This unique power company is a play on the Anthropic IPO. Bank of America says it has room to run

ERock shares jumped 23% after reporting a 1,000% increase in order backlog, pointing to $1.7B of forthcoming business tied to a 470MW power purchase agreement from Anthropic. Bank of America raised its ERock price target to $19 from $16 and reiterated a buy, citing the award as the company’s third major data-center contract. The deal is framed as part of the AI-driven data-center buildout that is straining U.S. power grids, supporting demand for ERock’s RockBlock modular generators (with installation/O&M expected only after separate contract scopes are finalized).

Analysis

The market is likely pricing a customer-validation story, but the real mechanism is a financing and conversion story: a large announced backlog can reduce perceived customer-acquisition risk and expand the multiple if investors start underwriting a repeatable pipeline into 2027-28. That said, the quality of earnings is still low near-term because the disclosed scope appears product-led, with installation and O&M deferred; that means revenue recognition, cash conversion, and maintenance annuity potential are all back-end loaded. The stock can stay volatile because the gap between headline backlog and near-term P&L is wide.

Second-order effects matter more than the single contract. If AI developers increasingly buy behind-the-meter generation to bypass grid bottlenecks, that is bearish for utilities with heavy capex plans and for grid-tied power-equipment vendors that rely on slower interconnection cycles, while it is supportive for gas supply, compression, and microgrid-adjacent names. The competitive risk is that this validates a broad category, not just one supplier: larger industrials can reprice into the same demand pool, and any execution hiccup at a small-cap can hand share to better-capitalized rivals.

The contrarian view is that the move may be overdone in the near term because investors are extrapolating backlog into funded revenue without enough visibility on installation timing, margin mix, or customer concentration. The key falsifier is if the company cannot convert this into signed O&M or additional awards within 1-2 quarters, or if project timelines slip into 2028+, which would compress the multiple back toward a hardware-only story. A near-term reversal could also come from a broad AI power trade de-rating if financing or data-center capex slows.

Key Decisions for Investors

• Tactically long EROC only on pullbacks into the mid-teens, with a hard risk line below the post-IPO low; this is a momentum/speculative order-flow trade, not a fundamental compounder call.
• Use a collar or call spread on EROC for the next 1-3 months; upside is tied to follow-on contract announcements, but downside is sharp if the market starts discounting 2027-28 revenue.
• Pair trade: long EROC / short CMI or GNRC as a relative-value basket on the thesis that smaller modular power providers are capturing AI grid-avoidance demand faster than diversified incumbents.
• Watch XLU and utility-exposed names for 6-18 month underperformance if behind-the-meter generation becomes a real procurement channel for AI data centers; use any strength there to fade if additional project awards keep landing.
• Alert level: if EROC fails to hold above its post-announcement breakout zone after the next earnings update or if backlog converts slower than modeled, take profits quickly and reassess the thesis.