NextFin News - A power company with no operating data centers and a $3.2 billion half-year loss just asked public-market investors to fund the most expensive real-estate bet in the artificial-intelligence boom. SB Energy, the SoftBank-backed infrastructure venture, has confidentially filed for a U.S. initial public offering that bankers expect to raise between $5 billion and $7 billion — and its draft filing reveals a business whose entire future rests on one tenant, one financier, and one majority owner.
The tell is in the terms SB Energy offered to land that tenant. OpenAI was issued stock warrants worth an estimated $5.5 billion as an inducement to sign as the anchor customer at SB Energy's flagship southern Ohio development, according to draft IPO documents reviewed by the financial press. The warrants were valued at $3.6 billion when issued in January; by the end of June, that paper claim had swollen to $5.5 billion. In plain terms: SB Energy is giving away equity worth more than the entire $5 billion to $7 billion it hopes to raise from public investors, just to secure a customer that does not yet exist.
That trade captures the central tension of the AI buildout in 2026. Chips get the headlines, but electricity has become the scarce input — and SB Energy is attempting to go public as the first pure-play bet on the right to consume power at scale. The question for investors is whether the company is selling genuine scarcity value, or whether it is selling a promise backed by a tenant, a chip supplier, and a parent company whose fortunes all move together.
The Filing, the Numbers, and the Three-Legged Stool
SB Energy expects to make its IPO filing public as soon as the week of August 31, 2026, with bankers targeting a listing as soon as the following month. The company has raised $2.4 billion across four funding rounds: $800 million from Ares Infrastructure Opportunities funds in 2025, and another $1 billion in January 2026 split evenly between SoftBank and OpenAI. Nvidia has committed $3 billion through two private transactions tied to the IPO, one of which gives the chipmaker the right to buy shares at a 10 percent discount to the offering price.
What those investors are buying into is a balance sheet that looks more like a venture bet than an infrastructure yield-co. SB Energy's net loss widened to $3.2 billion in the first half of 2026, from around $250 million in the same period a year earlier — a deterioration driven largely by changes in the estimated value of its warrant liabilities. Revenue from its renewable-energy unit, which develops solar and battery systems, reached approximately $140 million in the first half of 2026, up 66 percent year over year. The company has no data centers currently in operation; its first 800-megawatt phase is scheduled to begin construction in 2026, with capacity expected to come online in phases beginning in 2028.
Yet the pipeline is enormous — and almost entirely concentrated in one place. SB Energy has contracts for nearly nine gigawatts of computing capacity in total, the vast majority tied to the PORTS-Pike Technology Campus in Pike County, Ohio, a site that has not yet been built. OpenAI executed 17 individual lease agreements in August for around eight gigawatts of computing capacity at the campus, which is backed by 10 gigawatts of power infrastructure under a 20-year lease.
The structure is a three-legged stool, and each leg is a version of the same counterparty risk in disguise. OpenAI is both the anchor tenant and an investor — it put $500 million in earlier this year and is projected to own a low-single-digit percentage of shares after the IPO. Nvidia is both the exclusive compute provider and the financier, providing credit support on the land, power, and shell buildout for an initial 4.25 IT-GW with an option for the remaining 3.75 IT-GW, on the condition that only Nvidia AI compute is deployed at the site, while also investing $1.5 billion directly. SoftBank is both the majority owner and OpenAI's largest backer, with its total commitment to the AI lab expected to reach approximately $65 billion by October.
SB Energy's own filing flags the obvious danger: it cautions that any weakening of OpenAI's finances could have adverse consequences for the company. The disclosure is unusually candid for a pre-IPO document — the business is not diversified, it is not operating, and its sole anchor customer is a privately held AI lab burning cash at a pace that has become a subject of public debate.
The company frames the arrangement as a partnership of necessity rather than convenience. In the company's August 17 announcement of the Ohio deal, Rich Hossfeld, co-CEO of SB Energy, said:
Infrastructure is vital for the AI economy. With SoftBank Group, OpenAI and NVIDIA, SB Energy is building power-first infrastructure at unprecedented scale while strengthening the communities that make it possible – protecting ratepayers, creating tens of thousands of well-paying jobs, and investing in infrastructure to revitalize Southern Ohio that has long shaped America's future.
Why Power, Not Chips, Is the Scarce Input
To understand why SB Energy can demand such terms, consider what has happened to U.S. electricity demand. Grid power supplied to data centers rose 25 percent in 2025 to about 64.4 gigawatts and has nearly tripled since 2020. Goldman Sachs Research expects U.S. data-center power demand to climb from 31 gigawatts in 2025 to 41 gigawatts in 2026 and 66 gigawatts in 2027, with total capacity roughly doubling to 95 gigawatts by the end of 2027. An academic forecast covering 2026 to 2030 projects data-center energy use could reach approximately 600 terawatt-hours by 2030 — 10 to 12 percent of all U.S. electricity — up from about 147 terawatt-hours, or 3.7 percent of national consumption, in 2023.
The bottleneck has shifted. For most of the 2010s, efficiency gains kept data-center energy use flat even as computing grew. Around 2019, compute-heavy AI workloads reversed that trend, and the constraint moved from silicon to substations. Getting a gigawatt-scale site connected to the grid now takes years of permitting, transmission upgrades, and interconnection queues that are measured in hundreds of projects. That is why SB Energy's deal includes at least $4.2 billion in new regional grid infrastructure through a partnership with AEP Ohio, and why the company is building at least 10 gigawatts of new generation — including 9.2 gigawatts of natural-gas generation backed by roughly $33 billion in generation investment — rather than simply leasing space.
The Ohio site sits on remediated land at the U.S. Department of Energy's Portsmouth Site, a public-private partnership that gives SB Energy something no competitor can replicate easily: a shovel-ready parcel with federal land, pre-cleared environmental status, and a utility partner willing to shoulder grid upgrades. In this light, the $5.5 billion warrant package is not irrational generosity. It is the price of locking in the only customer large enough to absorb eight gigawatts of capacity, on a site where the alternative — empty land with no tenant — has a value of zero.
There is a second-order point that most coverage of the filing misses. The warrants are not just a customer-acquisition cost; they are a mechanism that transfers AI capex upside from SB Energy's public shareholders to OpenAI. The warrants unlock in stages as SB Energy's market capitalization reaches predetermined thresholds. If the AI buildout accelerates and SB Energy's valuation rises, OpenAI captures a growing slice of that appreciation — paid for with equity, not rent. If the buildout slows, OpenAI keeps its lease flexibility while SB Energy's public investors absorb the full downside. The filing's own language makes this asymmetry explicit: the warrants are an inducement, and inducements are paid by the party that needs the deal more.
Cyclical Boom, Structural Shift, and the Signal That Breaks the Thesis
Is the AI power buildout a structural regime change or a cyclical capex wave? The answer is both, and confusing the two is the most common error in this trade. The demand side is structural: AI training and inference workloads are fundamentally more electricity-intensive than the cloud workloads of the 2010s, and global data-center electricity consumption is projected to grow about 15 percent per year through 2030 — more than four times faster than all other sectors combined. That is a durable shift in the shape of electricity demand, not an inventory cycle.
But SB Energy's own valuation is a cyclical bet layered on top of that structural trend. The company has no operating assets, no diversified customer base, and a loss profile that widened more than tenfold in a year. Its revenue today comes from solar and battery development, not from the AI data-center business that public investors are being asked to underwrite. A cyclical claim of this kind requires evidence that the capex wave can sustain itself through a downturn — and here the evidence is thin. Goldman Sachs' own forecast assumes only about 60 percent of capacity scheduled for the next year will materialize on time, dropping to roughly 50 percent over the following two years. In other words, the market's own analysts expect half of the announced AI data-center pipeline to slip or vanish.
The strongest counter-thesis is straightforward: SB Energy is not a power company; it is a leveraged option on OpenAI's survival and Nvidia's willingness to keep guaranteeing its financing. If either counterparty blinks, the structure unravels faster than a traditional infrastructure asset ever would. This view has gained traction among infrastructure investors who note that Nvidia's credit support is conditional on exclusive deployment of Nvidia compute — a condition that ties SB Energy's fate to chip pricing, export controls, and the competitive fortunes of a single silicon vendor. Jensen Huang, Nvidia's founder and chief executive, framed the deal in exactly those terms:
AI is becoming infrastructure – the foundation for intelligence in every industry – and land, power and shell have become vital in the age of AI. Now is the time to scale the AI infrastructure that will power the next industrial revolution.
That counter-thesis is correct as far as it goes, but it misses what the structure is actually designed to do. The point of the three-way entanglement — SoftBank's ownership, OpenAI's tenancy, Nvidia's credit support — is precisely to make each party's exit prohibitively expensive. SoftBank cannot walk away from OpenAI without impairing its largest private holding. Nvidia cannot withdraw support without stranding the compute it has committed to deploy. OpenAI cannot replace eight gigawatts of contracted capacity in a grid-constrained market without losing years. The structure is fragile in appearance and sticky in practice — until it isn't.
The falsifying signal is specific and observable. If OpenAI's lease commitments at PORTS-Pike slip materially from the roughly eight gigawatts contracted, or if Nvidia's credit support is renegotiated or withdrawn before the site reaches commercial operation, the structural thesis for SB Energy breaks. Watch the 2028 timeline: capacity is expected to come online in phases beginning that year, and any delay beyond that window would convert the warrants from an asset into a liability that no public shareholder wants to fund.
What Comes Next: Scenarios and Time Horizons
Short term (the IPO window): The listing is being pitched into a market that has rewarded AI-adjacent stories generously. Shares of SoftBank Group jumped as much as 20 percent to 6,039 yen on optimism around the SB Energy filing and reports of an OpenAI IPO, after the parent company posted record annual profit driven largely by valuation gains on its OpenAI stake. A successful $5 billion to $7 billion raise would validate the model and likely pull more capital toward power-and-data-center developers. The risk is that public investors, seeing a $3.2 billion half-year loss and zero operating data centers, demand a steeper discount than the private-round pricing allows.
Medium term (2027–2028, construction and commissioning): This is where the thesis is actually tested. SB Energy must convert nearly nine gigawatts of contracted capacity into operating assets while spending at least $4.2 billion on grid infrastructure and building 10 gigawatts of generation. The base case is that the Ohio campus reaches partial operation in 2028 with Nvidia compute deployed and OpenAI taking delivery under the 20-year lease. The downside case is a construction or interconnection delay that pushes revenue recognition past the warrant-vesting thresholds, leaving OpenAI's equity claim unearned and public shareholders funding cost overruns.
Long term (2030 and beyond, the structural question): If AI power demand reaches the 600 terawatt-hour range projected for 2030, SB Energy's early-mover position in grid interconnection rights could prove immensely valuable — the company would own one of the few shovel-ready gigawatt-scale sites in the Eastern Interconnection. But that outcome requires the structural demand trend to persist through at least one AI capex downturn, and it requires SB Energy to diversify beyond a single tenant before that downturn arrives.
The beneficiaries of this trade extend beyond SB Energy itself. Utilities with data-center exposure, grid-equipment suppliers, and developers holding interconnection rights in power-constrained regions stand to gain if the model validates. The exposed are the public shareholders asked to fund a pre-revenue asset at a valuation that already embeds years of flawless execution — and the communities, like Pike County, that have committed $80 million in combined community benefits on the promise of 35,000 construction jobs and 2,500 operating positions.
SB Energy's IPO is the first public test of whether the AI boom can be securitized as infrastructure, or whether it remains a venture bet dressed in concrete and copper. The $5.5 billion warrant giveaway tells you which way the insiders are betting: they are paying OpenAI to stay, because the alternative — a world where AI demand cools and eight gigawatts of contracted capacity sits idle — is a world where this equity is worth nothing at all.
Explore more exclusive insights at nextfin.ai.
