Credible AI power orders.
A larger valuation hurdle than quote sites imply.

The first correction: this is not a $600m whole-business valuation

The June 30 reference count includes approximately 272m A/B/M equivalents, as management explained on the call, plus restricted awards. Class B exchange does not create new whole-business ownership; it moves existing ownership into public shares. Class M participation is an additional economic claim. Loss-period GAAP diluted EPS excludes these securities because they are anti-dilutive under accounting rules; investment valuation still needs them.

Capital bridge. June unrestricted cash was $626.636m; restricted cash was another $34.225m. Operating lease liabilities were $29.771m, with a small $0.397m short-term note in accrued payables. Using the reference diluted value gives approximately $2.85bn enterprise value after all unrestricted cash and these recorded financing claims. This is about 6.3× the $450m 2026 revenue-guide midpoint, or 6.8× the provider's $418.4m 2028 EBITDA estimate. Price and balance-sheet dates differ, future stock dilution is not fixed, and all cash is not surplus.

Tax and governance claims. ERock owes specified legacy owners 85% of realized or deemed specified tax savings under its tax receivable agreement; it retains 15%. This is not 85% of operating profit, but it can consume cash beyond a simple tax forecast. The prospectus illustrates approximately $1.19bn of undiscounted lifetime payments under specified exchange and taxable-income assumptions, and a $658.1m hypothetical immediate termination payment at the $21.50 IPO price. These are conditional illustrations, not current booked debt or amounts presently due. A $658.1m reserve would equal about $2.19 per 300m future shares, but is not an estimate of the actual future liability. Sponsor voting power is approximately 46.8%, a substantial influence; the Class B owners collectively hold about 78% of votes before exchanges. Potential lock-up expiry around December 7 is an inferred first trading date after the normal 180-day period, subject to exceptions or waiver.

Sources: Q2 10-Q, Notes 12–16 and cover; August 12 earnings transcript; LLC agreement and prospectus in the source register. Class M's complete grant-threshold distribution was not located, so an exact October-price conversion curve is not presented.

What the company does—and why the demand is real

ERock, formerly Enchanted Rock, designs and assembles modular natural-gas generator systems called RockBlocks. These are engine-driven generators, rather than BW's boiler-and-steam-turbine starting architecture. ERock integrates hardware, controls, installation and ongoing operation into systems that can provide bridge power before grid connection, backup power during outages and dispatchable capacity when the grid needs it.

Market fit. Rapid on-site power can be valuable when a data center's grid connection is years away. Once the grid arrives, systems can continue as backup or dispatchable assets. The value depends on customer economics, gas infrastructure, air permits, uptime and delivery—not simply on an AI spending forecast.

Competitive comparison. Caterpillar, Cummins and other engine suppliers compete directly or through integrators; GE Vernova and Siemens Energy address larger turbine configurations. BW offers a distinct thermal architecture. ERock's integrated controls, field experience and customer relationships may help, but no matched independent benchmark reviewed here establishes universal technical or cost superiority. The company says it controls engine designs, not every upstream component or supplier capacity.

Primary commercial evidence: August 11 release, August 12 presentation and inspected transcript. No current peer valuation screen was completed; peers above are product comparisons, not interchangeable valuation comparables.

Fundamentals snapshot: a fivefold step-up in half-year revenue

H2 revenue must be approximately 5.1–5.5× H1 revenue. At the midpoint, 2026 revenue of $450m would be 2.46× 2025's $183.145m, or 145.7% year-over-year growth. That acceleration is plausible given deliveries and customer orders, but leaves considerable execution risk over the next two reported quarters.

Earnings quality. The reported $0.06 loss per share covers only the June 11–30 post-IPO period and Class A attribution; it is not the consolidated loss for all of Q2 divided by the full economic share base. Provider standardized Q2 net income and historical EPS differ from the filing, so filed consolidated results control this analysis. Profit-growth percentages are not meaningful when crossing from losses to profits.

Cash flow is favorable—but deposits are doing the funding

H1 operating cash flow was $268.939m, versus $0.436m a year earlier. Contract liabilities increased $358.380m, to $528.405m. These customer advances improve contract credibility and fund delivery; they are obligations to perform future work. Subtracting that one inflow from reported operating cash flow gives −$89.441m, an illustrative bridge rather than normalized cash burn. Receivables and inventory also absorbed cash.

Capital expenditure was $8.835m. The large positive cash balance and low assembly capex are better starting conditions than a debt-heavy plant developer, but they do not yet prove high cash conversion from recurring profits. Treating the whole $626.6m cash balance as excess distributable cash would be aggressive.

Sources: Q2 10-Q and release. Contracts require advance milestone payments; this is commercial funding, not equivalent to free cash-flow earnings. One customer represented 62% of receivables at June 30. Execution and customer concentration are material.

Growth Estimates Consensus: strong growth is already expected

Provider revenue estimates grow from $451.7m in 2026 to $1.419bn in 2027 and $2.174bn in 2028: +214.2% and +53.2%, or 119.4% annual growth over two years. EBITDA rises from $5.5m to $195.6m and $418.4m, implying margins of approximately 1.2%, 13.8% and 19.2%. Merely meeting these numbers would be a major transformation; it is also already visible in published expectations.

At $12.65, the provider's 2028 EPS of $1.200 implies 10.5× forward earnings. At the 10x price of $126.50, it implies 105.4× that same estimate. A 30× exit multiple instead requires $4.217 EPS. Because the normalized estimate feed does not specify accounting basis or each analyst's exchange/dilution assumptions, this is an expectations illustration rather than a reliable fully diluted earnings valuation.

Fiscal.ai mean estimates retrieved October 3: revenue/EBITDA contributor counts 7, 7 and 6; EPS counts 4, 3 and 2. Individual estimate timestamps and EPS normalization were not disclosed. Company guidance and consensus are kept separate. No past EPS comparison bridges incompatible pre-IPO units and post-IPO public shares.

Ten-quarter revenue window and missing history

Four of the six reported-quarter positions were sourced from the filing; Q3 and Q4 2025 are not sourced as comparable standalone quarters and remain blank. The ten-quarter earnings chart is unavailable on a consistent economic-share/EPS basis. No formal Q3 revenue or EPS company guide was found; four future quarters use analyst estimates. Quarterly estimates and annual means are independently aggregated and do not perfectly reconcile. The provider's historical Q2 2026 estimate/actual field is inconsistent with the $39.878m filed revenue and is not used as the historical actual.

The 10x case needs much more than today's backlog

A tenbagger from $12.65 means $126.50. At an illustrative 300m future economic shares, that requires $37.95bn equity value. At 30× earnings, the company needs $1.265bn annual common profit. At a 15% net margin, that is $8.43bn revenue—18.7× the 2026 guidance midpoint, or approximately 165.6% annual revenue growth over three years to 2029.

300m is an analyst allowance for Class M sensitivity, current restricted awards and future issuance; not a disclosed forecast. The June-reference 272.8m count is not invariant with stock price. Earnings must include taxes, financing and all common economic interests; cash distributable to public investors can also be affected by tax-receivable payments.

Illustrative 2029 scenarios

All cases use 300m future economic shares and $150m residual net cash after delivery obligations, leases, financing and capital investment. The $150m is an assumption, not current cash. The price calculation is (EBITDA × exit multiple + net cash) ÷ shares. No separate noncontrolling-interest claim is added because all economic ownership is already included in the denominator. Future tax-receivable obligations remain unmodeled; the scenarios can overstate common cash value.

Successful scaling can produce meaningful upside without a 10x outcome. $3bn revenue and $600m EBITDA at 18× imply $36.50, or 2.89×. The stretch case requires $7.5bn revenue and $1.65bn EBITDA at 25×, yielding $138, or 10.91×. That is 155.4% annual revenue growth from 2026 to 2029 and a margin above the 2028 consensus. The failure case implies $2.37, about 81% downside. Outcomes can be worse, including zero; these are not probability-weighted targets.

Valuation sensitivity: at the 2028 consensus EBITDA of $418.4m, $150m net cash and 300m shares, 15× / 20× / 25× multiples imply approximately $21.42 / $28.40 / $35.37. Keeping all $626.6m as future net cash would add about $1.59 per share. Reducing future shares to 273m at the same operating outcome increases per-share value by roughly 9.9%. Execution and margins dominate the 10x math more than that share-count sensitivity.

Variant view. The interesting upside possibility is that additional assembly capacity and customer phases drive growth beyond today's orders while pricing raises retained profit. The known backlog by itself is not a variant thesis. Conversely, quote-site capitalization errors can create a false impression of extreme cheapness.

Price context: the IPO decline is real, the causality is incomplete

The stock is 41.2% below its $21.50 IPO price, but 40.7% above the July 29 closing low of $8.99 in this retrieved series. It rallied 22.8% on August 12 after the prior evening's earnings release, then reached $17.40 on August 17 before falling 27.3% to the latest close.

Those moves show volatile expectations for the ramp. A return to the IPO price would be only 1.70×. This review did not establish a company-specific explanation for every down day or verify exact selling flows. The official Fiscal.ai series retrieved here starts July 6, so the full June trading history and intraday extremes are not claimed. One weekend observation was excluded from the chart, while the raw data remains retained.

Market source: Fiscal.ai verified company page; price data as of October 2. IPO terms corroborated in the prospectus and call.

Growth-investment scorecard: what earns an upgrade

Next catalysts: the next Q3 and Q4 reports test production, delivery acceptance and EBITDA inflection; exact release dates were not confirmed. Hyperion's year-end capacity milestone and mid-2027 Meta-linked operations are management targets. New customer orders, additional assembly sites, December lock-up effects and Class B/M exchanges deserve monitoring.

Specific downside risks: engine/component supply, manufacturing quality and warranty costs, customer project financing, air permits and gas connections, dependence on a few major contracts, back-loaded deliveries, deterioration in deposit terms, stock dilution and tax-receivable cash leakage. Even if demand remains strong, weak conversion from backlog to profit per share can disappoint investors.

Compared with BW: I prefer ERock's current commercial evidence and delivery horizon. ERock has a commissioned installed fleet, named AI customers and advance payments; BW's key first project targets operation in 2030 and lost its original guarantee. That comparative evidence does not establish that EROC is cheaper or offers a higher probability of 10x at its larger effective equity value.

Sources, assumptions and remaining evidence

Source currency is USD; no exchange-rate conversion was needed. The August 12 earnings materials package was inspected: transcript, presentation slides and report PDF (the 10-Q). The primary release was checked separately. The share-dilution table and Hyperion slide were visually inspected. Provider fundamentals and forward estimates were retrieved October 3; filed historical facts control when sources conflict.

• Q2 2026 10-Q: capitalization, liquidity, EPS period, customers, obligations and cash-flow bridge.
• August 11 results release: orders, 2026 guidance and financial reconciliation.
• Issuer presentation register: August 12 presentation, project milestones and assembly claims.
• Final IPO prospectus: ownership, proceeds, lock-up and conditional tax-receivable illustrations.
• LLC agreement: Class B exchange and Class M conversion formula.
• Fiscal.ai company page: identity, current price, shares and provider estimates; raw retrievals retained for audit.

Unresolved conflicts. Public Class A market capitalization and full economic capitalization are different concepts. The normalized overview's small capitalization, enterprise-value fields and historical EPS are not accepted as whole-company valuation. Class M conversion is value-dependent; exact grant thresholds remain unavailable. A future model must reconcile taxable earnings, noncontrolling attribution and all economic shares. The release's IPO wording compresses primary and owner-liquidity flows: roughly $600m gross total, $554m net, approximately $184.7m owner repurchases and $369.3m retained proceeds before subsequent debt repayment—not $600m of fresh operating liquidity.

Evidence gaps. Independently verified lifetime unit economics, complete supplier commitments, detailed project-level margins, a full Class M threshold schedule, future tax-receivable cash forecasts and a comparable six-quarter history are not complete. Current short interest, borrow conditions, exact free float, ETF/index ownership and factor exposures were not verified; no portfolio sizing recommendation is made.

Underwriting status. Preliminary watchlist underwrite. The commercial proposition merits deeper work; valuation and after-tax cash per fully exchanged share require a detailed model before ownership conclusions. The next useful evidence is actual H2 deliveries and profit, not another broad AI market-size forecast.