An IPO Offering $430 Billion of Backlog, Three Data Centers That Don't Exist Yet, and Free Rent If It's Late
SB Energy wants to go public on a backlog that mostly doesn't start for eight years, from an anchor customer with no credit rating, funded largely with debt.
By Sean Anees Saifi · Capital Wealth · Published Friday, September 11, 2026 · Source: The Wall Street Journal, September 11, 2026 edition
Key Points
The pitch rests on $430 billion of contracted future data-center revenue, mostly from OpenAI with a partial Nvidia (NVDA) backstop.
About 82% of that backlog doesn't start for at least eight years, and none of the three planned data centers is operating yet.
One project — the 10-gigawatt PORTS-Pike campus in Ohio, phasing in 2028–2032 — accounts for 90% of contracted capacity.
Planned capital spending runs to about $174 billion, funded mostly with debt, while the anchor customer carries no credit rating.
Delays carry teeth: free rent that doubles after 90 days, and a buyout option at a Texas site priced materially below lease value.
$430B
Contracted future revenue on offer
82%
Of backlog starts 8+ years from now
90%
Of capacity in one unbuilt project
$174B
Planned capex, mostly debt-funded
Development-stage power and data-center companies typically absorb years of losses before cash flows begin, unlike listed operators that went public with assets already running.
In one line: A $430 billion backlog sounds like a thesis until you read the start dates, the concentration and the funding — which is why the book avoids it.
Somewhere in the paperwork for SB Energy — the SoftBank-backed power and data-center developer planning to go public as soon as this month — sits a clause promising OpenAI free rent if the buildings run late. The rent-free days double after 90. It's an unusual way to advertise confidence in your construction schedule.
Read the backlog, then read the start dates
The headline is a $430 billion book of contracted future revenue, mostly from OpenAI, with partial backstop support from Nvidia (NVDA). That's a real number. Here are the other real numbers. Roughly 82% of that backlog doesn't begin for at least eight years. About 90% of contracted capacity sits in a single project, the 10-gigawatt PORTS-Pike campus in Ohio, phasing in between 2028 and 2032 on contracts that run into the 2050s. And none of the company's three planned data centers is operating today.
Getting there takes roughly $174 billion of capital spending, funded mostly with debt. The flagship campus depends on a natural-gas plant that an unrelated SoftBank affiliate — not SB Energy — is meant to build, with no binding agreements yet. The anchor customer carries no credit rating at all. Meanwhile 360 companies have at least one U.S. data center on the drawing board, which is the sort of figure that turns a shortage into a glut roughly four years after everyone agrees there's a shortage.
Why this is the trade the book won't make
Listed operators such as NextEra Energy (NEE), Vistra (VST), Digital Realty (DLR) and Equinix (EQIX) came to market with assets already spinning. Development-stage companies post years of losses first. As Ted Brandt, chief executive of Marathon Capital, put it, “The earnings pattern sucks for a typical developer.” That's the honest version, and it's precisely why this sits in the category the model portfolios avoid outright — the lottery ticket bought on a theme, priced on a promise, with the payoff pushed well past most people's planning horizon.
None of which means SB Energy fails. It means the reward is a decade out while the risks are concentrated, leveraged and dependent on somebody else pouring concrete on time. If the AI build-out appeals — and there are sound reasons it might — the better question is which part of it already pays its own way today. That one gets answered against a plan, not a prospectus, and it's worth raising at your next review.
What It Means For Your Portfolio
Avoid — a decade-out payoff, concentrated and debt-funded
A backlog that mostly doesn't start for eight years, sits 90% inside one unbuilt project and is funded with debt isn't an investment thesis — it's a lottery ticket with a prospectus attached.
General planning principles, not advice for anyone in particular: a large contracted backlog says nothing about when cash actually arrives, and concentration in one project or one customer is a risk no headline number offsets. Development-stage companies typically lose money for years before they stop, and an unrated counterparty is a genuine variable rather than a footnote.
In the book, this is the category the model portfolios avoid outright — pre-operating listings bought on theme enthusiasm. AI build-out exposure is held instead through operating companies with customers and cash flow today, and nothing new is being bought this week under the cautious house bias.
