NextFin News - Mark Zuckerberg's $1.7 trillion empire is betting its future on artificial intelligence, and the bill is too large to pay alone. At a board retreat last December at his Kauai compound, the Meta Platforms chief executive officer pulled aside Dina Powell McCormick, a director who had joined the board only eight months earlier, and laid out the problem: to win the AI race, Meta would need to raise a tremendous amount of outside capital — and it would need an adviser with the Wall Street connections and political capital to make that possible. Eight months later, that conversation has reshaped Meta's leadership and exposed a deeper truth about the AI buildout: the winner may not be the company with the best model, but the one that can finance its infrastructure at the lowest cost.
The $145 Billion Question
Meta is no longer funding its AI ambitions out of its own cash flow. The company now expects to spend between $130 billion and $145 billion on capital expenditures in 2026, narrowing a forecast it raised to $125 billion to $145 billion after increasing it from $115 billion to $135 billion at the start of the year. That is roughly double the $72 billion it invested in 2025, which itself was nearly double the $39.23 billion spent in 2024. In the second quarter alone, capital spending surged 83 percent to $31.08 billion, nearly matching the $31.9 billion in operating cash flow the business generated. The result: free cash flow collapsed 91 percent to $784 million, the weakest reading since late 2022, when investors last revolted against Zuckerberg's spending — then on the metaverse, now on AI.
The market's verdict was swift. After Meta reported second-quarter earnings per share of $6.18, short of the $7.22 analysts expected, shares fell about 10 percent in extended trading. Revenue grew 28 percent to $60.8 billion, the fastest pace since the fourth quarter of 2021, and daily active users across its apps rose 3 percent to 3.6 billion. None of that mattered as much as the cash flow line.
Meta's AI spend was easier to celebrate when margins were expanding. It's harder to celebrate now that the costs are showing up in the numbers.
Mike Proulx of research firm Forrester captured the shift in investor psychology. The numbers explain why: second-quarter operating income fell 8 percent year over year, and the company booked $2.4 billion in charges related to legal proceedings plus $1.18 billion in severance expenses from its May layoffs. Chief Financial Officer Susan Li said operating income would have risen 9 percent excluding those items — a distinction that reassured some analysts but did little to ease the broader concern that Meta's cost base is being structurally reset higher.
The spending is physical, not abstract. Meta plans to double its computing capacity to 7 gigawatts this year and double it again to 14 gigawatts in 2027, operating or building 32 data centers worldwide. Each gigawatt of AI data center capacity carries an estimated price tag in the low single-digit billions of dollars once land, construction, servers, and grid connections are counted. At that math, a 14-gigawatt fleet represents a multi-hundred-billion-dollar asset base that must be financed, maintained, powered, and refreshed on a three-to-five-year hardware cycle.
Why the Capital Markets Became the Battlefield
Here is the mechanism investors often miss: AI infrastructure spending does not just hit the income statement — it transforms what kind of company Meta is. For most of its history, Meta was a self-funding machine. Its advertising business generated enough cash to build data centers, retire shares, and still return money to shareholders. The company has spent approximately $174 billion to buy back nearly 12.7 percent of its outstanding shares, one of the most aggressive return-of-capital programs in corporate history. That model is breaking. When quarterly capital spending alone exceeds $30 billion and free cash flow falls below $1 billion, internal cash flow can no longer carry the buildout. Meta has not repurchased shares since the third quarter of 2025, and reports have suggested it is weighing equity issuance to help fund the AI infrastructure push. The same buyback drought is spreading across the sector: of the four biggest AI spenders — Alphabet, Microsoft, Meta, and Amazon — only Microsoft bought back shares in the first quarter, and its $3.4 billion of repurchases was the lowest total among the group in nearly a decade.
The financing is already underway, and it is taking forms Meta has rarely used. In April, the company sold $25 billion of investment-grade bonds across six tranches to fund AI infrastructure, following a record $30 billion debt offering in 2025 and a comparatively modest $10 billion issuance in 2022. The trajectory of the debt program — $10 billion, then $30 billion, then $25 billion in little more than a year — maps the speed at which Meta's financial calculus has changed. Peer issuances tell the same story: Amazon raised about $54 billion across U.S. and European markets, Alphabet issued roughly $32 billion in dollar and euro notes, Oracle completed a $25 billion sale, and Nvidia tapped the market for $25 billion in its first corporate debt offering since 2021. Big Tech is no longer competing only for chips and engineers. It is competing for the same pool of fixed-income capital, and that competition is pushing up the price.
In one of the most striking examples, Meta is financing its roughly $14 billion, 1-gigawatt data center campus in El Paso, Texas, through a partnership with BlackRock. BlackRock-managed funds hold an 80 percent stake while Meta retains 20 percent. Meta contributed land and construction-in-progress worth about $2.3 billion, BlackRock injected about $4.9 billion in cash, and the remaining $12.5 billion is covered by debt from partners including Global Infrastructure Partners and HPS Investment Partners. Meta received a $1 billion distribution to align ownership and will lease the capacity back for up to 20 years. The structure lets Meta book the compute capacity it needs while pushing the immediate financing burden onto an asset manager — but a $13 billion residual value guarantee concentrates long-term asset risk back on Meta's balance sheet. It is a financing model invented for hyperscalers that can no longer self-fund at the pace the AI race demands.
This is where Powell McCormick enters as more than a ceremonial appointment. Announced on January 12, 2026, her role as president and vice chairman makes her a member of Meta's management team, tasked to "build new strategic capital partnerships and find innovative ways to expand our long-term investment capacity," and to partner with "governments and sovereigns to build, deploy, invest in, and finance Meta's infrastructure." Her résumé is the point: 16 years at Goldman Sachs, where she led the firm's Global Sovereign Investment Banking business and sat on its Management Committee, plus service as deputy national security adviser to President Donald Trump and as an assistant secretary of state under President George W. Bush. Most recently she was vice chair, president, and head of global client services at BDT & MSD Partners.
Dina's experience at the highest levels of global finance, combined with her deep relationships around the world, makes her uniquely suited to help Meta manage this next phase of growth.
Zuckerberg said in a statement. In the AI infrastructure race, relationships with sovereign wealth funds, pension funds, and governments are not soft assets — they are the channel through which $100 billion-plus capital programs get funded at viable rates. A sovereign anchor can provide patient, long-duration capital that public bond markets price at a premium for technology risk. It can also smooth the regulatory and political friction that comes with building gigawatt-scale facilities, securing power purchase agreements, and navigating cross-border data rules. Powell McCormick's dual fluency — in the language of Wall Street syndicates and in the corridors of national security — is precisely the profile a company needs when its capital plan outgrows its domestic debt market.
The Cost of Capital Is the New Moat
The second-order consequence of this shift is that artificial intelligence is turning infrastructure into a financing business. When every hyperscaler needs hundreds of billions of dollars, the scarce resource is not just chips or electricity — it is cheap, patient capital. And the price of that capital is rising. Lenders have begun demanding higher yields on bonds tied to Meta's data-center projects compared with similar deals a year earlier, a sign of debt-investor fatigue as big technology companies flood the market with issuance. S&P Global has rated Meta's planned debt investment-grade with a stable outlook but noted that the scale of AI spending is beginning to affect the company's credit profile. Meta ended the second quarter with $90.3 billion in cash and marketable securities against $83.7 billion in debt, after roughly $24.9 billion of net long-term debt issuance in the first half of 2026. Quarterly dividends of $1.353 billion exceeded the quarter's entire $784 million of free cash flow — a small but symbolic inversion of the old order, in which the operating business comfortably covered every claim on its cash.
Across the sector, Big Tech hyperscalers are expected to spend between $650 billion and $725 billion on AI infrastructure in 2026, with AI-related debt issuance projected to reach $570 billion. At that scale, the technology industry is becoming one of the largest borrowers in global credit markets, and the marginal cost of the next dollar of infrastructure funding is set by the market's tolerance for AI risk — not by any single company's balance sheet strength.
The divergence among peers is instructive. Microsoft's free cash flow fell 23 percent in the June quarter, yet its shares rose because Azure revenue grew 43 percent in its fiscal fourth quarter and operating margins held near 45 percent — the market believes Microsoft's cloud can self-fund its roughly $175 billion capital plan. Azure's commercial remaining performance obligations reached $627 billion, giving Microsoft far more revenue visibility than a company building capacity without committed customers. Meta, by contrast, remains an almost entirely advertising-driven business attempting to diversify, and its monetization of AI is still prospective. Zuckerberg told analysts on the earnings call:
We expect that a significant portion of our compute is going to go towards training our models, growing our core business and delivering personal agents and new products, but we also expect to grow a large business serving large customers as well.
The phrase "we also expect to grow" is doing a lot of work: external compute revenue is a promise, not yet a disclosed line item. Until Meta can point to a revenue stream from serving large customers that scales with its capacity additions, every new gigawatt is a bet that advertising efficiency and future agent businesses will absorb it. That is a different risk profile from Microsoft's contracted cloud backlog, and the debt market is beginning to price the difference.
That gap between cash generation and cash need is the structural heart of the story. It is not a cyclical dip in free cash flow that will normalize on its own. The capital intensity of frontier AI is a regime change in the economics of a social media company, and Powell McCormick's role exists because Meta has recognized, internally, that its future capital structure will look less like a technology company's and more like an infrastructure platform's — part balance sheet, part syndicate, part sovereign partnership.
The Counter-Thesis: The Ad Machine Still Prints Cash
The strongest argument against this reading is the simplest: Meta's core business remains powerful enough to fund the AI push without fundamentally reshaping the company. Revenue grew 28 percent in the second quarter. The company has 3.6 billion daily active users. Advertising pricing keeps rising as AI improves targeting and recommendation systems, and the underlying ad engine is what pays for everything else. Luke Stillman of research firm Madison and Wall put it plainly:
Meta's underlying ad business that's financing everything though is still performing well and is our main focus.
There is evidence on the bull side. Wall Street analysts at Goldman Sachs project Meta's AI-related spending will reach $534 billion from 2026 to 2028 as the company continues to monetize its compute buildout, and the bank maintains a Buy rating. Bernstein has reiterated an Outperform rating, arguing that Meta's advertising revenue trajectory is converging toward Google Search in its ability to capture high-intent demand. If AI-driven ad efficiency continues to compound revenue off a $60 billion quarterly base, today's cash crunch could prove to be a timing mismatch rather than a structural deficit — and Powell McCormick's capital-raising mandate would be a hedge, not a necessity.
That view deserves weight, but it rests on a specific sequence: monetization must arrive before the financing wall does. The falsifying signal is concrete. If Meta's 2027 capital expenditure guidance remains above $120 billion while quarterly free cash flow stays below $10 billion, and the company issues more than $50 billion in incremental debt or equity over the following 12 months, then the ad business alone cannot carry the buildout — and external capital becomes structurally required, not optional. A secondary signal would be a credit downgrade or a meaningful widening of Meta's credit spreads relative to peers with contracted cloud revenue, which would confirm that lenders view the company's financing need as durable rather than transitional. Until then, both readings remain live.
What Comes Next
In the short term, Meta's stock will keep reacting to each capital expenditure print and free cash flow number — the same sentiment cycle that produced the 10 percent post-earnings drop. Shares trade near $667, with a market value of about $1.7 trillion, still roughly 16 percent below the all-time high close reached in August 2025 and inside a 52-week range of $520 to $791. In the medium term, the key data points are the pace of external compute revenue, the cost of debt on new data-center financings, and whether free cash flow recovers toward $5 billion a quarter. The company expects third-quarter revenue in the range of $61 billion to $64 billion and has guided full-year 2026 total expenses to $165 billion to $169 billion, up from $117.69 billion a year earlier.
In the long term, the question is whether Meta can operate a hybrid capital model — advertising profits plus sovereign and institutional co-investment — without ceding control or accepting dilutive terms. Selling majority stakes in data centers to asset managers preserves liquidity today but creates a growing web of lease obligations and residual value guarantees that claim cash flow for two decades. If those obligations multiply as the 14-gigawatt target comes into view, Meta's equity holders will be funding the AI buildout indirectly, through contracted payments rather than direct capital expenditure — with less flexibility to slow down if the payoff disappoints.
The beneficiaries and the exposed are easy to name. Asset managers and sovereign funds with deployable capital gain pricing power and deal flow; the El Paso structure is likely to be replicated across the industry. Bond investors gain a new, high-volume issuance market — but only if yields compensate them for technology risk they have never underwritten. Meta's advertising customers benefit from better targeting regardless of who owns the data centers. The exposed are Meta's equity holders, who now bear both execution risk on AI and financing risk on the capital structure, and any hyperscaler that must raise money in a market where lenders are starting to ask harder questions.
Zuckerberg's walk with Powell McCormick in Kauai was, in one sense, a personnel decision. In another, it was an admission: the AI empire he is building cannot be self-funded, and the path to 14 gigawatts runs through Wall Street, Washington, and the sovereign wealth funds in between. The company that masters the cost of capital may outlast the company that masters the model.
Explore more exclusive insights at nextfin.ai.
