Title: ⚡️ How SB Energy Makes Money — The $439 billion bet behind AI infrastructure Show: App Economy Insights / How They Make Money (appeconomyinsights.com) — Premium edition Guest: Bertrand Seguin (author) Date: 2026-09-15 URL: https://www.appeconomyinsights.com/p/how-sb-energy-makes-money Length: written post (~3,500 words) — no timestamps Note: Paid (Premium) written post; API fetch truncated at the paywall, so the full text was captured verbatim via Stephen's logged-in Chrome session (custom-domain /account SSO warm-up first; title + byline "SEP 15, 2026" verified). Chart images viewed and their figures transcribed below as [Chart image — …] blocks at their position in the post; the header logo and the management headshot carry no data. Subscription-promo boilerplate and share/like counters omitted.
Welcome to the Premium edition of How They Make Money.
Over 300,000 subscribers turn to us for business and investment insights.
In case you missed it:
📈 How to Invest in IPOs
💍 How Oura Makes Money
📊 Earnings Visuals (8/2026)
☁️ Oracle: 850 Megawatts Later
⚡️ SB Energy is going public
If you have been watching the race to power the AI boom, this is one of the most aggressive bets on physical infrastructure yet.
The stakes just got higher. With AI leaders like Dario Amodei and Sam Altman openly discussing a more cautious pace at the frontier, investors face a fundamental question. How durable is the physical infrastructure buildout if frontier development slows?
SB Energy sits right at the center of that question.
The company generated just $139 million of revenue in the first half of 2026. It now reports roughly $439 billion of contracted backlog, almost entirely tied to data centers.
There’s one catch. None of those data centers are operating yet.
SB Energy started as a renewable-power developer. The AI boom has transformed it into something much bigger. It now develops gigawatt-scale data centers and the power infrastructure behind them, with OpenAI as its largest customer and NVIDIA and SoftBank helping fund the buildout.
In effect, SB Energy is becoming a physical landlord of the AI boom, owning the infrastructure underneath the compute.
The gap between SB Energy’s revenue today and the future revenue already under contract is the entire IPO story.
I condensed 300+ pages of the S-1 into a clean breakdown, supported by our signature visuals. By the end, you'll have a clear view of the SB Energy investment case.
Today at a glance:
Overview Business model Financial highlights Risks & challenges Management Use of proceeds Future outlook Personal take
[Chart image — "How SB Energy Makes Money" (income statement, 6M YTD FY26, ending June 2026): Customer contracts $58M (−5% Y/Y) + Power swap derivatives $80M (+271% Y/Y) → Revenue $139M (+66% Y/Y). Costs & expenses ($3,324M): cost of operations ($41M); D&A and accretion ($48M); interest & other ($48M); general & admin ($602M), includes $590M of stock-based compensation; fair value revaluation of OpenAI warrants ($2,573M). Net loss ($3,192M). Contracted backlog $439B; projected CapEx $178B.]
1. Overview
SB Energy’s transformation is easy to see in one place: Milam County, Texas.
In 2024, the company completed the Orion Solar Belt, a 900 MW solar complex with Google as its anchor customer. Two years later, SB Energy is building a 1.2 GW data center campus for OpenAI in the exact same county.
That captures the entire evolution of the company. SB Energy went from supplying clean power to third-party facilities to building and owning the physical data centers themselves. Milam County was only the beginning.
Mastering the grid (2019–2023)
SB Energy was founded in 2019 as a SoftBank Group company focused on large-scale renewable energy. It spent its early years mastering the hardest parts of infrastructure development: finding land, securing grid connections, contracting power, arranging project debt, and managing construction over multi-year cycles.
Those capabilities once looked like the plumbing of a slow-moving utility business. The AI boom changed that overnight.
Hyperscalers need massive compute, but chips are only part of the bottleneck. Gigawatt-scale data centers require land, high-voltage electricity, transmission access, environmental permits, and years of buildout. SB Energy already knew how to solve many of those problems.
Moving up the stack
Instead of remaining an upstream power supplier, the company moved up the stack. It now operates as an integrated data center and power platform, developing the physical infrastructure where AI hardware will live.
The scale changed rapidly. At its PORTS-Pike campus in Ohio, SB Energy has contracted 8 GW of capacity for OpenAI, with NVIDIA providing the compute infrastructure and initial phases coming online in 2028.
[Chart image — "From Powering to Building Data Centers: How SB Energy moved up the infrastructure stack": 2024 / Renewable power — Orion Solar Belt, Milam County, Texas, 900 MW solar complex; role: power generator; anchor customer: Google; use case: supports data-center power demand. → 2028+ / AI infrastructure — PORTS-Pike Campus, Pike County, Ohio, 8.0 GW AI data-center campus; role: data-center developer / landlord; tenant: OpenAI; compute + strategic partner: NVIDIA. "Same land-and-power expertise. Much bigger role in the AI economy."]
Across its portfolio, the company reports roughly 5 GW of power projects operating or under construction, along with three AI campuses in development.
SB Energy doesn't design chips, train AI models, or rent cloud instances. Its bet is on the land, power, and buildings that AI systems require.
Takeaway: SB Energy spent its first five years learning how to build power infrastructure at utility scale. The AI boom turned that expertise into physical infrastructure for the AI economy.
2. Business model
SB Energy monetizes that infrastructure through three segments:
🏢 Data Centers: Develop, build, own, and lease AI data-center infrastructure.
⚡ Standalone Power: Develop and own large-scale power generation and battery projects.
🛠️ Solutions: Provide the design, construction, and operating capabilities supporting those assets.
The data-center model is the most important to understand.
🏢 Data Centers
SB Energy’s preferred model is turnkey delivery.
It secures the land and power, develops the site, and delivers a completed “white-space” data center ready for the customer to install its computing equipment. SB Energy then owns the facility and collects rent.
So SB Energy essentially stops below the GPU.
[Chart image — "SB Energy Stops Below The GPU: It owns the power, land, and data center shell": layer 1 Power & Grid Interconnection (solar, batteries, substations, interconnection) and layer 2 Land & Data Center Shell (powered land, cooling, white space) — SB Energy owns / develops; "SB ENERGY STOPS HERE"; layer 3 Compute Equipment (racks, GPUs, networking) and layer 4 AI Models & Applications (foundation models, inference, applications) — tenant installs / operates. "The tenant upgrades the chips. SB Energy keeps collecting rent." 15–20 year triple-net leases, annual escalators. "SB Energy monetizes the physical layer of AI infrastructure."]
Its leases generally run for 15 to 20 years and are structured as triple-net leases. That means the tenant ultimately bears most ongoing property-level expenses, while SB Energy receives contractual rent.
The rent itself is based on a yield-on-cost model. In simple terms, SB Energy invests the capital required to build a facility and prices the lease to earn a contractual return on that investment. Annual escalators then increase rent over time.
That creates a fairly intuitive model:
Secure land + power Build the data center Deliver capacity Collect rent for decades
The hard part happens upfront. SB Energy assumes the development and construction risk, while the long-term lease creates predictable cash flow once the facility is delivered.
What could the economics look like?
Property-level margins should look very attractive once facilities turn on. At Cosmos, for example, the lease includes 100% operating-expense pass-through, meaning the tenant covers the ongoing property costs while SB Energy collects rent tied to a return on project cost.
But investors shouldn’t confuse high property-level margins with software-like free cash flow. These assets require enormous amounts of capital, with depreciation and, more importantly, interest expense absorbing part of those economics. The business may look highly profitable at the property level, but shareholder returns will ultimately depend on the spread between lease yields and financing costs.
⚡ Standalone Power
SB Energy uses a similar playbook in power. It develops, builds, owns, and operates utility-scale solar and battery-storage projects. It primarily sells electricity through long-term power purchase agreements (PPAs) with utilities, corporations, and government entities, generally lasting 15 to 25 years.
Some of these assets remain standalone businesses. Others can be purpose-built to support SB Energy’s own data centers.
That integration is important. For most data-center developers, power is something they need to procure. For SB Energy, power development is already part of the product.
🛠️ Solutions
The third segment is less about generating standalone revenue and more about execution. SB Energy has brought design, engineering, construction management, and operations capabilities in-house, including through its acquisition of data-center specialist Studio 151. The Solutions business primarily supports its Data Center and Standalone Power segments today.
That lets SB Energy control more of the project lifecycle instead of relying entirely on outside contractors.
Takeaway: SB Energy is effectively an infrastructure landlord. It takes on the difficult and capital-intensive work of securing land and power and building the asset upfront, then converts that investment into contractual cash flows that can last for decades.
3. Financial Highlights
SB Energy’s financials are unusual because they barely reflect the business investors are being asked to value.
Revenue reached $232 million in FY24, before falling 8% to $214 million in FY25. Growth returned in the first half of FY26, with revenue rising 66% Y/Y to $139 million. Almost all of it still came from the standalone power business.
The data center segment generated only a negligible amount of revenue during the period. SB Energy expects that to begin changing when the first phase of the Cosmos Technology Campus starts paying rent, currently expected in Q4 FY26.
So a genuine financial inflection lies ahead. Most of the company’s future value has barely appeared on the income statement yet.
The $3.2 billion loss is mostly accounting
At first glance, profitability looks disastrous.
SB Energy reported a $3.2 billion net loss in the first half of FY26. But two non-cash expenses explain almost all of it:
$2.6 billion from the revaluation of warrant liabilities
$590 million of stock-based compensation
The warrant expense largely reflects an increase in the estimated value of warrants issued to OpenAI. As SB Energy’s implied equity value increased, so did the accounting liability attached to those warrants.
That makes the GAAP loss a poor representation of the economics today. SB Energy used only about $56 million of cash from operations during the first half.
I wouldn’t dismiss the non-cash charges entirely. Stock compensation is still dilution, and OpenAI’s warrants represent a very real economic cost of securing its most important customer. But the $3.2 billion headline loss makes the operating picture look far worse than it actually is.
$439 billion backlog does not equal near-term revenue
The more important financial disclosure is when that backlog actually becomes revenue. The company reports roughly $439 billion of contracted revenue, including about $430 billion from data centers. The sheer size is remarkable, but the timing matters just as much.
Roughly $357 billion is expected to be recognized more than eight years from now. The weighted-average remaining term of its data center contracts is almost 20 years.
That means this is very different from the backlog at a software company or chipmaker. SB Energy has contracted decades of future rent, not hundreds of billions of revenue that will arrive anytime soon. And those contracts still need to be turned into physical assets.
SB Energy estimates approximately $178 billion of future capital expenditures associated with its existing backlog, including $174 billion for data centers.
[Chart image — "SB Energy's Backlog Comes With a Bill": contracted backlog (future contracted revenue over decades) $439B = $82B years 1–8 (19% of total) + $357B year 9+ (81% of total). Associated future CapEx (estimated capital required to build the infrastructure) $178B = $174B data centers (98%) + $4B standalone power (2%). Key takeaways: 81% of backlog sits beyond year 8; ~$178B of future CapEx needs to be deployed.]
That number puts the backlog in perspective.
The opportunity is enormous, but so is the amount of capital required to capture it. SB Energy will need to finance and build these projects over many years before most contracted revenue reaches the income statement.
Takeaway: SB Energy’s current results tell us very little about its future scale. The reported loss is distorted by non-cash accounting, while the real economics sit in decades of contracted revenue that still require an extraordinary amount of capital and execution to materialize.
4. Risks and Challenges
SB Energy has secured extraordinary demand. Now it has to actually deliver the infrastructure behind it.
🏗️ Execution risk is enormous: SB Energy’s large data center campuses are still under development. Each project requires land, permitting, power, grid connections, equipment, and construction to come together on schedule. Delays can postpone rent commencement, and some leases give customers rent abatements, reimbursement rights, or termination rights if SB Energy fails to deliver contracted capacity.
🎯 A few customers drive the entire story: OpenAI anchors the massive PORTS-Pike campus, while SoftBank is another major data-center customer. Customer concentration is expected to increase as the data center business scales. NVIDIA provides substantial credit support for portions of OpenAI’s leases, reducing some counterparty risk, but SB Energy’s future remains heavily tied to the spending plans and financial health of a small group of AI companies.
💰 Financing determines the returns: SB Energy must fund projects years before collecting most of the associated rent. It expects to rely heavily on project-level debt and other external financing. Even with committed tenants, higher interest rates, construction overruns, or less favorable financing terms could materially reduce returns to equity holders.
⚡ Power is both the moat and the bottleneck: SB Energy’s renewable-development expertise gives it an advantage in securing power for enormous campuses. But gigawatt-scale projects still depend on transmission capacity, grid interconnections, generation availability, utilities, and regulators. A finished data center has little value if sufficient electricity cannot reach it on time. PORTS-Pike alone is expected to require 9.2 GW of new gas generation, while the filing warns that gas-turbine lead times can stretch to seven years.
🤖 AI infrastructure demand has to remain durable: Long-term leases protect much of the revenue once facilities are delivered, but SB Energy’s ability to keep developing new campuses depends on hyperscalers continuing to spend aggressively on AI infrastructure. A slowdown would hurt future development opportunities and potentially the residual value of assets built specifically for very large AI workloads.
Takeaway: SB Energy’s biggest risk is no longer finding customers. It is turning enormous contractual commitments into functioning infrastructure while preserving attractive returns. The leases provide visibility, but financing, construction, and power availability will determine how much of that opportunity ultimately reaches shareholders.
5. Management
SB Energy is unusual among IPO candidates for having two co-CEOs with complementary backgrounds.
Rich Hossfeld brings the finance and investment side. Before leading SB Energy, he worked in renewable-energy strategy and investing, including roles at First Solar and True North Venture Partners.
Abhijeet Sathe comes from the operating side. He has more than three decades of engineering and renewable-energy experience, with previous leadership roles at SunEdison, SunPower, and GE.
[Image — co-CEO headshots (Source: SB Energy); no data.]
That combination makes sense for a company whose success depends equally on raising enormous amounts of capital and actually building complex infrastructure.
The broader leadership team includes:
Gaetan Frotte, CFO: Joined in 2023 after more than 16 years at NRG Energy, where he worked across corporate and project finance and helped raise more than $25 billion of financing.
Ryan Bates, General Counsel: Oversees legal and compliance as SB Energy transitions into the public markets.
The board mirrors the AI ecosystem
SB Energy’s proposed board includes executives and investors closely connected to the companies that fund and use its infrastructure.
Alex Clavel leads SoftBank Investment Advisers, while longtime SoftBank executive Ron Fisher is also on the board. Sachin Katti, who leads compute strategy and infrastructure efforts at OpenAI, is another nominee.
The filing also names directors with deeper public-market and financial experience, including Alexi Wellman, a former Yahoo Finance executive and KPMG audit partner.
OpenAI also retains the right to designate a director as long as it owns more than 5% of SB Energy’s outstanding capital stock.
SoftBank will remain in control
This is not a conventional one-shareholder-one-influence IPO.
SoftBank will retain more than 50% of the voting power after the offering, making SB Energy a “controlled company” under Nasdaq rules. SB Energy therefore plans to rely on exemptions that can allow it to operate without a majority-independent board or fully independent compensation and nominating committees.
Takeaway: SB Energy’s leadership looks well matched to the challenge ahead. The co-CEO structure combines infrastructure execution with capital markets expertise, while the board brings SoftBank and OpenAI directly into the governance structure. The trade-off is that public shareholders will have limited influence while SoftBank remains firmly in control.
6. Use of proceeds
SB Energy hasn’t disclosed how much it plans to raise yet. The initial S-1 leaves the share count and IPO price blank.
Management expects to use substantially all of the proceeds for working capital, operating expenses, and CapEx required to develop its data centers, power generation, storage, and related infrastructure. A portion may also go toward technology development, acquisitions, or repayment of debt under its Hickory facilities.
Unlike Oura (that we discussed last week), there is little ambiguity about why SB Energy wants access to public capital. The company has contracted a buildout far beyond what its current operations can finance internally.
The IPO won’t fund that expansion by itself. Instead, it becomes another piece of a much larger financing stack that will include public equity, project-level debt, and strategic capital.
NVIDIA is investing alongside public shareholders
NVIDIA has agreed to purchase $1.5 billion of non-voting Class N shares directly from SB Energy in a private placement at the IPO price. The transaction is expected to close alongside the offering, although neither deal is conditional on the other.
NVIDIA also has a separate $1.5 billion prepaid forward agreement with SB Energy’s parent, Energy Global. NVIDIA prepaid the amount in August and will receive shares priced at 90% of the IPO price if the offering occurs.
So NVIDIA has roughly $3 billion of IPO-linked equity exposure, but only half represents fresh capital being invested directly into SB Energy at the IPO.
That distinction matters because NVIDIA is simultaneously an investor, infrastructure supplier, and financial backstop for SB Energy’s largest planned data-center campus.
[Chart image — "SB Energy's AI Ecosystem: Customers, capital, governance, and suppliers in a single closed loop": OpenAI (demand / tenancy) — largest customer / anchor tenant; 8 GW PORTS-Pike campus; warrants + board designation right; drives future demand. SoftBank (control + sponsorship) — >50% voting power; controlled company; strategic sponsor; broader AI ecosystem. NVIDIA (capital + compute + credit support) — $1.5B direct Class N investment; $1.5B Energy Global forward; compute supplier; credit support (OpenAI leases). "SB Energy sits at the center of a circular AI infrastructure ecosystem."]
NVIDIA is not just investing equity. Its credit support helps SB Energy finance a campus that will exclusively host NVIDIA infrastructure, while OpenAI remains the tenant. That tightly aligns NVIDIA’s chip sales with SB Energy’s ability to build the real estate.
Takeaway: SB Energy isn’t going public simply to create liquidity for existing shareholders. Access to public equity is part of the financing model itself. The IPO should provide fresh capital for the initial buildout, but the much bigger test will be whether SB Energy can repeatedly raise project financing on attractive terms as its contracted campuses move into construction.
7. Future outlook
SB Energy’s near-term growth is largely spoken for. The more interesting question is what comes after the current backlog.
I see three main growth vectors.
Make existing campuses bigger
Once SB Energy controls the land, interconnection, permits, and local relationships at a site, adding another phase should be easier than starting from scratch.
Management already plans multi-phase expansion at Milam, Pike, and other campuses. A successful site can therefore keep growing long after the first building is delivered. That makes each campus potentially more valuable than its initial contracted capacity suggests.
Add more power around the data centers
SB Energy’s biggest differentiator may be its ability to develop both sides of the equation.
More data-center load can support additional nearby generation and storage, while greater control over power can make future campuses easier to develop. That creates a potential flywheel between SB Energy’s two core businesses.
The company does not need every power project to sit directly next to a data center, but the ability to originate both gives it more ways to monetize the same infrastructure ecosystem.
More sites and more customers
Beyond the current portfolio, SB Energy is already pursuing multiple gigawatts of additional data-center development.
Its focus is primarily on Texas, the Mid-Atlantic/Midwest, and the Western US, while management is also evaluating other US and international markets. The customer base could broaden as well, with SB Energy targeting additional hyperscalers, sovereign customers, and large enterprises.
OpenAI and SoftBank helped SB Energy reach enormous scale quickly, but the broader question is whether customers outside of that ecosystem will choose the platform too.
Takeaway: SB Energy’s current backlog may only be the first layer of the opportunity. The bigger upside comes from expanding successful campuses, building more power around them, and proving the model can attract new customers across new markets.
8. Personal take
I always come back to the same golden rule for IPOs:
IPO stands for “It’s Probably Overpriced.”
SB Energy may be one of the hardest companies to apply that rule to.
Reports suggest it could seek a valuation of more than $50 billion, although the initial S-1 does not yet disclose a price range. And good luck choosing the multiple it should apply to.
Current revenue is almost beside the point. The data-center business investors are really trying to value has barely started contributing.
The entire thesis depends on one question:
How much of the economics from SB Energy’s contracted projects will ultimately accrue to shareholders?
That comes down to the spread between the return SB Energy earns on each project and the cost of building and financing it.
Think of the project math like a rental property at massive scale. Data-center developers typically target an annual return of roughly 8% to 11% on what they spend to build the asset, before using debt. In industry terms, that’s the unlevered yield on cost. If they can borrow at 6% to 7%, the equity returns look great. But that spread is unforgiving. If construction costs jump by 15% or loan terms worsen during the multi-year buildout, that profit margin shrinks fast.
A 20-year lease looks extraordinarily safe on paper. But if SB Energy has to inject more equity to absorb cost overruns, far less value flows down to public shareholders.
📈 Bull Case
The bull case is that SB Energy has positioned itself around powered land at enormous scale, which is genuinely scarce
AI companies can buy GPUs. Finding sites with enough land, power, grid access, permits, and infrastructure to support multiple gigawatts is much harder.
SB Energy already has those development capabilities and has translated them into long-duration contracts.
If the company can consistently build campuses at attractive yields and finance a large portion of the cost with project-level debt, the equity economics could be very powerful.
There is also an important advantage to stopping below the GPU. SB Energy does not need to keep replacing the compute itself as technology changes. The tenant can upgrade the chips while SB Energy continues collecting rent on the land, power, and buildings.
If the first projects prove those economics, the model becomes much easier to replicate across additional phases and campuses.
📉 Bear Case
The bear case is that the returns on the backlog could disappoint.
Infrastructure investing is highly sensitive to relatively small changes in construction costs, financing rates, leverage, and project yields. When the capital base becomes this large, even modest differences can dramatically change the value left for equity holders.
That makes SB Energy very different from a software company where incremental revenue can fall through at extremely high margins. Here, enormous future revenue comes attached to an enormous amount of capital.
Public investors may therefore be paying north of $50 billion before SB Energy has demonstrated the economics of its first major operating data centers.
The backlog may already lock in enormous future cash generation, but it does not lock in enormous shareholder value. That depends on how much equity SB Energy must invest, when the cash arrives, and what the assets are worth at the end of the leases.
Takeaway: SB Energy may have secured one of the most valuable positions in the AI infrastructure buildout. But investors are not buying the backlog itself. They are buying the returns SB Energy can earn on the capital required to serve it. At a potential valuation above $50 billion, I’d want to see the first major campuses operating and the project-level economics proven before getting comfortable with the price.
That’s it for today.
Happy investing!
Thanks to Fiscal.ai for being our official data partner.
Author's Note (Bertrand here 👋🏼): The views and opinions expressed in this newsletter are solely my own and should not be considered financial advice or any other organization's views.
Disclosure: I own NVDA in App Economy Portfolio. I share my ratings (BUY, SELL, or HOLD) with App Economy Portfolio members.