Meta’s $50B Hyperion Expansion: The CRE Read on a 5 GW Campus

Site selection, land pricing, incentive negotiation, and power procurement are accountable human decisions. So is the choice not to chase: a market where the headline number is large but the interconnection queue is unsigned deserves patience, not a land position taken on momentum.

The honest caution runs in both directions. Do not underwrite data center demand into rents without a signed load or an executed lease, and do not assume announced totals are contractual. Cost allocation for new generation is contested in public dockets, and those proceedings, not the press cycle, will set the durable facts around Hyperion.

At Hyperion scale, even temporary construction demand needs to be separated from the operating market it may leave behind. A local exposure file should distinguish workforce lodging, laydown and storage demand, supplier space, permanent operations headcount, and the utilities that enable each phase. Those are different products, with different lease durations and different downside cases. A hotel full of construction crews can be a real signal for near-term cash flow without proving a permanent rate reset; a land parcel near a proposed transmission line can gain optionality without being entitled for a data center. The file should make those distinctions explicit before a rent-growth or residual assumption enters a model.

The committee also needs a sequencing test. What must happen first for the thesis to hold: a utility approval, new generation, a water permit, an executed customer commitment, or a construction notice to proceed? Put those conditions in order and identify the party that controls each one. Then model a delay at the slowest gate, not just the base-case delivery date. This turns a giant investment announcement into an inspectable chain of dependencies. It also makes the right diligence calls obvious: utility and commission contacts for load and generation, local officials for permitting, and brokers for current, not projected, supply and labor pressure.

Finally, preserve the difference between a company’s stated economic impact and a decision-ready local market fact. The public announcement is useful because it defines what to investigate. It is not evidence that every surrounding parcel should trade as data-center land, or that every contractor-facing use has durable demand. The useful operator read is therefore a living record: announce, verify, date, and revise as filings and milestones arrive. That discipline lets an investment team benefit from the signal without turning the signal into a substitute for underwriting.

For an investment committee, the deliverable should end with an evidence register: claim, primary record, as-of date, site-level relevance, and the next verification event. It is a modest control with an important effect. It prevents a large announcement from hardening into a permanent model input after the conditions, timing, or responsible party have changed. The result is a local thesis that can become more confident as the record improves, rather than a national narrative that grows more certain simply because it has been repeated.