AI data center firm SB Energy, which is backed by Softbank, Nvidia and OpenAI, files for IPO

SB Energy filed for an IPO, targeting $5B–$7B, while reporting net losses of about $3.2B in 1H 2026 on only ~$139M in revenue and noting none of its data centers are operational yet. The S-1 highlights heavy concentration risk: revenue and project financing plans are “substantially dependent” on OpenAI’s performance as both tenant and equity investor. Despite backing and Nvidia’s $105B financing commitment for an OpenAI Ohio data center, the filing flags community opposition and potential hyperscaler capex deceleration as key risks.

Analysis

This looks less like a normal growth IPO and more like a public-market financing wrapper around a single customer relationship. Near term, that can be constructive for NVDA because it legitimizes the capex cycle and may pull forward funded GPU demand, but the second-order effect is a higher skepticism premium on any AI infra business whose cash flows are effectively tenant-credit risk.

The bigger market consequence is that the cost of capital for the whole buildout may rise if the book shows investors want hard contracts, not narrative. If that happens, the winners are balance-sheet providers with contracted backlog and asset-heavy vendors that can finance themselves; the losers are developers that need repeated equity rounds and depend on one anchor tenant to keep project financing open. Watch for contagion into data-center REITs and electrical infrastructure names if the IPO trades poorly.

The contrarian point is that public backlash and permitting risk may matter more than GPU supply over the next 6-18 months. Consensus is likely underestimating how quickly a "strategic AI asset" can be re-rated as a levered, pre-revenue industrial project if OpenAI or broader hyperscaler capex slows. Falsifiers: a materially diversified tenant base, investment-grade project funding at tight spreads, or a stable first-quarter post-IPO trading range despite weak AI sentiment.

Key Decisions for Investors

• Do not chase the IPO on day 1; fade any opening pop if the stock prices like a software multiple rather than a project-finance multiple. Conditional starter short only if the debut values the business at >5x forward contracted revenue or the book is dominated by related-party demand.
• Stay long NVDA on pullbacks, but hedge with a small SOXX/SMH short if the market starts reading this as circular financing rather than incremental end-demand. Time horizon: 1-3 months, until the IPO book and first filings clarify the customer mix.
• Prefer long ETN or PWR over SB Energy exposure to express the AI power buildout. These names monetize the infrastructure wave without taking OpenAI concentration or permitting risk.
• Set a watch item on financing spreads and permitting headlines. If project debt widens materially or local moratoria spread, reduce exposure to AI infra developers and shift toward balance-sheet-light suppliers.