U.S. technology companies pushed their future data center lease commitments past $850 billion in the first quarter of 2026, according to a Bloomberg analysis of quarterly filings published June 24. The figure marks a 63% jump from a year earlier and a 31% increase from the previous quarter alone. Meta and Microsoft drove most of the new activity, while Oracle now holds the single largest pile of future lease obligations of any company tracked in the dataset.
The number is not spending that already happened. It is a stack of rent obligations that hyperscalers have signed up to pay landlords, colocation firms and power providers over roughly the next two decades. But the pace of growth, and the market’s reaction to it, says a lot about where the AI infrastructure race stands in mid-2026: still accelerating, increasingly expensive, and starting to draw real skepticism from investors and the communities hosting the buildings.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
The $850 Billion Number: What Just Happened
Bloomberg’s June 24 report tracked future data center lease commitments disclosed in the latest round of corporate filings from the largest cloud and social media companies. The headline figure, $850 billion, represents the total rent hyperscalers have committed to paying data center landlords for capacity that, in many cases, has not been built yet, according to Benzinga’s review of the filings.
Two numbers explain why this quarter stood out. The total grew by $570 billion compared with the same period a year earlier, a 204% increase. It also grew $200 billion versus the prior quarter, a 31% quarter over quarter. Bloomberg had already flagged the trend in March, when it reported commitments topping $700 billion after Meta and Microsoft each added roughly $50 billion in a single quarter. The June update shows the pace didn’t slow down. It sped up.
Meta, Microsoft and Oracle: Breaking Down Who’s Spending What
Three companies account for most of the growth in data center lease commitments this cycle, and each is playing a different game.
Meta added the most new commitments of any company in the quarter: $79 billion, a 76% jump from the prior quarter, pushing its total future lease obligations to roughly $182.9 billion, per filings reviewed by BiGGo Finance. Microsoft added $41 billion, up 26% quarter over quarter, bringing its total to about $196.6 billion. Oracle did not disclose a single new-commitment figure in the same format, but it now holds the largest total in the group: roughly $250 billion in future lease obligations, tied in part to site commitments linked to its cloud infrastructure contracts. Amazon added a comparatively modest $10 billion in new lease commitments during the quarter, a gap worth sitting with given how aggressively AWS has otherwise been spending on its own infrastructure.
| Company | New Commitments Added (Q1 2026) | Total Future Lease Commitments | Quarterly Change | Stock Performance, YTD 2026 |
|---|---|---|---|---|
| Meta | $79 billion | ~$182.9 billion | +76% QoQ | -10.38% |
| Microsoft | $41 billion | ~$196.6 billion | +26% QoQ | -17.43% |
| Oracle | Not separately disclosed | ~$250 billion (largest total) | Leads all holders | -28.33% |
| Amazon | $10 billion | Not disclosed in dataset | Smallest addition among top spenders | +7.14% |
| Industry total | +$200 billion | $850 billion | +31% QoQ / +204% YoY | — |
Source: Bloomberg quarterly filings analysis, June 24, 2026, as reported by Benzinga and BiGGo Finance. Stock performance figures are year-to-date as of publication.
Why Data Center Lease Commitments Jumped 204% in a Year
The short answer is capacity anxiety. Every major AI lab and cloud provider is racing to secure power and shell space years before it’s needed, because the alternative, showing up late to a site with available grid capacity, can mean a multi-year wait. Leasing lets a company lock in a building and its power allocation without carrying the multi-billion-dollar construction cost on its own balance sheet the way a fully owned data center would.
Oracle’s position illustrates the dynamic well. A large share of its roughly $250 billion in commitments traces back to site deals tied to its cloud infrastructure contract work, including capacity later associated with OpenAI’s data center buildout. Securing that scale of power and shell space years in advance requires exactly the kind of long-dated lease commitment now showing up in the aggregate $850 billion figure. Meta and Microsoft, meanwhile, are leasing aggressively to keep pace with training and inference demand for their own AI products, from Meta AI to Copilot.
Leases vs. Capex: Two Numbers Behind the Same AI Story
It’s worth separating two figures that get conflated in AI infrastructure coverage. Lease commitments are future rent obligations, money owed to a landlord over the life of a contract. Capital expenditure, or capex, is cash a company actually spends building or buying servers, chips and its own facilities in a given year. They overlap but they are not the same metric, and 2026 has produced record numbers on both fronts.
On the capex side, Amazon is guiding to roughly $200 billion in 2026 spending, Microsoft is tracking toward about $190 billion, Alphabet has guided to $175 billion to $185 billion, and Meta has guided to $115 billion to $135 billion. Combined, the four companies are on pace to spend $725 billion on capex in 2026, up 77% from 2025’s already record $410 billion, according to figures reported by Tom’s Hardware and Yahoo Finance. One analyst estimate puts the aggregate baseline capex bill for compute, data centers and power at $7.6 trillion between 2026 and 2031.
| Company | 2026 Capex Guidance | Primary Cloud Platform | Change vs. 2025 |
|---|---|---|---|
| Amazon | ~$200 billion | AWS | Part of $725B combined, +77% |
| Microsoft | ~$190 billion | Azure | Part of $725B combined, +77% |
| Alphabet | $175–185 billion | Google Cloud | Part of $725B combined, +77% |
| Meta | $115–135 billion | Meta AI / infrastructure | Part of $725B combined, +77% |
| Combined (4 firms) | $725 billion | — | +77% vs. 2025’s $410 billion |
Source: Tom’s Hardware and CNBC, based on company earnings guidance for calendar year 2026.
Wall Street’s Verdict: Why the Biggest Spenders Have the Worst Stock Returns
Here’s the part that should worry AI infrastructure bulls. The three companies leading the data center lease boom, Oracle, Microsoft and Meta, are also the three worst-performing large-cap tech stocks of 2026 so far. Oracle is down 28.33% year-to-date. Microsoft is down 17.43%. Meta is down 10.38%. By contrast, Alphabet is up 14.20% and Amazon is up 7.14%, according to the stock data compiled by Benzinga.
That split maps almost exactly onto capital discipline. Alphabet and Amazon are spending heavily too, but investors appear more comfortable with how each is pacing its commitments relative to visible cloud revenue growth. Oracle’s stock decline is the sharpest, and it lines up with real balance-sheet strain: heavy debt-funded capex, thinner margins on its OpenAI-linked contracts, and now a round of layoffs meant to free up cash. Microsoft and Meta are seeing a milder version of the same skepticism: strong core businesses, but growing investor unease about whether AI-driven revenue is showing up fast enough to justify the lease and capex numbers both companies are now carrying.
The AWS Exception: Why Amazon Is Pumping the Brakes on New Leases
Amazon’s $10 billion in new Q1 lease commitments looks small next to Meta’s $79 billion, and that gap is deliberate. AWS owns roughly 24 million square feet of data center space and leases a similar amount, a more balanced mix than peers that lean harder on colocation deals. Wells Fargo analysts, cited in reporting from IT Pro, have observed that hyperscalers are growing more selective about leasing large power clusters and are shortening pre-lease windows for capacity due before the end of 2026. AWS has reportedly paused some of its own colocation commitments as part of that shift.
That doesn’t mean AWS is stepping back from AI infrastructure. Roughly 40% to 50% of AWS’s revenue backlog requires net-new data center capacity, which implies somewhere between 12 and 18 gigawatts of additional AWS-occupied capacity over the next three to five years. Amazon just prefers to own more of that capacity outright rather than lease it, a strategy the company has said it believes is cheaper to operate over a facility’s full 20-year lifespan. It is a bet on patience over speed, and so far the market has rewarded it: Amazon’s stock is one of only two among the group posting gains this year.
Oracle’s Trade-Off: Cutting 21,000 Jobs to Fund the AI Pivot
Oracle’s position at the top of the lease-commitment table came with a cost. On March 31, the company cut roughly 21,000 positions, trimming its headcount from about 162,000 to 141,000 employees, a 13% reduction delivered by mass email at 6 a.m., according to details reported by BiGGo Finance. Severance costs for fiscal 2026 came to $1.84 billion, up from just $374 million the year before, and unlike Meta, Microsoft or Cloudflare, Oracle canceled unvested equity immediately on termination. At least one affected employee reportedly lost close to $300,000 in unvested stock awards as a result.
The logic behind the cuts is straightforward even if the execution drew criticism: Oracle expects the layoffs to free up $8 billion to $10 billion a year, cash it plans to funnel toward roughly $70 billion in net capex for fiscal 2027 and $50 billion in AI data center spending this fiscal year alone. The company can point to results that partly justify the bet. Oracle posted record fiscal 2026 revenue of $67.4 billion, up 17%, with cloud services revenue surging 39% to $34 billion. Whether that growth curve holds long enough to justify a $250 billion lease book is the question investors are still pricing in, and Oracle’s 28% stock decline this year suggests they are not fully convinced yet.
The Neocloud Angle: CoreWeave’s Expanding Meta Contracts
Hyperscalers aren’t the only ones absorbing AI infrastructure risk. Specialized “neocloud” providers like CoreWeave are doing it too, just through a different financing structure: instead of leasing raw shell space and power from landlords, CoreWeave leases GPU capacity to hyperscalers and AI labs directly. In April 2026, CoreWeave announced an expanded agreement with Meta worth approximately $21 billion, running through December 2032 and building on a prior $14.2 billion deal from September 2025. The combined relationship now totals roughly $35 billion, and the new capacity includes some of the first commercial deployments of Nvidia’s Vera Rubin platform.
CoreWeave shares rose as much as 7.8% in premarket trading the day the deal was announced, while Meta shares moved more modestly, according to coverage from CNBC and Yahoo Finance. But CoreWeave’s growth still leans on customer concentration and debt: Microsoft alone accounted for roughly 67% of CoreWeave’s revenue last year, and around the time of the Meta expansion, CoreWeave was lining up as much as $4.25 billion in new debt, plus convertible notes and junk bonds, to help fund up to $35 billion in 2026 capex. It’s the same trade every player in this market is making in one form or another: spend now, on borrowed money or long leases, to not miss the AI capacity window.
Community Pushback: $130 Billion in Projects Blocked or Delayed
The lease boom is running into resistance on the ground. A Gallup survey found that 71% of Americans oppose new data centers being built near their communities, and the research firm Data Center Watch estimates that roughly $130 billion worth of U.S. data center projects were blocked or delayed in the first quarter of 2026 alone, per figures cited by Benzinga and tracked by Data Center Watch. Local objections typically center on electricity prices, water usage for cooling, noise and strain on regional power grids.
That pushback is a real constraint on how fast the $850 billion in committed leases can actually turn into working capacity. Some of that friction is already showing up in public commentary from tech investors, who have started weighing in on how far AI companies should go in overriding local objections versus responding to documented harm once facilities are running. Expect zoning fights, utility rate cases and state-level legislation to become as important to this story over the next year as the financing side has been so far.
What the Lease Boom Means for AWS, Azure and Google Cloud Customers
For businesses buying compute, the immediate effect of $850 billion in new lease commitments is more capacity coming online over the next few years, concentrated in AI-optimized regions with GPU and power access. The market context helps explain the urgency: global cloud infrastructure spending hit $129 billion in the first quarter of 2026 alone, up 35% year over year, with AWS holding 28% share, Azure 21% and Google Cloud 14%, a combined 63% of the market, according to industry tracking data cited by CloudZero. AI infrastructure now makes up 19% of total cloud spending, up from just 8% in 2023.
Growth rates tell a different story than market share alone. Azure grew 40% year over year and Google Cloud grew 63%, both outpacing AWS’s 19% growth, largely because AI workloads are flowing disproportionately to Azure through its OpenAI relationship and to Google Cloud through Gemini and TPU infrastructure. For customers, that means GPU-backed instances should get somewhat easier to reserve as new leased capacity comes online, but it also means the fixed costs baked into these 20-year lease commitments will eventually work their way into on-demand and reserved-instance pricing. Our recent coverage of cloud waste hitting 29% of spend is a preview of the FinOps discipline this capacity wave is going to demand from enterprise buyers.
Historical Context: From $700 Billion to $850 Billion in One Quarter
The speed of this run-up stands out even against a sector that’s been breaking spending records since the generative AI boom started in 2023. Bloomberg’s March 2026 reporting put aggregate lease commitments above $700 billion for the first time, driven largely by Meta and Microsoft each adding close to $50 billion in a single quarter. Three months later, the same tracking methodology showed the total at $850 billion, a $150 billion jump in one quarter alone. Amazon, Alphabet and CoreWeave have all ranked among the largest data center lessors going back to 2023, but the size of individual quarterly additions has grown sharply since then.
Zoom out further and the trajectory matches broader market forecasts. IDC has projected that global public cloud services spending will reach $1.6 trillion by 2028, roughly double the 2024 total. Flexera’s 2025 cloud research separately found that organizations are increasingly standing up dedicated FinOps teams specifically to manage the cost side of this buildout. The lease numbers released this quarter aren’t an isolated spike. They’re the infrastructure funding mechanism behind a spending curve that’s been forecast for years and is now arriving faster than most of those forecasts assumed.
Competitive Comparison: Hyperscalers, Neoclouds and Challengers
Three distinct strategies are visible in this data. Oracle and Meta are leasing aggressively and fast, prioritizing speed to capacity over balance-sheet caution, a strategy that’s delivered real revenue growth for Oracle but has hammered both stocks this year. Amazon and, to a lesser extent, Alphabet are pacing themselves, favoring owned infrastructure and longer planning cycles, a strategy the market is currently pricing as the safer bet. And neoclouds like CoreWeave are running a hybrid model, taking on debt and customer-concentration risk to build GPU capacity that hyperscalers would rather rent than build themselves.
Our own AWS vs. Azure vs. Google Cloud comparison breaks down how that market-share gap is showing up in product decisions, and CoreWeave’s $30 billion capex gamble digs deeper into how far a neocloud can stretch its own balance sheet before the debt load becomes the story. None of the three strategies has clearly “won” yet. What’s clear is that the market is no longer rewarding scale for its own sake. It’s rewarding scale that comes with a credible plan for paying for itself.
What Industry Data and Reporting Say
“Microsoft Corp. and Meta Platforms Inc. each committed nearly $50 billion in additional data center leases in their most recent quarters, underscoring an escalating bet the tech industry is making on artificial intelligence.”
Bloomberg, March 2026 filings analysis, as reported by IT Pro
That March figure was the first sign of how fast this cycle was moving. By June, Bloomberg’s follow-up reporting confirmed the trend had not cooled off.
“Meta Platforms Inc. and Microsoft Corp. have both pledged billions of dollars in new data center leases during their latest financial quarters, contributing to the substantial investments being made in artificial intelligence within the industry.”
Bloomberg, June 24, 2026, as reported by Benzinga
Market commentary account marketsday, relaying figures originally compiled by The Kobeissi Letter, summarized the scale of the shift bluntly: “U.S. technology companies have committed a record $850 billion in data center leases over the next several years, highlighting the accelerating investment cycle driven by artificial intelligence.” The same post put Oracle at the top of the list, citing “approximately $250 billion in total data center lease commitments,” a figure that matches the company breakdowns reported separately by Benzinga and BiGGo Finance.
5 Predictions for the Rest of 2026
- The $850 billion figure keeps climbing. If the current pace of roughly $150-200 billion in quarterly growth holds, aggregate lease commitments could approach or pass $1 trillion by the end of 2026.
- The stock-market split widens before it narrows. Expect continued pressure on Oracle, Microsoft and Meta shares until each company can show AI revenue growing as fast as its lease and capex commitments, while Amazon and Alphabet keep getting rewarded for capital discipline.
- More owned-infrastructure pivots follow AWS’s lead. Other hyperscalers are likely to quietly shift more new capacity toward owned or long-term-controlled sites rather than short-cycle colocation leases, echoing the pullback AWS has already made.
- Neocloud expansion deals keep coming. Expect more CoreWeave-style agreements, where AI labs and hyperscalers offload GPU capacity risk to specialized providers willing to carry the debt, especially as Vera Rubin-class hardware rolls out.
- Community and regulatory friction becomes a bigger line item. With $130 billion in projects already blocked or delayed in a single quarter, expect zoning battles and state utility fights to start showing up explicitly in company risk disclosures, not just local news coverage.
Related Coverage
- Microsoft’s $150 Billion AI Capex Gamble: Inside the Azure Surge and Copilot’s Adoption Crisis
- CoreWeave’s $30 Billion Capex Gamble: Inside the AI Cloud Company’s Debt-Fueled Rise
- Alphabet Q1 2026 Earnings: $109.9B Revenue, 63% Cloud Surge [2026]
- Google Cloud Hits 82% Growth Ahead of AWS Earnings [2026]
- AWS vs Azure vs Google Cloud 2026: Market Share Breakdown
- Cloud Waste Hits 29% as AI Spend Breaks Budgets [2026]
- CoreWeave’s Anthropic Deal: 12% Surge, $6.8B Backlog [2026]
For more ongoing coverage of hyperscaler spending and cloud infrastructure trends, see our cloud computing section.
Frequently Asked Questions About the Data Center Lease Boom
What are data center leases, and why are they suddenly an $850 billion story?
Data center leases are long-term rent agreements hyperscalers sign for building shells, power capacity and colocation space, often before construction is finished. Bloomberg’s June 24, 2026 analysis found aggregate future lease commitments across major cloud and social media companies had reached $850 billion, up 204% from a year earlier.
How much did Meta, Microsoft and Oracle each commit to new data center leases?
Meta added $79 billion in new commitments in the first quarter of 2026, a 76% quarterly jump, for a total of about $182.9 billion. Microsoft added $41 billion, up 26%, for a total near $196.6 billion. Oracle did not break out a quarterly figure but holds the largest total at roughly $250 billion.
Is data center leasing the same thing as capital expenditure?
No. Lease commitments are future rent obligations owed over the life of a contract, often 15 to 20 years. Capital expenditure, or capex, is cash spent in a given year on servers, chips and facilities. Amazon, Microsoft, Alphabet and Meta together are guiding to $725 billion in 2026 capex, a separate figure from the $850 billion lease total.
Why are Oracle, Microsoft and Meta’s stocks underperforming despite record AI spending?
All three are down year-to-date in 2026 (Oracle -28.33%, Microsoft -17.43%, Meta -10.38%), while Alphabet and Amazon are up. Investors appear to be rewarding companies that pace their AI infrastructure spending against visible cloud revenue growth, and penalizing heavier, faster lease and debt commitments where the payoff is less certain.
What does the lease boom mean for AWS, Azure and Google Cloud pricing?
In the near term, more leased capacity should ease GPU and compute availability. Over time, the fixed costs of these 20-year lease commitments are likely to filter into on-demand and reserved-instance pricing, making cloud cost governance and FinOps practices more important for enterprise buyers.
Why is Amazon adding fewer new data center leases than Meta or Microsoft?
AWS owns roughly as much data center space as it leases and has reportedly paused some colocation commitments, per Wells Fargo analysts cited by IT Pro. Amazon has said it believes owning facilities is more efficient over a 20-year operating life, even as it still plans roughly $200 billion in 2026 capex.
What is CoreWeave’s role in the data center leasing story?
CoreWeave is a neocloud that leases GPU capacity to hyperscalers and AI labs rather than leasing shell space itself. Its expanded April 2026 agreement with Meta, worth about $21 billion through 2032, shows how hyperscalers are offloading some AI infrastructure risk to specialized providers willing to take on debt to build it.
Could this level of data center spending be an AI bubble?
That debate is unresolved. Data Center Watch estimates $130 billion in U.S. projects were blocked or delayed in a single quarter, and a Gallup survey found 71% of Americans oppose data centers near their communities. Combined with sliding stock prices for the heaviest spenders, those are real signs of friction, though none of it has slowed the underlying commitment totals yet.
