Meanwhile, other open-source AI models, such as DeepSeek, which outperform Llama, have emerged in the market, which may further reduce other companies' willingness to fund Meta.

AI Burn Rate is Astonishing, Tariffs May Further Increase AI Costs

Meta is expected to spend between $60 billion and $65 billion on capital expenditures this year, an increase of about 60% compared to 2024, primarily for the development of more AI data centers. This expenditure accounts for about one-third of its expected revenue this year.

Although Microsoft, Google, and Amazon have each invested similar or even greater amounts in capital expenditures, the revenue scale of these three companies is larger than that of Meta.

As of December 31 last year, after considering debt factors, Meta still had as much as $49 billion in cash on its books, and generated $91 billion in cash flow from operations last year.

Despite Meta's strong financial position, it may become increasingly difficult in the coming years to continue investing heavily in AI research and development while also meeting shareholder expectations for stock buybacks and dividends Recent media reports indicate that private equity firm Apollo is in talks to lead up to $35 billion in debt financing for Meta's data centers.

In addition, tariffs that the Trump administration may impose are expected to further increase the development costs of AI. Given the uncertainty surrounding future spending by businesses and consumers on AI services, the market's widespread skepticism about whether the collective investment of hundreds of billions of dollars by technology companies in data centers and chips over the past two years was wise will also intensify