By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
While Meta's AI demands and hyperscalers' outsourcing indicate near-term neocloud demand, the panel agrees that Meta's cloud ambitions are unlikely to disrupt the market immediately. Key risks include regulatory scrutiny on hyperscalers bundling data with AI training and potential antitrust issues if Meta controls the base AI model. The opportunity lies in neocloud providers' ability to extract premium pricing for the next 24-36 months.
Risk: Regulatory scrutiny on hyperscalers bundling data with AI training and potential antitrust issues if Meta controls the base AI model
Opportunity: Neocloud providers' ability to extract premium pricing for the next 24-36 months
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Meta Platforms (NASDAQ: META) has caused some angst for neocloud stockholders. The company allegedly plans to launch its own cloud business to lease access to AI computing power and AI models, and given the scale of the data centers it's constructing, it could become a serious competitor in the space.
Those concerns intensified after reports emerged that Meta Platforms and Anthropic are discussing a compute deal for up to $10 billion that would run over 2 years. No official announcements have been made, so there are no details yet about how many gigawatts such a deal could involve.
Where to invest $1,000 right now? Our analyst team just revealed what they believe are the 10 best stocks to buy right now, when you join Stock Advisor. See the stocks »
Neither company has confirmed the talks, which were described as early-stage conversations by the unnamed source. However, it appears that Meta Platforms is serious about entering the cloud infrastructure industry, and its entry could affect key players.
While Meta Platforms is building a 5-gigawatt data center in Louisiana and gobbling up as much computing hardware as possible, AI demands present a challenge for the company. Many investors are wondering how it will be able to reliably sell computing power when it is buying compute heavily from other companies for its own needs.
For instance, Meta Platforms recently signed a five-year, $27 billion deal with Nebius (NASDAQ: NBIS) for additional compute.
However, Meta Platforms might have plenty of compute available within the next few years. In January, Mark Zuckerberg shared in a Threads post that he expects the company to build tens of gigawatts of AI infrastructure this decade, and "hundreds of gigawatts or more over time."
Meta Platforms still needs compute for its social networks, large language models (LLMs), and AI glasses. Since the company closed Q1 with 3.56 billion daily active users across its social networks, it has greater compute needs than almost any other company.
Meta Platforms has used the copycat model to take out rival platforms like Vine and limit Snapchat's growth prospects, but it likely won't play out that way with the neoclouds. The AI industry is moving too quickly, and Meta Platforms' own compute needs will continue to grow. By the time it gets its cloud leasing business off the ground at meaningful scale, the winners of the industry may have already been decided.
Hut 8 (NASDAQ: HUT) recently announced that it had expanded an existing contract with a "high-investment-grade tenant." The updated lease has a 15-year term covering 352 megawatts. The contract's total value is $9.8 billion, averaging $653 million per year. The company's 1-gigawatt Beacon Point AI data center is poised to produce more than $1.75 billion in net operating income from its contracts.
Meta Platforms is one of several companies that fit the "high-investment-grade tenant" description. If it was Meta Platforms expanding its contract, that would be a clear signal that the social media company still needs a lot more compute than it has online, in which case it won't be in a good position to sell capacity for a while. However, if the unnamed tenant wasn't Meta Platforms, the deal still highlights something investors should pay attention to: Companies that need AI compute aren't waiting around for Meta to figure out its cloud segment.
Meta Platforms is a hyperscaler already, and if it has enough excess compute that it can start selling some of it, it theoretically won't have to buy compute from competitors. Granted, Meta Platforms would actually have to build tens of gigawatts' worth of AI data centers this decade, but investors should consider what would happen if this scenario is realized.
The extremely pessimistic take for the neoclouds is that Meta Platforms and other hyperscalers will build enough of their own data centers to fully meet their needs, after which they will not need neoclouds like Nebius or CoreWeave. Those companies have lucrative contracts with the giants right now, but they may be renegotiated to less profitable rates or simply not renewed at all if hyperscalers have sufficient capacity within their own AI data centers.
Compute seller Iren (NASDAQ: IREN) has demonstrated that it can diversify its customer base beyond the hyperscalers. The company announced this week that it had secured $2.8 billion in new customer contracts with "leading AI developers." The number of megawatts was not specified, but the average contract length is four years.
The company now has 10 customers, including "a new leading AI developer," the name of which it may reveal in its upcoming earnings report. Iren also raised its 2026 annual recurring revenue target to over $4 billion, indicating that demand is not slowing.
Even if hyperscalers were to wind down their usage of neocloud providers, that might not be a problem for them. Neoclouds like Nebius could ink agreements to sell their critical IT load to new customers one to two years in advance, getting those commitments on the books ahead of the expirations of their current contracts. AI developers and hyperscalers have already demonstrated they are willing to wait that long for power.
With that in mind, investors shouldn't be worried about the latest headlines around Meta Platforms. If the neoclouds need to look elsewhere for customers, they should have little trouble finding them.
Before you buy stock in Meta Platforms, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Meta Platforms wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $364,562! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,247,668!
Now, it’s worth noting Stock Advisor’s total average return is 894% — a market-crushing outperformance compared to 207% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of July 22, 2026. *
Marc Guberti has positions in Iren. The Motley Fool has positions in and recommends Meta Platforms. The Motley Fool has a disclosure policy.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
Four leading AI models discuss this article
"META's own compute hunger and the industry's speed ensure neocloud long-term leases remain secure through at least 2027-28, outweighing any distant hyperscaler self-sufficiency risk."
The $10B potential Meta-Anthropic compute lease over 2 years signals hyperscalers are still outsourcing AI capacity aggressively, supporting neocloud demand in the near term. META's 5GW Louisiana buildout and Zuckerberg's "tens of gigawatts this decade" rhetoric underscore its own insatiable needs (3.56B DAUs, LLMs, AR glasses), making near-term cloud-selling unlikely. Deals like HUT's $9.8B/15yr 352MW lease and IREN's $2.8B diversification to 10 AI customers (raising 2026 ARR >$4B) show the market is booking capacity years ahead. NBIS's $27B 5yr Meta contract further illustrates ongoing reliance. Article correctly downplays immediate risk; AI growth pace favors specialized neoclouds.
If META successfully scales to hundreds of GW and launches its own cloud business, it could flip from largest buyer to competitor, forcing neoclouds like CoreWeave, Nebius, and IREN into margin compression or non-renewals once hyperscalers internalize capacity—especially if power constraints ease and copycat execution accelerates.
"Meta’s reliance on third-party compute providers proves that the supply-demand imbalance in AI infrastructure is structural and will favor specialized neoclouds for several years."
The narrative that Meta is pivoting to a cloud-leasing competitor is premature and ignores the capital intensity of infrastructure. Meta’s $27 billion commitment to Nebius underscores a structural deficit in their own internal compute, not an excess. If Meta were truly building a multi-gigawatt cloud business, their CapEx would need to scale exponentially beyond current guidance, likely compressing margins. The real story isn't Meta becoming a cloud provider; it's the 'neoclouds' (IREN, NBIS, HUT) acting as necessary utility-like bridges to bridge the gap between AI demand and the multi-year lag in massive data center construction. I remain neutral on Meta as a cloud play but bullish on the neocloud sector's ability to extract premium pricing for the next 24-36 months.
If Meta successfully executes vertical integration of their silicon and power supply, they could commoditize the neoclouds, turning high-margin compute leasing into a race-to-the-bottom utility business.
"Meta's cloud threat is real but distant; the immediate risk to neoclouds is demand cliff or customer concentration loss, not competition from a company still aggressively *buying* external compute."
The article frames Meta's potential cloud ambitions as a long-term threat to neocloud providers, but the math doesn't support urgency. Meta signed a $27B/5-year deal with Nebius weeks ago—if internal capacity were imminent, why lock in external supply at that scale? The $10B Anthropic deal (unconfirmed, early-stage) is a rounding error against Meta's stated 'tens of gigawatts' target. More critically: the article conflates *building* capacity with *monetizing* it. Operating a cloud business requires sales infrastructure, support, pricing discipline, and competing against AWS/Azure/GCP. Meta's core competency is social platforms and ads, not infrastructure services. Neocloud providers face real risks, but not from Meta's cloud ambitions—from AI demand saturation or capex discipline tightening.
If Meta achieves even 50% of Zuckerberg's stated capex goals and successfully monetizes excess capacity at scale, neocloud valuations could compress sharply due to supply shock and margin pressure across the sector.
"The material risk to neoclouds hinges on Meta’s ability to monetize cloud at scale and with healthy margins, not on a rumored deal, and the article’s capacity numbers are likely exaggerated, making near-term risk overstated."
Stories front-run headlines with a 10B, two-year compute deal as if it will reshape markets. But capacity figures cited (5 GW site, tens of GW this decade) look hyperbolic for a single campus; even the largest data centers are measured in hundreds of MW. The real risk is economics and timing: can Meta monetize cloud at healthy margins, and on what basis? A potential Anthropic deal may reflect internal workloads more than a scalable external cloud offering. If Meta's cloud ambitions stay anchored to internal needs or narrow commercial agreements, neoclouds’ leverage could persist. The ramp and capital constraints give incumbents room to adjust pricing and contracts over time.
Even if Meta signs such a deal, the revenue flow and margins are unproven; scale alone does not guarantee displacing neocloud incumbents.
"Regulatory and data-privacy hurdles on Meta's AI training could prolong neocloud premium pricing window."
Claude correctly flags the sales/support gap for Meta entering cloud, but misses that the $27B Nebius deal is likely a mix of committed capacity and potential resale—Meta already subleases excess to partners. The real overlooked risk: regulatory scrutiny on hyperscalers bundling social data with AI training could slow internal builds, extending neocloud runway beyond 36 months.
"Meta’s infrastructure strategy is a play for ecosystem dominance via Llama, not a traditional cloud-reselling business."
Claude and ChatGPT are missing the 'sovereign AI' angle. Meta isn't building a traditional cloud; they are building a private, proprietary training cluster that effectively functions as a walled garden. By leasing capacity to Anthropic, they aren't competing with AWS—they are building a coalition to standardize Llama as the industry's base layer. The risk isn't margin compression from commoditization; it's the antitrust trap of Meta controlling the infrastructure that powers their competitors' models.
"Meta's antitrust exposure stems from data+model control, not from infrastructure leasing alone."
Gemini's 'sovereign AI' framing is sharp, but conflates two separate risks. Meta leasing to Anthropic isn't antitrust exposure—it's just capacity arbitrage. The actual antitrust trap is if Meta uses training data from 3.56B DAUs to build Llama, then leverages infrastructure dominance to force adoption. That's the bundling story. Leasing spare cycles to a competitor doesn't trigger it; *controlling the base model* does. The regulatory risk exists regardless of cloud strategy.
"Power and permitting constraints could cap Meta’s expansion, delaying a disruptive external cloud shift and prolonging neocloud pricing power."
Challenging Grok: the near-term outsized capacity narrative hinges on power and permitting that may cap Meta’s expansion. Grid constraints, high energy costs, and lengthy approvals could slow internal scale to tens of GW, keeping external demand and pricing power with neoclouds longer. Nebius/HUT/NBIS look like bridge capacity, not a durable external market disruption. If Meta stalls, neoclouds could avoid margin compression only by sustained pricing power.
While Meta's AI demands and hyperscalers' outsourcing indicate near-term neocloud demand, the panel agrees that Meta's cloud ambitions are unlikely to disrupt the market immediately. Key risks include regulatory scrutiny on hyperscalers bundling data with AI training and potential antitrust issues if Meta controls the base AI model. The opportunity lies in neocloud providers' ability to extract premium pricing for the next 24-36 months.
Neocloud providers' ability to extract premium pricing for the next 24-36 months
Regulatory scrutiny on hyperscalers bundling data with AI training and potential antitrust issues if Meta controls the base AI model