# Meta becomes an AI vendor, a confession disguised as strategy > On July 1, 2026, Meta Platforms confirmed it is developing a new cloud business called Meta Compute, designed to resell its surplus - Author: Maxime Marquette - Publisher: MadMax (https://mad-max.co) - Section: Analysis - Published: 2026-07-06 - Canonical URL: https://mad-max.co/en/article/meta-devient-vendeur-d-intelligence-artificielle-un-aveu-deguise-en-strategie ## Introduction: the day Meta stopped being just an AI buyer ### An announcement that made Wall Street jump On **July 1, 2026**, **Meta Platforms** confirmed it is developing a new cloud business called **Meta Compute**, designed to resell its surplus computing capacity as well as access to its **artificial intelligence** models, notably **Muse Spark**, according to information reported by **Bloomberg** and confirmed by **Reuters**. The company's stock jumped nearly **9 percent** that same day, closing at an all-time high. This announcement marks a strategic **turning point** for a company that, until now, stood apart from the other three American tech giants by having no commercial cloud division whatsoever, while **Amazon**, **Microsoft**, and **Google** have dominated this lucrative market for years. ### Why this decision deserves a closer look I believe this announcement, presented by several outlets as a simple case of smart diversification, deserves a more critical examination. Selling off excess capacity is not necessarily a sign of strength: it could just as easily reveal that internal demand for **Meta's** **artificial intelligence** is not keeping pace with its breakneck infrastructure spending. This analysis examines the economic motivations, the fallout on **financial markets**, the reaction of semiconductor makers, and the broader geopolitical implications of this strategic pivot in the global technology race. *I remain fascinated by how quickly Wall Street can interpret the same move in two opposite ways: a sign of strategic genius one day, an admission of overinvestment the next. The truth, as usual, probably lies somewhere in between.* ## What Meta Compute actually reveals ### Two distinct business models in the works According to **Bloomberg**, **Meta Compute** is weighing two complementary commercial approaches: the first involves offering programmatic access to models hosted on **Meta's** **infrastructure**, including **Muse Spark** and the **Llama** family, an approach comparable to **Amazon Web Services'****Bedrock** service. The second involves directly renting out raw computing capacity, in the manner of specialized providers like **CoreWeave**. This **dual architecture** would let **Meta** target both developers looking for a turnkey solution and more sophisticated tech companies wanting direct control over their own software stack on cutting-edge hardware. ### An internal initiative already in place for months **Meta Compute** already existed as an internal effort to coordinate the company's infrastructure ambitions, co-led by **Santosh Janardhan**, head of global infrastructure, and **Daniel Gross**, a well-known investor in the **artificial intelligence** sector, according to information picked up by several tech outlets. **Meta** is said to have already amassed more than **600,000 graphics processing units** equivalent in power to **Nvidia's** H100 **chips** across its data centers, a computing fleet larger than that of most companies not specialized in cloud services worldwide. *Stockpiling the equivalent of 600,000 top-tier chips is a colossal bet. Turning it into a revenue stream rather than pure waste follows a commercial logic I understand, without applauding it unreservedly.* ## The implicit admission behind Zuckerberg's decision ### Remarks that date back several weeks **Mark Zuckerberg** had already raised this possibility at **Meta's** annual **shareholder** meeting on **May 27, 2026**, stating, according to **CNBC**, that launching a cloud business was **"definitely on the table."** He noted that outside companies were approaching **Meta** almost every week seeking access to its models or its computing power. That May statement, relatively little commented on at the time, takes on new significance in light of the July announcement: it confirms that **Zuckerberg** was already anticipating, weeks earlier, the possibility that his company had **"overbuilt"** its **computing capacity** relative to its actual internal needs. ### A more critical reading from some analysts **Jochen Stanzl**, an analyst at **Consorsbank**, quoted by several European news agencies, summed up this ambivalence with a striking line: this announcement amounts to **"an admission that demand for its own artificial intelligence applications isn't broad enough to actually use the computing power it built so quickly."** This critical reading contrasts with the immediate stock-market enthusiasm, illustrating the financial markets' persistent difficulty in properly assessing massive infrastructure strategies in the **artificial intelligence** sector, where technological optimism constantly rubs shoulders with doubt about actual profitability. *I find Stanzl's quote particularly clear-eyed. The market first celebrated the news, then some more attentive investors began wondering whether it was actually a warning sign dressed up as a business opportunity.* ## The staggering scale of Meta's AI investments ### An infrastructure budget larger than several countries' GDP **Meta** plans to spend between **125 and 145 billion dollars** on **artificial intelligence** infrastructure during **2026**, according to the revised budget forecasts shared during its earnings release in April. According to **Forbes**, that sum exceeds the gross domestic product of several small nations. This budgetary scale fits into a broader race among American tech giants, where the sector's cumulative **artificial intelligence** infrastructure spending could reach **725 billion dollars** this year according to estimates cited by **Cryptopolitan**. ### Rapidly expanding energy capacity According to **GIGAZINE**, **Meta** already has roughly **20 gigawatts** of energy capacity worldwide to power its data centers, with expansion plans aiming to add roughly **14 more gigawatts** in the years ahead, a level that would bring the company closer to the operating scale of established cloud providers. This massive energy expansion raises legitimate questions about the long-term sustainability of such an intense infrastructure race, at a time when several American local communities are beginning to worry about the impact of these data centers on their regional power grids. *Twenty gigawatts is a scale that boggles the mind for most people, myself included. This race for energy to feed artificial intelligence deserves a far broader public debate than the one it is currently getting.* ## The immediate reaction from financial markets ### A sharp drop in semiconductor stocks Meta's announcement triggered a notable drop in semiconductor makers' stocks, particularly in **South Korea**, where shares of **Samsung** and **SK Hynix** fell in the days following the announcement, according to information reported by the business daily **Chosun**. American manufacturer **Micron** reportedly fell more than **10 percent** on July 1, according to market data cited on specialized financial networks. This negative reaction reflects a new investor worry: if a player as advanced as **Meta** ends up with excess computing capacity to sell off, that could signal that global demand for **artificial intelligence** semiconductors may not be as insatiable as the market had assumed until now. ### A challenge to the perpetual-shortage thesis The premise underpinning the stock-market growth of the **artificial intelligence** sector rested largely on the idea that demand would indefinitely outstrip available computing supply. Meta's announcement shakes that thesis, with several analysts cited by **Chosun** now raising legitimate doubt about a possible **oversupply** rather than a persistent shortage of computing capacity. This narrative shift, if confirmed, could have lasting repercussions on stock valuations across the entire **artificial intelligence** supply chain, from chipmakers to specialized memory suppliers. *I am wary of binary narratives in finance: yesterday perpetual shortage justified every valuation, today sudden abundance calls them all into question. Economic reality is probably more nuanced than these two extremes.* ## The competitive backdrop with established hyperscalers ### Meta joins a race already well underway, and late **Meta** becomes, according to **CNBC**, the only one of the four big American tech players without an already-established commercial cloud division, a gap it is now seeking to close by taking on competitors as entrenched as **Amazon Web Services**, **Microsoft Azure**, and **Google Cloud**. This late entry carries real risks: these three competitors already generate tens of billions of dollars in quarterly revenue from their mature **cloud infrastructures**, with business relationships built over more than a decade with corporate clients around the world. ### A potential pricing edge through depreciation According to an analysis by **Enterprise DNA**, several analyst estimates suggest **Meta** could afford to offer prices **20 to 30 percent** lower than established cloud providers, given that the company has already paid for this **infrastructure** to power its own products and could now spread that cost across additional external revenue. This potential pricing edge, if it materializes, could significantly disrupt the pricing structure of the entire cloud market dedicated to **artificial intelligence**, forcing established players to adjust their own business models in the face of this new competition. *I acknowledge Meta's cost advantage is real on paper, but I remain skeptical of its ability to quickly rival the operational maturity and customer trust built over years by Amazon, Microsoft, and Google.* ## The SpaceX precedent, a model explicitly copied ### Meta follows a strategy already tested by Musk According to **CNBC**, **Meta** is explicitly following the example of **SpaceX**, **Elon Musk's** company, which also began reselling its excess **computing capacity** this year, notably to companies like **Anthropic**, said to be paying roughly **1.25 billion dollars a month** for this access, and **Google**, said to be paying roughly **920 million dollars a month** according to the same source. This precedent shows that **monetizing excess capacity** is no longer an isolated exception but is becoming an increasingly common practice among tech companies that have invested massively in their own **artificial intelligence** infrastructure without necessarily using it at full capacity right away. ### A trend that could accelerate among other tech giants If this trend becomes widespread, other tech companies with significant underused computing capacity could be tempted to follow a similar path, gradually turning the cloud market dedicated to **artificial intelligence** into a far more fragmented and competitive ecosystem than it was just months ago. This potential fragmentation could, in time, benefit end consumers and smaller tech companies through a general drop in prices, even as it considerably complicates the strategic reading of the sector for institutional **investors**. *I find it telling that even the richest and most ambitious companies in American tech, from SpaceX to Meta, arrive at the same conclusion: better to sell the surplus than let it sit idle.* ## What this means for the West's technological position ### A consolidation of American dominance in AI infrastructure In my view, this diversification by **Meta** into cloud computing reinforces the dominant position of American companies in global **artificial intelligence** infrastructure, at a moment when international competition, particularly from China, is actively trying to close its gap in advanced computing capacity. This American consolidation of the underlying infrastructure of **artificial intelligence** remains a major strategic asset for the **West** in its global technological competition with rival powers like **China**, whose tech companies are developing domestic alternatives due to their lack of full access to **Nvidia's** most advanced chips. ### A risk of excessive concentration of technological power This growing concentration of global **artificial intelligence** infrastructure in the hands of a limited number of American companies, however reassuring it may be geopolitically in the face of **China**, also raises legitimate questions about the concentration of economic and technological power within the Western ecosystem itself. This tension between Western geopolitical solidarity and healthy internal **competition** among American companies deserves close monitoring by regulators, who will need to balance support for innovation against preventing excessive dominant positions in this strategic market. *I remain unreservedly pro-Western when it comes to China, but I refuse to be naive about the risks of excessive concentration of technological power in just a few American hands. Both concerns can, and must, coexist.* ## The lingering doubts about actual profitability ### A strategy still unclear on several essential points According to **Bloomberg** and several other sources, **Meta Compute's** plans remain under active development and could still evolve significantly before any official commercial launch. No launch date, no precise pricing structure, and no confirmed customer pipeline have been made public at this stage. This persistent uncertainty, acknowledged by **Meta** itself through its lack of a detailed official comment, illustrates the still-preliminary nature of this initiative, despite the scale of the immediate stock-market reaction it triggered. ### Skepticism from some financial analysts Analyst **Mandeep Singh**, quoted during a **Bloomberg TV** broadcast, suggests this announcement could also reflect internal pressure at **Meta** to demonstrate a concrete return on its massive infrastructure spending, as its consumer **artificial intelligence** agent offerings are taking longer than expected to generate significant revenue at scale. This reading, if confirmed in coming quarters, would suggest that monetizing excess capacity responds as much to an immediate financial necessity as to a genuine long-term strategic ambition in the cloud sector. *I always prefer cautious analysis over hasty enthusiasm. Until Meta confirms pricing, a launch date, and signed pilot customers, I will treat this announcement as a strategic intention rather than an achieved commercial success.* ## The implications for AI startups and developers ### Potentially wider access to scarce computing resources For startups specialized in artificial intelligence, often facing considerable wait times to access cutting-edge graphics processing units, the arrival of another player like **Meta** in this market could offer welcome relief, diversifying the available supply beyond the three historic hyperscalers. This diversification of supply, if it materializes at competitive prices, could accelerate innovation among smaller Western tech companies, which often struggle to compete with larger companies for priority access to the most advanced computing resources. ### Direct access to the Llama and Muse Spark models Beyond simply renting raw capacity, the potential offer of direct access to the **Llama** and **Muse Spark** models hosted by **Meta** could attract developers seeking an alternative to **OpenAI's** or **Google's** proprietary models, particularly those who value **Meta's** partially open approaches to artificial intelligence. This dimension of the project, less discussed than the simple sale of raw capacity, could prove strategically more significant in the long run, positioning **Meta** as an unavoidable player in the Western artificial intelligence application development ecosystem. *I believe wider access to Llama models could, over time, prove more strategically important than simply renting raw computing capacity, cementing Meta's influence at the heart of the Western artificial intelligence development ecosystem.* ## Comparing this to historical precedents in the tech sector ### Amazon Web Services, a precedent clearly inspiring Meta The trajectory of **Amazon**, which turned its excess internal infrastructure into **Amazon Web Services**, since become the group's main profit engine, clearly serves as the reference model that **Meta** is seeking to replicate with **Meta Compute**, according to several analysts cited by specialized financial outlets. This historical analogy, while appealing, has its limits: **Amazon Web Services** took more than a decade to become this dominant, in a cloud market that was then far less mature and competitive than the one **Meta** is about to face today. ### Lessons to draw from this historical comparison Replicating **Amazon Web Services'** success would require considerable strategic patience from **Meta** and sustained investment over several years, at a time when financial market expectations for a quick return on artificial intelligence investment remain particularly high and impatient. This tension between necessary strategic patience and the impatience of today's financial markets could be one of the main internal challenges **Meta** will have to manage to turn this initial announcement into a truly lasting commercial success. *Amazon Web Services' success took more than ten years to fully materialize. I seriously doubt that today's financial markets, more impatient than ever, will grant Meta the same generous time horizon to prove the value of this diversification.* ## The strategic role of Daniel Gross and the leadership team ### Oversight entrusted to recognized figures in the sector The co-leadership of **Meta Compute** by **Santosh Janardhan** and **Daniel Gross**, the latter recognized as an influential investor and operator in the artificial intelligence ecosystem, shows, according to several observers, the strategic seriousness **Meta** is placing on this new commercial initiative. This dedicated governance structure, separate from Meta's traditional infrastructure teams, suggests a clear intent to treat **Meta Compute** as a genuine standalone business unit rather than a minor side project attached to the company's historic advertising business. ### An ambition stated on a very long time horizon According to information picked up by **Enterprise DNA**, **Zuckerberg** reportedly spoke of infrastructure ambitions running into **"tens of gigawatts this decade, and hundreds of gigawatts or more over time,"** a scale that far exceeds the immediate needs of **Meta's** current products and confirms a very long-term vision for this strategic diversification. This outsized ambition, if realized, would place **Meta** among the most important cloud players on the planet by the end of the decade, an audacious goal given the company's current lag behind its three main established competitors. *Hundreds of gigawatts down the line is an ambition bordering on the excessive. I respect the audacity, but I seriously question whether Western power grids can even absorb such growth without major strain.* ## What Wall Street will be watching in the coming months ### The key indicators to watch next Investors will closely watch, in the coming months, whether an official launch date for **Meta Compute** is confirmed, along with the first concrete pricing signals and the identification of significant pilot customers, all elements that will help determine whether this July announcement amounted to a genuine strategy or simply a financial communications maneuver. **Meta's** next quarterly earnings release should also provide valuable insight into how infrastructure spending is evolving and how the company plans to concretely structure this new commercial division financially. ### A credibility test for the entire AI sector Beyond Meta's specific case, this sequence of events amounts to a broader credibility test for the entire artificial intelligence sector, whose stock-market valuation rests largely on a promise of continuously growing demand that, according to some recent signals, is beginning to be questioned more seriously by investors who were previously unanimously enthusiastic. This period of heightened scrutiny could prove healthy for the sector in the long run, forcing more rigorous financial discipline on tech companies whose infrastructure spending had, in recent years, sometimes escaped sufficiently critical examination by the markets. *I think a bit more healthy skepticism from financial markets toward massive artificial intelligence infrastructure spending would be broadly beneficial, including for the companies themselves, forced into greater financial discipline.* ## The question of Western technological sovereignty ### An issue that reaches far beyond Meta's case This strategic pivot by **Meta** fits into a broader debate over Western **technological sovereignty** in the face of Chinese competition and growing dependence on a limited number of semiconductor suppliers, an issue that American and European governments have been watching with growing attention for several quarters. The ability of major Western tech companies to maintain their lead in computing infrastructure is, according to several technology geopolitics analysts, a decisive factor in preserving Western leadership in the global race for artificial intelligence. ### A signal sent to strategic rivals By consolidating its computing capacity and now offering it to other Western players in the sector, **Meta** is also sending a signal to its strategic rivals, notably China, whose tech companies are actively trying to close their gap in artificial intelligence infrastructure despite American restrictions on exporting advanced semiconductors. This positioning fits directly into the logic championed by several Western officials, according to which technological dominance should remain in the hands of allied democracies rather than authoritarian regimes seeking to close their gap by any means available. *I believe this geopolitical dimension is too often overlooked in media coverage of these business announcements: behind every one of Meta's infrastructure decisions also plays out a chapter in the technological rivalry between the West and its strategic rivals.* ## Conclusion: a strategic pivot with still-uncertain consequences ### A decision that embodies the contradictions of the AI era The announcement of **Meta Compute** perfectly embodies the current contradictions of the artificial intelligence sector: a company pouring staggering sums into infrastructure while simultaneously trying to monetize a surplus it implicitly admits it cannot fully use for its own products in the short term. This tension between outsized ambition and immediate commercial pragmatism will likely continue to define the Western tech sector in the months ahead, as investors seek a healthier balance between technological optimism and traditional financial rigor. ### A story to watch very closely in the coming months I will continue to closely follow the evolution of this initiative, between eventual confirmation of a concrete commercial launch and the persistent skepticism of the most cautious financial analysts, a story emblematic of the current tensions running through the entire Western artificial intelligence ecosystem. *I end this analysis convinced that the coming months will separate the Western tech companies able to turn their infrastructure promises into concrete results from those that simply rode the passing financial hype surrounding artificial intelligence.* **By Maxime Marquette, columnist** ## Columnist's transparency note ### Editorial stance I am not a journalist but a **columnist and analyst**. My expertise lies in observing and analysing the geopolitical, economic and strategic dynamics that shape our world. My work is to take political strategies apart, to understand global economic movements, to put the decisions of international actors in context and to offer analytical perspectives on the changes that are redefining our societies. I do not claim the cold objectivity of traditional journalism, which stops at reporting facts. **I claim analytical clarity**, rigorous interpretation and a deep understanding of the complex issues that concern us all. My role is to give meaning to the facts, to place them in their historical and strategic context, and to offer a critical reading of events. ### Method and sources This text keeps the basic distinction between **verified facts and interpretive analysis**. The rule never changes: a factual claim is published only when a verifiable source supports it, and the sources this article actually used are listed under **Sources**, never here. **Primary source families** used by the publication, when the story has them: official statements from governments and international institutions, public statements by political leaders, reports from intergovernmental organisations, dispatches from recognised international news agencies. **Secondary source families**: specialised publications, internationally recognised news media, analyses from established research institutions, reports from industry organisations. When an article cites statistical, economic or geopolitical data, it comes from the institutions that produce it (intergovernmental organisations, central banks, national statistics offices), and the exact institution is named under **Sources**. ### Nature of the analysis The analyses, interpretations and perspectives in the analytical sections of this article are a **critical, contextual synthesis** based on the information available, the trends observed and the expert comments cited in the sources consulted. My role is to interpret these facts, to set them in the context of today's geopolitical and economic dynamics, and to give them a coherent meaning within the larger story of the changes shaping our time. These analyses reflect expertise built through continuous observation of international affairs and an understanding of the strategic mechanisms that drive global actors. This article describes a documented state of affairs on its publication date, not a prediction: later developments may shift its outlook. No update is promised in advance; when an article is corrected or completed, the change is dated in the text. ## Sources ### Primary sources - Meta building cloud business to sell excess AI capacity — Reuters, July 1, 2026 - Meta Is Planning a Cloud Business to Sell AI Computing Power — Bloomberg, July 1, 2026 ### Secondary sources - Meta stock pops on cloud push to sell excess AI compute — CNBC, July 1, 2026 - Meta's AI Cloud Entry Triggers Semiconductor Stock Drop — Chosun, July 2, 2026 - Meta Makes Cloud Play To Sell Excess AI — Forbes, July 2, 2026 - Meta planning cloud infrastructure business — GIGAZINE, July 2, 2026 - Mark Zuckerberg says a Meta cloud computing business 'definitely on the table' — CNBC, May 27, 2026 --- Source: Maxime Marquette, MadMax. When citing, name the author and link https://mad-max.co/en/article/meta-devient-vendeur-d-intelligence-artificielle-un-aveu-deguise-en-strategie