Key Points:
- A prominent Wall Street brokerage issued five key takeaways evaluating Meta’s potential entry into renting out excess artificial intelligence computing capacity.
- Analysts believe a raw “neocloud” bare-metal service would pose significantly lower execution, technology, and hiring risks than a full-scale hyperscaler model.
- Leasing just 250 megawatts of excess first-party capacity at $40 per watt for a year could add $3, or 8%, to the company’s 2028 earnings per share.
- The company’s capital expenditures are projected to swell to $175 billion in 2027 and $205 billion in 2028 to bring online 3.5 gigawatts of new capacity.
The global technology landscape is closely watching a major strategic shift as social media giant Meta Platforms evaluates a potential entry into the cloud computing business. The planned initiative, internally dubbed “Meta Compute,” aims to monetize the company’s vast and rapidly expanding artificial intelligence infrastructure by renting excess graphics processing capacity to external developers. This development represents a direct competitive threat to both legacy hyperscale giants like Amazon Web Services, Microsoft Azure, and Google Cloud, as well as specialized, high-growth “neocloud” startups like CoreWeave and Nebius. The strategic shift turns what was previously viewed as a massive, non-revenue-producing capital expenditure pile into a potentially lucrative, high-margin revenue stream.
As the company explores this new market, management is reportedly considering two distinct operational pathways: a hosted model-access service similar to Amazon’s Bedrock, or a raw, neocloud-like bare-metal silicon offering. Analysts at a prominent global investment bank recently evaluated these options, outlining five critical takeaways regarding the company’s competitive advantages and financial hurdles. The bank’s researchers believe that renting out raw GPU capacity is much easier for the social media company to execute successfully than attempting to build a full-scale hyperscaler service. A raw computing rental model bypasses the heavy software engineering, customer support, and system integration costs that define the traditional cloud business.
In contrast, trying to construct a full-service hosted API platform carries substantial technology, hiring, and execution risks. To compete with the sophisticated software stacks of the big three cloud giants, the company would have to dramatically improve the performance of its own proprietary models. Currently, the company’s Muse model suite lags behind frontier systems like Google’s Gemini on prominent coding and third-party benchmark evaluations, including TerminalBench and SWE Bench Verified. Because building a competitive ecosystem of enterprise software tools and developer applications presents a high barrier to entry, analysts advise viewing any full hyperscaler ambitions with a cautious “show me” attitude.
The company’s sheer scale of infrastructure investment provides a powerful foundation for this new business unit. Financial models estimate that the company will bring online approximately 2 gigawatts of incremental owned-and-operated IT capacity in 2026, followed by an additional 3.5 gigawatts in 2027. This rapid expansion builds on top of an estimated year-end 2025 capacity base of roughly 3 gigawatts. While this massive footprint allows the company to lease out excess capacity temporarily, it still trails the long-term capital plans of the established cloud giants. For comparison, Amazon and Google will likely add 5 gigawatts and 9 gigawatts of IT capacity, respectively, in 2027 alone, demonstrating the immense scale of the competition.
The company’s extensive contracts with external infrastructure providers also complicate its ability to lease out this hardware. Over the past year, the company has signed multi-billion-dollar deals to rent approximately 2.5 gigawatts of computing power from third-party platforms, including CoreWeave, Nebius, Oracle, and Google Cloud, to keep pace with its immediate internal research needs. The bank’s analysis clarifies that the social media giant will not have the legal or operational flexibility to sublease these third-party systems. However, these external arrangements do free up the company’s first-party, self-built data centers, giving it the necessary breathing room to lease its own proprietary hardware to external clients.
From a financial perspective, a raw neocloud rental model offers highly attractive, near-term earnings accretion. Because computing power remains a scarce, highly coveted resource, the market is willing to pay premium rates for immediate access to high-performance GPUs. Recent transactions in the specialized cloud sector feature highly favorable terms, including shorter lease durations, smaller volume sizes, and dual-sided opt-out clauses. The bank’s financial analysts project that for every 250 megawatts of excess first-party capacity the company leases for one year at a rate of $40 per watt, it will add approximately $3—or roughly 8%—to its 2028 earnings per share estimates, providing a powerful financial boost.
Despite this lucrative upside, the investment bank emphasizes that the neocloud opportunity should be viewed as a temporary stopgap rather than a permanent, core business to scale. The firm maintains its positive, long-term investment rating on the stock based on the company’s ability to develop and monetize new consumer-facing products. The true drivers of sustainable, multi-year revenue growth remain the scaling of MetaAI, commercial messaging tools, diffusion-based creative models, and new subscription revenue streams. Utilizing excess compute to generate rental income acts as a valuable financial bridge, supporting the stock’s valuation multiple while these primary, consumer-facing software products undergo long-term development.
The bank’s long-term capital expenditure projections reflect this strategic orientation. The current financial model projects that the company’s capital spending will rise from $145 billion in 2026 to $175 billion in 2027, and eventually swell to $205 billion in 2028. Importantly, this model assumes that the company is constructing this massive infrastructure footprint to run its own first-party consumer products, rather than to support a permanent, public-facing cloud business. If the firm chose to transition into a permanent cloud competitor, it would have to commit billions of dollars in additional capital to build out software platforms, sales forces, and global enterprise support teams.
Ultimately, the potential arrival of this social media giant into the cloud computing market has already sent shockwaves through the technology sector, triggering sharp double-digit stock declines for specialized GPU providers like CoreWeave and Nebius. By proving that it can successfully monetize its excess hardware, the company has demonstrated impressive operational agility. However, the ultimate success of the program will depend on how the company manages the balance between its internal research requirements and its external commercial ambitions. The coming months will reveal whether the firm can successfully convert its massive capital investments into a durable, multi-billion-dollar cloud revenue engine.
