The Multi-Gigawatt Megaproject

The multi-gigawatt megaproject is a new object in the infrastructure landscape, and it does not obey the physics of the data centers that came before it. Named, single-purpose, purpose-financed campuses drawing power at the scale of a large industrial complex have moved from proposal to construction. Treat the megaproject as its own category, because underwriting it with data-center intuitions produces nonsense.

Start with scale, because scale changes kind, not just degree. A conventional large campus draws power at a level a utility can serve from existing headroom. A multi-gigawatt megaproject draws more than many entire metropolitan areas, and it wants that power as firm base load, continuously, for years. At that scale you cannot simply request a connection — you must co-develop generation. The megaproject stops being a customer of the power system and becomes a co-planner of it, negotiating new plants, dedicated substations, and transmission upgrades as part of the project scope. The building and the power plant are designed together.

Next, the financing. A megaproject cannot ride on a single company's operating cash flow without distorting it, so it is carved into a purpose-built financing vehicle — a joint venture or project-finance structure that isolates the capital, brings in outside investors, and matches long-dated debt to long-dated power and offtake contracts. This is the same machinery that builds pipelines and toll roads, now pointed at compute. The structure matters because it determines who actually bears the risk: the lab may be the anchor and the reason the project exists, but the capital may sit with infrastructure investors who require a firm, contracted demand signal to fund at all. The megaproject is as much a financing artifact as an engineering one.

Then, the schedule. A megaproject's critical path is not the servers — it is the long-lead physical inputs. Turbines, high-voltage transformers, switchgear, and the interconnection study each run on multi-quarter to multi-year clocks. The project is orchestrated backward from a target energization date, with equipment ordered years early and construction phased so the first hall can train while later halls are still steel. This phasing is why megaprojects are announced as staged builds: a first tranche of power energizes, then successive tranches follow as generation and equipment arrive. The named capacity is a destination, not a day-one fact.

Now the concentration risk, which is the defining feature. A megaproject is a single enormous bet on a single demand thesis. There is no diversification inside it — one anchor, one purpose, one roadmap. If the anchor's compute need materializes as planned, the project is a triumph of foresight. If the thesis slips — a cheaper training method, a demand plateau, a model that under-delivers — the project is a stranded industrial asset with a power contract it must honor and no obvious second tenant. This binary is why megaprojects attract both spectacular capital and spectacular skepticism. They are not hedged; they are conviction expressed in concrete.

Consider the location logic, which inverts the old rulebook. Conventional campuses cluster near population, network hubs, and existing data-center gravity. Megaprojects go where the firm power is, even if that is far from every traditional hub. Available generation, permitting speed, and land at scale outrank latency to end users, because a training campus talks mostly to itself, not to consumers. This is how regions with no data-center history — places with stranded gas, surplus generation, or fast permitting — become frontier compute sites. The megaproject follows energy, and energy is not where the old map said to build.

Finally, the ripple. A single megaproject reorganizes its host region: it commits a large block of a utility's future generation, it pulls construction labor and long-lead equipment out of the regional pool, and it becomes a political and economic anchor that governments court and communities contest. The project is never just a building — it is a regional infrastructure event. For the practitioner, the megaproject is best read as a fusion of three things that used to be separate: a power plant, an industrial facility, and a compute cluster, commissioned as one object by a buyer who needs all three to arrive on the same schedule. That fusion is the signature of the principal buyer operating at full scale.