Firmus's $5bn Listing Prices the Contract

Firmus's Australian listing, tracked in PWD's deal log this week, asks whether the market will pay $5bn for megawatts it has already sold, putting a public number on a contracted AI capacity backlog that other neoclouds keep off the page. Most of the quarter's energy deals arrived with names, technologies, and sometimes a headline euro figure, but almost never with the price that settles the trade.

Where a number did appear, the disclosure only sharpened the problem: Naturgy is paying €330m to repower Spanish grid it already holds, pricing the connection rights that permit additional megawatts, with the power itself left out of the number, and while the program adds capacity at connections already in hand, the coverage stops short of naming the offtake that will absorb the new output. Without that offtake, even a public number remains a private guess. Palisade's 1.4GW Australian pipeline purchase carries no disclosed price even though connection rights are the scarce asset in that market, and Hy24's OPAL purchase names Uniper and the pipeline but omits the figure that would tell the market what the existing midstream asset is worth.

Where a headline figure did exist, it washed out on inspection: Masdar and Luxcara announced a €5bn tie-up covering battery storage and offshore wind with no capacity, counterparty, or structure disclosed, amounting to a statement of strategic intent rather than a priced transaction. Latvenergo is pursuing 250MW of wind with no seller, price, or offtake named, and Nordex's 34-MW repowering order arrived as a capacity figure with no buyer and no site. The market is left tracking appetite rather than value.

Qualitas Energy, Ørsted, Alinta, and Sonntag all appear in the quarter's round-up, but the details carry no price, counterparty, or capacity, and the gap between announcing a deal and pricing one has become normal. Firmus's listing is different because an announced deal without a price is an expression of interest, while a listing demands a number.

Against that unpriced pipeline, Firmus's $5bn ask is the quarter's clearest attempt to put a public price on electricity that has already been sold, and the distinction matters because infrastructure investors buy cash flows rather than capacity. A contracted megawatt has a revenue stream, a counterparty, and a term; a repowered turbine or a 250MW development target has a future market and a permitting risk. Because the neocloud capital stack has been built on contracted revenue, the backlog is now the balance sheet, and taking it public tests whether public investors will underwrite AI capacity as infrastructure at all.

What the market will be asked to underwrite is a contract, not a data center: $5bn implies the megawatts already have buyers, with the value sitting in offtake agreements instead of in steel and transformers. The listing documents will eventually have to show who signed and for how long, because a contracted megawatt is only as good as the counterparty behind it, and without counterparties and tenors the market cannot tell whether it is buying a long-term hyperscaler lease or a shorter placeholder. The multiple gap between those outcomes is the gap between a utility and a development company. Firmus has chosen to put the number before the names, and the next round of disclosure will show whether that order was confidence or necessity.

The unpriced alternative is clearest in Egypt, where Heca Data is studying power with the state transmission company; the step that will decide whether the integrated hyperscale zone becomes a project is the power study rather than the September 3 meeting. The project cannot name a tenant before it names a watt, and the watt depends on the grid rather than the data hall. That is the inverse of Firmus's listing: Egypt has no contracted revenue to show, so there is no number to attach.

The same bottleneck shows up in the retrofit fast path for AI capacity: the fastest new capacity sits inside buildings already standing, where the electrical ceiling is lowest, and a retrofit project can move quickly until it reaches the substation and joins the same queue as every greenfield development. Contracted megawatts are worth more per unit than any unpriced project in the quarter because they are selling through that bottleneck rather than toward it, and a developer who controls a connection can sell contracted megawatts while one who does not is selling a site and a plan.

If the Australian listing clears at $5bn, contracted AI capacity leaves the private capital stack and becomes a public infrastructure class, with consequences for every neocloud holding a backlog and every energy developer still unable to name a buyer. The first disclosure to watch is who signed those contracts and for how long, because the number means little until the market knows whom it is underwriting.