Key Takeaways
- Meta’s stock rose nearly 9% after reports that the company is exploring a cloud‑computing business that would sell access to its AI models and excess data‑center capacity.
- The initiative would leverage Meta’s massive AI‑focused infrastructure—estimated at 20 GW of current capacity with an additional 14 GW planned—to compete with established cloud providers such as AWS, Google Cloud, and Azure.
- Meta could offer its own “Muse Spark” models (and possibly raw compute) via an API‑style service, similar to Amazon Bedrock, while also selling spare processing power to third parties.
- Analysts note that Meta’s data‑center footprint already rivals that of major cloud players, giving it a credible chance to capture a share of the fast‑growing AI compute market.
- The market reaction punished pure‑play “neocloud” stocks (CoreWeave down ~14%, Nebius down ~17%), reflecting investor concerns that Meta’s entry could intensify competition.
- CEO Mark Zuckerberg had previously signaled that selling compute is “definitely on the table,” noting frequent external requests for API access or compute purchases, and viewing excess capacity as a strategic option if internal demand wanes.
Meta’s shares jumped almost 9% on Wednesday after Bloomberg reported that the social‑media giant is preparing to enter the cloud‑computing arena by monetizing the vast computing resources it has amassed for its AI ambitions. The move would allow Meta to sell access to its own AI models—such as the internally developed Muse Spark—as well as surplus “raw” processing power to external customers, mirroring offerings like Amazon Web Services’ Bedrock platform. By running the data centers and chips that power these models, Meta would charge developers for API‑based access, creating a new revenue stream that taps into the insatiable demand for AI‑driven compute.
The rationale behind the potential shift stems from Meta’s aggressive investments in data‑center infrastructure over the past few years. Analyst Madison Rezaei of Bernstein highlighted that the company already controls roughly 20 gigawatts (GW) of global capacity, with plans to bring another 14 GW online in the coming years. This scale, she noted, “easily rivals cloud provider footprints,” positioning Meta to become a formidable player in a market currently dominated by Amazon, Google, and Microsoft. The prospect of a new entrant with such resources sent shockwaves through specialized cloud firms: CoreWeave’s shares fell nearly 14% and Nebius dropped about 17% as investors reassessed the competitive landscape.
Meta has not officially confirmed the report, but CEO Mark Zuckerberg had previously hinted at the idea. In a May earnings call, he told investors that selling compute is “definitely on the table,” citing a steady stream of outside companies requesting either API services or the chance to purchase Meta’s compute at a premium. Zuckerberg added that the company has not yet pursued such sales because it currently needs the capacity for its own AI workloads, but that the option remains attractive should internal demand ever slacken. This flexibility, he argued, bolsters confidence in continued investment in data‑center expansion.
If Meta follows through, the impact could be twofold. First, it would diversify the company’s revenue beyond advertising, reducing reliance on a single volatile income stream. Second, by offering its proprietary AI models as a service, Meta could deepen ecosystem lock‑in, encouraging developers to build on its platform and thereby reinforcing its AI research advantages. The move also aligns with broader industry trends where cloud providers are increasingly bundling AI model access with raw infrastructure to capture higher‑margin workloads.
Market observers caution that success will depend on execution. Meta must navigate the complexities of multi‑tenant cloud security, performance guarantees, and pricing models that compete with entrenched players who benefit from mature ecosystems and extensive customer relationships. Moreover, the company will need to balance internal AI research priorities with external sales commitments, ensuring that allocating capacity to third parties does not impede its own product roadmap.
Nevertheless, the enthusiastic investor response—reflected in the stock’s upward swing—suggests that the market perceives strategic value in Meta’s latent compute assets. As AI continues to drive unprecedented demand for processing power, Meta’s potential to become a significant cloud compute provider could reshape competitive dynamics, prompting existing cloud giants to accelerate their own AI‑focused offerings and possibly prompting further consolidation or partnership activity in the sector.
AI is moving fast. Axios AI+ keeps you ahead. Sign up free at Axios.com.