A real power opportunity.
A much harder 10x thesis.

The decision in one minute

I am downgrading the earlier screen's confidence in BW. Deeper primary-source work found that Applied Digital's guarantee ended May 29 and that commercial operation of the first project is planned for 2030. BW still has a signed contract and recognized revenue; neither finding establishes cancellation. They materially weaken the original two-to-three-year case.

Market fit. Power availability is a credible AI bottleneck. BW offers a potential delivery advantage by starting with boilers and steam turbines. Investors need evidence that this advantage survives financing, permitting, project economics and competing supply.

Research posture. Watch for proof before promoting BW to a high-conviction tenbagger. The scenarios below are assumption tests, not probability-weighted forecasts or price targets.

Evidence confidence: high for filed contract terms, reported financials and closing prices; moderate for management's delivery claims; low for project financing completeness, independently matched technology economics and precise causes of daily selling.

How the decline actually unfolded

The approximately 75% claim is correct from May's closing high. From May 1 ($14.07), the decline is 59.5%. The data includes closes rather than intraday highs; no 2026 split was found. The equity capitalization fell approximately 71.5%, slightly less than the price because the share count increased.

Daily data cannot identify sellers or isolate event effects. Some large down days have no verified company-specific announcement in the reviewed chronology. Hover/tap markers, drag the date slider and toggle the legend. Tables retain the key observations if scripts are blocked.

May 15 → October 2: BW −73.9%, APLD −40.4%, GEV −5.8%, NVDA +3.8%. These are company comparisons, not a broad-market benchmark, matched-risk portfolio or statistical event study. BW's severity goes well beyond a uniform AI-sector selloff.

Why the stock fell: evidence versus interpretation

What the company does—and the genuine bull case

BW supplies boilers, related power equipment and aftermarket services. Its FastPower growth products aim to serve data centers that need electricity sooner than conventional gas-turbine projects can deliver it.

1. Start with equipment BW can supply: burn natural gas in a boiler, create steam and run a Siemens steam turbine to make electricity.
2. Add a gas turbine later: the boiler can become a fired heat-recovery steam generator, creating a more flexible combined configuration.
3. Replicate the design: use one successful plant to win more projects, spread engineering costs and generate equipment, construction and later service income.
4. Improve earnings per share: grow enough retained profit and cash to outweigh warrants, stock compensation and additional financing.

The first proof point: four 300MW systems, or 1.2GW, under the Base Electron agreement. BW recognized $131.7m of related revenue in H1, including $100.7m in Q2. This establishes activity; it does not establish final financing, full payment or attractive lifetime project returns.

Long-term growth opportunity: management describes a $14bn opportunity pipeline and reserved 20 × 50MW steam turbines in August. The pipeline is uncontracted, and the 1GW turbine reservation is a supply commitment rather than a customer order. Treat both as commercial ambition until funded contracts and cash terms appear.

Competitive comparison. GE Vernova and Siemens Energy offer conventional gas-turbine solutions; EPC contractors and alternative on-site power compete for the same customers. BW may have a delivery advantage during a turbine shortage. Management claims favorable economics, but no matched independent operating benchmark reviewed here proves lower lifetime costs or superior efficiency. Technology can be commercially useful without becoming a durable monopoly.

August investor presentation · August 11 turbine agreement

Fundamentals snapshot: scale is improving, quality still needs proof

The issuer's Q2 adjusted-net-income reconciliation is approximately $14.3m GAAP net income − $0.3m stock-appreciation adjustment − $4.9m customer-warrant adjustment = $9.1m. It is not simply GAAP earnings less the gross fair-value gain. H1 GAAP net loss was $62.7m, while issuer adjusted net income was $14.7m before preferred dividends; large noncash warrant movements explain part of the difference.

Management raised the 2026 adjusted EBITDA target to $80–105m. The release distinguishes these targets from GAAP guidance. Backlog was $2.569bn at June 30: approximately $403.8m expected in the remainder of 2026, $674.5m in 2027 and $1.491bn after 2027.

Sources: BW Q2 10-Q and Q2 earnings report PDF in the inspected earnings materials package. Continuing-operation reclassifications mean historical provider figures should not be treated as a uniform unrevised series.

Ten-quarter revenue window: six past quarters plus four future estimates

Q4 2025 is deliberately blank: standardized provider history says $138.8m, while BW's March presentation says $161.0m. Reclassifications are not fully reconciled. The issuer's August TTM figure, $834.3m, is the valuation growth base. A comparable ten-quarter earnings chart is unavailable: historical GAAP/common earnings and provider consensus EPS do not have a verified common accounting basis. No company Q3 revenue/EPS guidance was substituted for an analyst estimate.

Growth estimates consensus: useful progress, still far from 10x

Revenue rises from $1.049bn in 2026 to $1.614bn in 2028: 24.0% annual growth over two years (+22.2% in 2027 and +25.9% in 2028). At 25× the provider's 2028 EPS of $0.643, price would be approximately $16.08, or 2.82× today's close. A $57 stock at 25× earnings needs $2.28 EPS—3.54× that 2028 estimate.

Fiscal.ai mean estimates retrieved October 3; annual revenue/EBITDA counts 4, 4 and 3; EPS counts 3, 4 and 3. Analyst vintages and GAAP versus adjusted EPS basis were not disclosed by this normalized feed. These figures are a directional expectations anchor, not an audited comparable EPS model. The 2026 EBITDA mean of $84.7m is below the issuer target midpoint, illustrating that provider consensus and management targets differ.

What a $57 stock actually requires

Using approximately 149m current shares and 180m future shares, $57 requires roughly $10.26bn of common equity value. At 25× earnings that means $410.4m annual common net profit. At a 10% common net margin it requires approximately $4.10bn annual revenue, versus the issuer's $834.3m TTM base. That is 70.1% annual revenue growth for three years, or 121.8% for two years.

Change the hurdle

Future shares include exercised warrants, stock compensation and new issuance. This P/E approach uses earnings after interest, tax and preferred dividends; EBITDA is not an interchangeable input.

Illustrative 2029 enterprise-value scenarios

The stretch case: $4.5bn revenue at a 12% EBITDA margin produces $540m EBITDA. A premium 20× multiple supports about $10.8bn enterprise value. After modeled debt, cash and preferred claims, 180m shares imply $59.49, or 10.44× today's price. It requires 75.4% revenue CAGR from $834.3m over three years, and 85.4% EBITDA CAGR from the $84.7m 2026 consensus base to 2029. These are substantially beyond the currently published estimates.

The repeatable-business case: $1.8bn revenue, a 12% EBITDA margin and 12× multiple imply $14.41. The failure case implies $1.23. Actual recovery could be worse, including zero; these scenarios do not bound outcomes.

Growth-investment scorecard and thesis gates

Next catalysts and monitoring: Q3 cash collections and operating cash flow; county and state permit decisions; evidence of financing for Missouri Bend; a second full notice to proceed backed by real funding; contract margin disclosure; warrant exercise, preferred costs and share count. Management's expected next award is not an executed contract.

Falsifiers: persistent uncollected balances, escalating concessions or cancellation, lack of project financing, repeated permit delays, inadequate retained margins, or share issuance that absorbs operating growth. A cheap share price does not repair those failures.

Source register, chronology and method

Source currency is USD; no exchange-rate conversion was needed. Prices and USD capitalizations are matched official Fiscal.ai observations retrieved October 3. Financials were checked against BW's August 10 earnings release/report PDF, 10-Q, transcript and August investor presentation slides. The transcript package returned no slides; the separate SEC-filed August presentation was located and inspected. The original developer brochure and 10-Q PDF were visually checked for schedule and contract/cash details.

Management claims, filed facts, third-party reporting and analyst assumptions are explicitly separated. The August 2026 presentation is used for TTM revenue because provider quarterly restatement data is inconsistent. Publication dates and event dates are distinct. The report does not infer exact trading causality from a close-to-close move, treat a sales pipeline as backlog, or characterize unproven short-seller allegations as fact.

Updated broader AI buildout screen