Meta launches AI cloud business to compete with AWS, Google, and Azure

What happened?

According to an exclusive report from TechCrunch, Meta is developing plans to launch a cloud infrastructure business focused on artificial intelligence. The company would sell access to its AI computing power and its own models, directly competing with Amazon Web Services (AWS), Google Cloud, and Microsoft Azure. The strategy echoes SpaceX, which turned its excess launch capacity into the Starlink business, generating additional revenue from an underutilized resource. Meta has already begun internal discussions about the feasibility of offering its GPU clusters (such as those based on Nvidia H100) and its custom MTIA (Meta Training and Inference Accelerator) chips to external companies, according to sources close to the project.

Why is it important?

Meta has invested tens of billions of dollars in AI infrastructure. In 2023 alone, the company allocated over $35 billion in capital expenditures, a figure expected to exceed $40 billion in 2024, according to its financial reports. This investment includes building massive data centers, acquiring hundreds of thousands of GPUs (Meta owns more than 600,000 equivalent Nvidia H100 GPUs, per analyst estimates), and developing custom chips. Until now, that capacity was used internally for its products (Facebook, Instagram, WhatsApp, Reality Labs). However, the company has acknowledged that its AI infrastructure is underutilized during certain periods, opening the door to monetizing the surplus. By opening it to third parties, Meta not only seeks a new revenue stream — the AI cloud market is estimated to grow to $200 billion by 2030, according to Gartner — but also to position itself as a relevant player in the cloud market, currently dominated by three giants (AWS with ~32% share, Azure with ~23%, and Google Cloud with ~11%, per Synergy Research). Additionally, by offering its own open models (such as the Llama 3 family, which has surpassed 300 million downloads), Meta can compete with rivals' AI services like Amazon Bedrock, Google Vertex AI, and Azure OpenAI Service. This also reinforces its open-source strategy, differentiating it from the proprietary models of OpenAI and Google.

Consequences for the market

Meta's entry could intensify the price war in AI cloud, benefiting startups and companies that need computing power. Currently, the cost of training a model like GPT-4 is estimated at over $100 million, and cloud inference fees can be prohibitive for small businesses. Meta, with its idle capacity, could offer more aggressive pricing, forcing AWS, Google, and Azure to reduce their margins. However, it also poses challenges: Meta must demonstrate it can offer reliability and security comparable to AWS or Azure, which have decades of experience in enterprise services. Additionally, the company will have to address perceptions that Meta's use of customer data could raise privacy concerns, especially in Europe under GDPR. For users, more competition means more options and potentially lower costs. Moreover, as Meta is a provider of open models (Llama), it could foster a more open ecosystem compared to the proprietary models of Google and OpenAI, driving innovation and transparency. However, some analysts warn that Meta could use its position to favor its own models, creating a conflict of interest similar to what Google faced with its search and advertising business.

What readers should know

• Not a done deal: Meta is still developing the plans and there is no launch date. TechCrunch sources indicate the project is in early stages and could change or be canceled.
• Business model: It is expected to sell access to GPUs and TPUs, as well as inference and model fine-tuning services. Meta may also offer its own AI software, such as PyTorch, which is already widely used in the research community.
• Direct competition: AWS, Google Cloud, and Azure already offer similar services; Meta will need to differentiate itself, perhaps with lower prices or integration with its social media and messaging ecosystem. For example, companies could use Meta's infrastructure to train models with data from Instagram or WhatsApp, though this raises privacy questions.
• Regulation: Meta's use of customer data could raise privacy concerns, something the company will need to address. The U.S. Federal Trade Commission (FTC) already has Meta in its sights after previous fines for privacy violations. In the EU, GDPR imposes strict restrictions on data transfer and processing.
• Impact on the job market: Meta's expansion into cloud could create new jobs in engineering, sales, and support, but could also pressure existing providers to cut costs, potentially leading to layoffs.

"Meta seeks to turn its excess AI computing into a profitable business, following SpaceX's example with Starlink," notes TechCrunch. However, unlike SpaceX, Meta faces a highly regulated and competitive market with established players that have long-term contracts with large enterprises.

Historical context

This is not the first time a tech giant has sold its excess infrastructure. Amazon did the same by creating AWS from its internal capacity in 2006, transforming the IT industry. Google has also sold its cloud infrastructure, though with less initial success. Meta, with its massive AI investment, appears to be following a similar path. However, the current cloud market is much more mature and competitive than in 2006 when AWS launched its first services. Moreover, Meta is not starting from scratch: it already has experience offering services to developers through its advertising platform and APIs. But the scale of the cloud business is different: it requires an enterprise sales team, regulatory compliance, and service-level agreements (SLAs) that Meta does not yet have. A closer precedent is Microsoft, which transformed its server business into Azure, but that took years and billions in investment. Meta could accelerate the process by acquiring a smaller cloud company or partnering with an existing cloud provider, though this has not been mentioned in leaks. In any case, Meta's move reflects a broader trend: tech giants are seeking new ways to monetize their massive AI investments, whether through subscriptions (like Microsoft Copilot), advertising, or infrastructure.