Which antitrust theories threaten Meta's $14B data center deal?

The antitrust fact in Meta’s El Paso data center venture is not simply that the project is large. It is that Meta is the sole initial tenant in a BlackRock-majority special purpose vehicle, under a four-year lease with renewal options, while Meta also manages construction of the campus. Reuters reported the venture as a $14 billion project, with BlackRock holding 80% and Meta holding 20%; Forbes likewise described BlackRock’s 80% ownership and Meta’s minority stake, and Meta’s own announcement identifies the structure as a strategic venture to develop a data center in El Paso.[1][2][3]

That combination gives agencies a way into the file even if it does not give them an easy way out of court. A sole-tenant lease is legible. A repeat 80/20 SPV pattern is legible. Construction control by the downstream AI platform is legible. The harder question is what those facts prove: ordinary infrastructure finance, a financing workaround for enormous AI-capacity needs, or a practical method of reserving scarce inputs before rivals can get to them.

Vinson & Elkins’ July 27, 2026 AI infrastructure framework is useful because it separates four theories that too often get mashed into one regulatory mood: supply banking, killer acquisitions, exclusive dealing, and vertical integration.[4] For this deal, the first and third theories do most of the work. Supply banking and exclusive dealing are the most plausible investigation hooks. Vertical integration is worth a narrower inquiry. Killer acquisition should not carry much weight on the disclosed facts.

Supply Banking Is the Theory That Needs the Pattern

A one-off anchor-tenant data center lease is usually a financing story before it is an antitrust story. The tenant wants committed capacity; the capital partner wants predictable cash flow; the project needs creditworthy demand before construction risk becomes tolerable. The El Paso structure starts to look more interesting only when it is placed next to Meta’s earlier 80/20 SPV pattern.

Data Centre Magazine described the BlackRock deal against the backdrop of Meta’s earlier Blue Owl/Hyperion structure, reported as an October 2025, $30 billion, 5 GW Louisiana data center arrangement using the same 80/20 ownership split.[5] That does not prove foreclosure. It does, however, create the kind of cumulative record an agency lawyer would want to understand before accepting the explanation that each SPV is just project finance in isolation.

The legal difficulty is real. V&E’s framework treats supply banking as a live AI-infrastructure theory but also notes that applying it to input purchasing would be a significant extension of existing law.[4] That matters. Existing doctrine has been more demanding where the alleged exclusion comes from buying or reserving inputs rather than selling under exclusionary terms. A plaintiff would need more than a large buyer, a large project, and anxious competitors.

Investigation risk is a different animal. Agencies do not need to have a trial-ready theory before asking why a hyperscaler repeatedly uses minority-owned SPVs to secure multi-gigawatt AI infrastructure. They would ask whether capacity is scarce in the relevant geography or power market, whether competing AI developers had realistic access to comparable sites, and whether Meta evaluated the deals as a way to deny rivals timely access to capacity. They would also ask whether the 80/20 structure was selected for accounting, financing, control, HSR, tax, construction, or competitive reasons.

The documents that hurt in a supply-banking review are rarely the polished board sentences saying the project supports AI growth. The problem documents are the ones that translate capacity into rival delay: decks describing how much third-party AI demand the campus would otherwise serve, memos treating the SPV as a way to keep scarce power interconnection or data center capacity away from others, or emails celebrating that a financing partner lets Meta reserve more capacity than it could hold directly. If those documents do not exist, counsel still needs contemporaneous materials explaining why the structure was chosen before anyone started drafting for antitrust optics.

Exclusive Dealing Is the Cleanest Fact, Not Necessarily the Cleanest Case

The sole-tenant lease is the fact an investigator can explain in one sentence. Meta is the initial sole occupant for four years, with four renewal options, in a venture where BlackRock owns the majority and Meta manages construction.[1][2] Whether the contract uses the phrase “exclusive dealing” is beside the point. Functionally, the question is whether the arrangement reserves a meaningful share of a scarce input for Meta and away from rivals.

That does not make a Sherman Act case easy. Traditional exclusive-dealing analysis would look for substantial foreclosure in a properly defined market, not merely a contract that prevents one asset from serving others during an initial term. The relevant market could be narrow or broad: AI-ready data center capacity in a particular power-constrained region, large-scale leased AI compute infrastructure, or something wider that includes self-built campuses and alternative locations. Each framing changes the foreclosure story.

The FTC Act route is more flexible. If the agency sees the lease as an incipient method of locking up AI infrastructure before rivals can obtain it, Section 5 pressure could arrive before anyone has a litigable exclusive-dealing case. That is why labeling the risk “low” misses the operational point. The courtroom endpoint may be weak under current law; the subpoena endpoint is not.

The useful questions are concrete. Did BlackRock require a sole initial tenant to finance the campus on acceptable terms? Did Meta need construction management because of specialized AI data center specifications, supply-chain control, or delivery timing? Were shorter commitments or partial third-party leasing considered and rejected? Did the renewal options track investment recovery, or did they simply extend Meta’s control over scarce capacity? Those questions are not cosmetic. They decide whether the same lease reads like financeable infrastructure or foreclosure by contract.

The answer may be perfectly defensible. A 1 GW, 1,000-acre AI data center campus is not a generic office lease; Reuters reported both the scale and the sole initial tenant structure.[1] A capital partner may reasonably want committed demand, and a hyperscaler may reasonably want build-speed certainty. The problem comes when those rationales are reconstructed only after the antitrust theory has been named. Agencies read chronology as closely as they read wording.

Vertical Integration Is Narrower, but Construction Control Keeps It Alive

The vertical theory is not the strongest fit, mainly because Meta is not disclosed as controlling the SPV. The public reports describe BlackRock as the 80% owner and Meta as the 20% owner.[1][2] On those facts, a classic vertical foreclosure case has to work harder. A minority stake, standing alone, does not show the ability to deny rivals access to the asset, especially if BlackRock or the SPV retains leasing authority after the initial Meta commitment.

But the construction role matters. Meta is not merely signing a power-hungry lease after someone else builds a facility. It is managing construction of the campus.[1][3] That can be innocuous and even necessary: hyperscale AI workloads may require specifications that a passive real estate investor cannot design alone. It also means an agency will want the governance documents, not just the ownership percentages. Construction control can affect timing, technical compatibility, expansion sequencing, and practical access even where formal equity control is absent.

Recent enforcement climate makes that inquiry more plausible, though not necessarily more successful. Goodwin’s first-half 2026 antitrust update reported that large AI and technology transactions continued to clear without challenge, while also noting that the FTC had flagged AI partnerships for increased scrutiny.[6] That is the right level of signal: heightened interest, not a presumption that every infrastructure JV is unlawful.

The vertical documents should therefore answer ability and incentive without overclaiming. Who decides whether third parties can lease unused or future capacity? Can Meta influence expansion phases? Are technical specifications tailored only to Meta’s stack, or are they commercially justified for AI workloads more generally? Does Meta’s role as a downstream AI platform give it a reason to disadvantage other compute customers? The disclosed 20% stake cuts against a simple foreclosure theory; the operational role prevents the issue from disappearing.

Killer Acquisition Does Not Fit the Public Facts

The killer-acquisition theory should be eliminated quickly. The El Paso SPV is not disclosed as Meta buying a nascent AI rival, absorbing an independent compute platform, or acquiring a data center operator that otherwise would have competed against it. It is a financing and development vehicle for a specific infrastructure asset.

That discipline matters. AI antitrust analysis gets weaker when every large transaction is forced into every fashionable theory. If later facts showed that the venture displaced an independent capacity provider or prevented a would-be competitor from entering the market, the analysis could change. Nothing in the public deal description requires that move now.

Process Risk Is Not the Same as Illegality

HSR status is an uncertainty, not a hidden answer. Quinn Emanuel has warned that AI data center SPVs can be structured in ways that fall below Hart-Scott-Rodino notification thresholds, depending on the rights and assets involved.[7] Goodwin also reported that expanded HSR rules were vacated in February 2026, creating a lighter filing environment than the one deal teams had been preparing for.[6]

None of that establishes that the El Paso venture avoided a filing, should have filed, or was structured for that purpose. The public information is not enough to determine reportability. Asset valuation, control rights, governance rights, formation mechanics, and exemptions would matter. The safer point is narrower: if a transaction of this size and strategic significance does not generate a pre-closing filing, agencies may still learn about it from public reporting, market complaints, competitor submissions, state attorneys general, or later conduct inquiries.

The FTC and DOJ also opened a joint inquiry into competitor-collaboration guidance in February 2026, which adds regulatory uncertainty around joint ventures and collaborations.[6] That inquiry should not be inflated into a deal-specific threat. It does, however, make counsel’s record more important because agencies may be recalibrating how they describe collaborations in markets where capacity, data, chips, power, and compute access are strategically bundled.

The Record Counsel Needs Before the Questions Arrive

The practical risk map is uneven. Litigation risk remains low under current law because supply banking would require an extension of existing input-purchasing doctrine, exclusive dealing would need substantial foreclosure proof, vertical integration is softened by Meta’s minority stake, and killer acquisition does not fit the disclosed transaction. Investigation risk is real because supply banking and exclusive dealing give agencies plausible entry points into documents, governance rights, market conditions, and the cumulative SPV pattern.

The file should explain the sole-tenant structure in ordinary business language, not in language reverse-engineered from an antitrust theory. If the lease was necessary to finance a 1 GW campus, say who required that commitment and why.[1] If Meta managed construction because AI workloads required specific design, supply-chain security, or accelerated delivery, the record should show that those considerations existed when the decision was made.[3] If the 80/20 structure repeated the Blue Owl/Hyperion pattern because it solved capital allocation or risk-sharing problems, the record should explain why the pattern made business sense rather than leaving agencies to infer that repetition means capacity banking.[5]

Good records will not prevent scrutiny if agencies decide AI infrastructure deserves a closer look. They can, however, determine whether scrutiny finds contemporaneous business necessity or post hoc antitrust hygiene. On the public facts, the stronger threat is not a near-term courtroom loss, but a document-driven inquiry into why Meta needed sole initial occupancy, construction control, renewal options, and another 80/20 SPV.

References

1. Meta, BlackRock partner on $14 billion El Paso data center, Reuters, July 28, 2026
2. BlackRock Will Own 80% Of Meta’s Massive New AI Data Center In El Paso, Forbes, July 28, 2026
3. Meta Announces New Strategic Venture with BlackRock to Develop Data Center in El Paso, Meta Investor Relations, 2026
4. The Next Antitrust Frontier: AI Infrastructure, Vinson & Elkins, July 27, 2026
5. Why Meta Chose BlackRock for its US$14bn AI Data Centre Deal, Data Centre Magazine
6. Antitrust & Competition Technology 1H 2026, Goodwin, July 2026
7. Client Alert: Emerging Litigation Risks in Financing AI Data Centers Boom, Quinn Emanuel