SB Energy, a SoftBank subsidiary, files for AI power infrastructure IPO
SB Energy, a subsidiary of SoftBank, has submitted an IPO application to the U.S. SEC, planning to list on both Nasdaq and Nasdaq Texas under the ticker "SBE"; the company has disclosed that its recent revenue, project financing, and development plans are "highly dependent" on OpenAI's continued fulfillment of leases and related agreements.
OpenAI is both a tenant and a shareholder of SB Energy, with OpenAI CEO Sam Altman being an early personal investor in the company. SB Energy stated in its prospectus that this concentration of customer and equity relationships directly links the revenue certainty of recently commenced projects, project-level financing arrangements, and expansion pace to OpenAI's operational performance.
As of the filing date, the company has no operational data centers, and its AI data center business has yet to generate revenue. In the first half of 2026, SB Energy reported a net loss of approximately $3.2 billion, with revenue of about $139 million, primarily from its existing energy business; the loss reflects significant upfront capital investments for its data center strategy.
The company’s construction of AI data center parks heavily relies on external financing and support from partners. Nvidia previously announced it would provide up to $105 billion in financing support for the Ohio data center being built for SB Energy and leased by OpenAI; SB Energy CEO Rich Hossfeld stated that Nvidia's involvement would help secure investment-grade financing for the project.
According to the Wall Street Journal, SB Energy aims to raise approximately $5 billion to $7 billion through the IPO, which could begin trading as early as this month; the company has not disclosed the offering price, offering size, or exact listing date, so the aforementioned financing targets do not constitute final terms.
SB Energy also cautioned that community opposition, local bans, and public resistance to AI infrastructure could impact its data center and power generation business; risks also include lower-than-expected corporate adoption of AI, technological advancements rendering facilities obsolete, regulatory changes, and large-scale cloud vendors slowing capital expenditures.
From a market mechanism perspective, IPO investors are not purchasing cash flows generated by operational data centers but rather development-stage assets supported by OpenAI leases, Nvidia financing, SoftBank control, and future AI computing demand. Beneficiaries include infrastructure developers with long-term power resources, grid capacity, land, and financing channels, as well as suppliers who can convert project demand into GPU and cloud service revenue; those under pressure are small and medium-sized data center developers unable to secure low-cost capital, lacking anchor clients, or facing local approval hurdles.
Source: Public Information
ABAB AI Insight
SB Energy originated as SoftBank's platform for renewable energy, primarily focusing on solar project development, construction, and operation; this IPO shifts the company's focus towards AI data centers, power, and park development. Through the Vision Fund and ongoing investments in OpenAI, SoftBank has transitioned its capital path from individual internet projects to a closed loop of "model companies—chip supply—power assets—data centers." SB Energy is being integrated into this loop, taking on the most capital-intensive, longest construction cycle, and most reliant on local permits segment.
OpenAI is not just acquiring server capacity but is becoming a credit anchor for project financing through leases. Nvidia's provision of up to $105 billion in financing support is also aimed at more than just selling GPUs: if the funds encourage OpenAI to long-term lease SB Energy's parks, and those parks deploy Nvidia computing devices, the demand for chips, leases, and infrastructure financing will mutually reinforce each other. This structure can reduce the initial financing difficulty for projects but also tightly binds the risks of clients, developers, chip manufacturers, and investors.
This differs from Amazon, Microsoft, and Google's past self-built global data centers: large-scale cloud vendors typically support capital expenditures with their own balance sheets, mature cloud revenues, and diverse customer bases; SB Energy currently lacks revenue from operational data centers but hopes to advance large-scale projects through the IPO and partner guarantees. A closer historical case is the fiber networks built around anticipated bandwidth demand during the telecom bubble, and liquefied natural gas projects locking in financing with long-term "take-or-pay" contracts: what truly determines asset value is not predicted demand but tenant credit, contract duration, financing terms, and actual production speed.
Essentially, this represents a restructuring of the industrial chain: the bottleneck in AI competition is shifting from model parameters and GPU procurement to available power, grid permits, construction capacity, and long-term capital. Model companies need computing power, computing power needs chips, chip clusters need power and land; thus, data center developers are gaining strategic positions that were previously absorbed internally by cloud vendors. However, when anchor clients are concentrated in OpenAI and assets are not yet operational, infrastructure assets have not escaped the volatility of tech demand, instead concentrating AI adoption risks, financing risks, and construction risks onto the same balance sheet.
ABAB News · Cognitive Law
Without power and grid connection, computing power commitments are merely future orders
Anchor tenants can finance but also amplify concentration risks
The bottleneck of the AI race will ultimately fall on infrastructure assets