Energy
PrivateA real software-margin turnaround stapled to a ~$332M net-debt stack the ~$10–15M-EBITDA business cannot service from cash — the equity is a ~$67M option on refinancing, not on the operations.
Research
The verdict
A real software-margin turnaround stapled to a ~$332M net-debt stack the ~$10–15M-EBITDA business cannot service from cash — the equity is a ~$67M option on refinancing, not on the operations.
Primary sources
Source documents — open to read in full
What it is. Stem, Inc. (Delaware; HQ 1400 Post Oak Blvd, Houston TX; NYSE: STEM) describes itself as "a global leader reimagining technology to support the energy transition". In plain terms: it sells software and services that monitor, control and optimize solar, storage and hybrid energy assets for the people who own and operate them. The flagship is the PowerTrack™ suite — PowerTrack Software (solar monitoring/analytics), Energy Management System (EMS), SCADA, Power Plant Controller, Logger, and PowerTrack Optimizer (the rebranded Athena® AI-optimization engine, renamed Sept 2024).
How it makes money — two revenue lines:
Customers. Project developers, asset owners, EPC (engineering-procurement-construction) firms, and distributors — i.e. B2B/infrastructure, not consumers. The research-layer customers.csv is empty (no per-name concentration on disk); the 10-K names a marquee win: a 484 MW PowerTrack deployment for Neovolt, a large Hungarian renewable owner (early 2025), and the Q1-26 call adds an Everyray German project (100+ MWh) and a Hungarian solar-hybridization deal (50+ MWh).
Contract structure. Software fees are "recurring fixed monthly payments throughout the term," often with an upfront/installation fee; hardware is recognized at delivery, with milestone pre-payments held as deferred revenue. The recurring base is large relative to the company: deferred revenue of $128.9M ($43.6M current + $85.3M noncurrent) and ARR of $61.2M at Q1-26. Notably, some historical contracts carried a hardware-value guarantee (Stem agreed the purchased hardware value would not decline) — the source of the $104M "impairment of parent company guarantees" in 2024.
Scale. 36–37.5 GW of solar AUM, ~1.7 GWh of operating storage AUM, 55 countries, ~90 patents, nine registered trademarks, 423 employees (post-RIF).
Stem sits midstream as a software/services intermediary — it is not a battery or hardware manufacturer, which is the single most important fact about its supply chain (and the reason it survived 2025; see Lens 13). Mapping upstream → Stem → end customer:
Chokepoints / single-source dependencies: As software, Stem has fewer hard chokepoints than a hardware peer — but it is exposed to (1) third-party cloud providers (concentration not disclosed; supply-chain.md for energy is missing ); (2) OEM battery availability and tariffs for the resale line (mitigated by being asset-light and hardware-agnostic); and (3) OBBB/FEOC "material assistance" sourcing-verification rules that push compliance complexity onto the whole clean-energy chain (Lens 10). Named partner in the chain: Nuvation Energy (April 2026 partnership to market a "cell-to-cloud BESS and hybrid control stack" in North America).
Provenance note: the energy commercial-layer files (supply-chain.md, bottlenecks.md, positioning.md) are all absent on disk, so this lens is built from the 10-K business section + the call, not a pre-compiled supply map. Names present; chokepoint granularity limited by source.
The honest verdict: a narrow, data-and-switching-cost moat that is real but not wide. Stem's differentiation is being the hardware-agnostic, AI-software layer that sits across a customer's whole heterogeneous fleet — "standardization of energy portfolios on one hardware agnostic application".
Moat sources, graded:
Bargaining power. Over suppliers: moderate-high (cloud is commoditized; OEM batteries are buyer's-market post-2024 glut). Over customers: weak — the 23% YoY bookings decline (Q1-26) and the need to "increase pricing modestly" say Stem is a price-taker competing for budget against in-house tools and larger platforms. The distressed balance sheet itself weakens bargaining power: bankability matters to infrastructure buyers, and a company with negative equity and 12% secured debt is a harder counterparty to underwrite for a 10-year asset.
Stem reports two product lines, no separate operating-segment P&L (single reportable segment); the research-layer segments.csv is empty, so the breakout below is lifted directly from the income statement and MD&A.
By product line:
| Line | FY2023 | FY2024 | FY2025 | FY25 YoY | Trend |
|---|---|---|---|---|---|
| Services & other revenue | $62.5M | $67.8M | $87.7M | +29% | Accelerating — the engine |
| Hardware revenue | $399.0M | $76.8M | $68.6M | −11% | Decelerating by design |
| Total revenue | $461.5M | $144.6M | $156.3M | +8% | Inflected up in FY25 |
The story in one line: hardware fell 83% across 2023→2024 (the battery-resale exit) while services compounded ~18%/yr — and in FY25 total revenue actually grew 8% for the first time post-pivot, because services growth finally outran hardware decline. Q1-26 broke that briefly (total −11% YoY) because there were zero battery resales in the quarter (management says they are "weighted to 2H 2026") while services kept growing +5%.
By geography: Not separately tabulated in the statements; international is ~5% of Q1-26 revenue — i.e. despite "55 countries" of AUM, revenue is overwhelmingly US. The European push (Neovolt/Hungary, Everyray/Germany, raicoon's EU footprint) is the stated growth vector but is small today.
Margin by line (derived): FY25 services gross margin ≈ ($87.7M − $52.7M)/$87.7M = 40%; hardware gross margin ≈ ($68.6M − $43.6M)/$68.6M = 36% — note hardware margin is positive in FY25 (vs deeply negative in 2024 when cost of hardware $103M > revenue $77M, i.e. selling inventory below cost to clear it). The mix shift to services is the margin lever, but FY25 hardware was also cleaned up.
Headline: Revenue $29.0M, −11% YoY, missing consensus of ~$34.5M by ~16% — and the stock fell ~26% after-hours. But adjusted EBITDA was +$2.0M (vs −$4.6M) — the first-ever positive Q1 adjusted EBITDA and the fourth straight positive EBITDA quarter.
The print, decomposed:
Guidance (reaffirmed at Q1): FY26 revenue $140–190M (software/services/edge $130–150M + battery resales up to $40M), non-GAAP GM 40–50%, adjusted EBITDA $10–15M, operating cash flow $0–10M (positive), year-end ARR $65–70M.
FY2025 full-year context: Revenue $156.3M (+8%); GAAP gross profit $59.96M (38%, vs −$11.1M loss in 2024); operating loss ($55.7M); net income $137.8M — but entirely from a $220.0M gain on debt extinguishment (see Lens 5/10); adjusted EBITDA +$6.7M (vs −$22.8M); operating cash flow +$6.9M (vs −$36.7M). The GAAP "profit" is an accounting artifact of the debt restructuring; the operating business still lost money on a GAAP basis but reached cash-flow and EBITDA breakeven.
transcripts/ is empty on disk; this lens is web-sourced from the Q4-25 (2026-03-04) and Q1-26 (2026-05-07) calls.
Tone shift over the last ~year: from "stabilize/survive" to "execute/scale," with consistently high management confidence — but a widening gap between management optimism and market reception.
Recurring themes management now leads with:
What they stopped saying: the 2024-era language of "strategic review," "slashing guidance," and delisting risk has disappeared. The CFO's recurring defensive note — Q1 cash burn is "timing, not a change in underlying cash generation" — is the tell that the analyst pushback is on cash.
Closing tone (Narayanan, Q1-26): "We have the right strategy, the right team and the right momentum… we remain confident in achieving all our full year commitments". Confident, scripted, turnaround-CEO register. The −26% stock reaction the same day is the market's counter-vote.
The peer set is thin and ugly — clean-energy storage/software is a graveyard sector. Pure-play comparables: Fluence Energy (FLNC) (closest public name — storage system integrator + software, AES/Siemens heritage) and the captive energy arms of Tesla and Enphase (not clean comps — different scale/model). The most instructive comp is not a multiple but a fate: Powin (3rd-largest US BESS integrator) went Chapter 11 in June 2025 (Lens 13).
| Company | Ticker | Mkt cap | EV | EV/Sales | P/S | EV/EBITDA | Notes |
|---|---|---|---|---|---|---|---|
| Stem | STEM | $67.0M | $399.5M | 2.62x | 0.44x | n/m (EBITDA ~$10–15M guide → ~27–40x EV/EBITDA fwd) | Net debt ~$332M dwarfs equity |
| Fluence Energy | FLNC | ~$3.0–4.2B | n/a | ~0.99x | ~1.2x | ~37x fwd | Hardware-heavy, ~$3.4B FY rev, unprofitable |
| Tesla (Energy seg.) | TSLA | n/a — captive segment | n/a | n/a | n/a | n/a | Not a clean comp |
| Enphase | ENPH | n/a | n/a | n/a | n/a | n/a | Residential micro-inverter/storage, different model |
| Powin | private | $0 (Ch.11, Jun 2025) | — | — | — | — | Asset sale to FlexGen; cautionary comp |
The provenance-critical read: On the equity (P/S 0.44x), Stem looks cheap vs Fluence (~1.2x). On enterprise value (EV/Sales 2.62x), Stem looks EXPENSIVE vs Fluence (~1.0x) — because Stem's ~$332M net debt is ~5x its equity market cap. The comp that matters is EV/Sales, and on that basis Stem is not a cheap software company — it is a leveraged turnaround where the debt has eaten most of the enterprise value. Any thesis must be a view on the capital structure, not just the operating multiple. (I did not source FLNC's exact EV; reported EV/Sales ~0.99x is from a third-party stat page — labelled, not fabricated.)
Stem is a textbook SPAC-bust chart. The moves that mattered:
Pattern → what the market actually reacts to: (1) revenue trajectory (down = punished, regardless of margin), (2) cash/solvency (delisting, burn), and (3) capital-structure events (the debt exchange was the single biggest positive catalyst). It does not reward EBITDA/margin improvement in isolation. Translation: the next real upside catalyst is a clean refinancing/equitization of the debt OR a quarter of genuine top-line re-acceleration — not another EBITDA beat.
Archetype: a professional-manager turnaround team, freshly installed, software-credentialed — not founders. The original founders are gone; this is a hired crew brought in to execute a pivot and (critically) to be credible to lenders.
Tenure & skin in the game: CEO/CFO both <18 months tenured. Insider ownership is not meaningful via open-market buying — recent Form 4s show only RSU grants (e.g. director David Buzby, 12,168 RSUs, June 2026); the revised Oct-2025 insider-trading policy prohibits shorts/hedging/derivatives. No conviction insider-buy signal — a notable absence for a stock the team calls deeply undervalued.
Capital-allocation history (the company's, not just this team's): Value-destructive on the legacy record — the Also Energy / battery-resale roll-up generated a $547M goodwill impairment and a cumulative $1,488.7M accumulated deficit. Disciplined under the new team: SBC cut from $45.1M (2023) → $10.2M (2025); two tuck-in software acquisitions (raicoon, undisclosed price; cell-to-cloud via Nuvation partnership) rather than capital-heavy deals; and the June-2025 debt exchange that cut debt face ~$195M. The new team's allocation looks rational; the inheritance is a wreck.
Red flags: (1) the hardware-value guarantees that blew up in 2024 ($104M) — a legacy underwriting sin; (2) CEO/CFO churn (this is the 2nd CEO in ~2 years); (3) repeated metric re-definitions (Bookings, Backlog, CARR all "redefined" beginning Q1-25) — common in turnarounds but it resets comparability and can flatter trends. No related-party or comp-abuse flags surfaced.
Acting as a forensic analyst. Auditor: Deloitte & Touche LLP (since 2018), UNQUALIFIED opinion on both the FY25 financials and internal control over financial reporting, dated 2026-03-04.
Income statement.
Balance sheet.
Cash-flow vs earnings divergence. FY25 OCF (+$6.9M) is below the operating loss would suggest only because of the working-capital release (+$30.4M, much from AR collection +$17.7M and project-asset wind-down +$14.8M). The working-capital tailwind is largely one-time — once the legacy balance sheet is fully harvested, run-rate OCF is thinner than the FY25 headline, which is exactly what Q1-26's ($8.3M) shows. This is the most important forensic caveat: do not extrapolate FY25 positive OCF.
Critical Audit Matter: fair-value of the 2030 Senior Secured Notes (Level 3, Black-Derman-Toy lattice, "synthetic credit rating" + yield + volatility inputs) — flagged by Deloitte as the area of greatest management judgment. Not an irregularity; a complexity disclosure.
Going concern. Both filings are prepared "assuming the Company will continue as a going concern" and disclose negative working capital and the $1.49B accumulated deficit — but neither contains "substantial doubt" language, and Deloitte's opinion is unqualified. So this is not an active going-concern qualification — management resolved substantial doubt via the June-2025 recap — but it is a live watch item given the cash trajectory.
Regulatory findings (required sub-section).
Built bottom-up from FY25 actuals + management's FY26 guide + the operating trajectory. Pre-revenue/EPS lines are GAAP-loss-making for years due to ~$45M/yr intangible amortization + ~$24M/yr interest; the more decision-relevant lines are revenue, adjusted EBITDA, and cash runway, so I project those and treat GAAP EPS as deeply negative throughout.
Anchors: FY25 revenue $156.3M, adj EBITDA +$6.7M, OCF +$6.9M, cash $36.6M (Q1-26), net debt ~$332M, interest ~$24M/yr, ~9.0M shares.
| Scenario | FY26 Rev | FY26 Adj EBITDA | FY27 Rev | FY28 Rev | GAAP EPS (FY26) | Cash-runway read |
|---|---|---|---|---|---|---|
| Base | ~$165M (guide midpoint) | ~$12M (guide mid) | ~$185M | ~$210M | ~($6) to ($8) | Survives FY26 on ~$37M cash + ~$0–10M OCF + 12% PIK toggle (can defer cash interest); refinancing needed before 2028–30 maturities |
| Bull | ~$190M (top of guide; resales + utility ramp) | ~$15M | ~$220M | ~$260M | ~($5) | OCF positive, EBITDA covers cash interest if PIK deferred; debt refinanced/partly equitized at higher equity price — re-rating |
| Bear | ~$135M (resales slip, bookings keep falling) | ~$6–8M | ~$130M | ~$140M | ~($9)+ | Cash to ~$15–20M by YE26; forced dilutive raise or distressed exchange #2; equity impaired |
Reasoning lines (all, inputs labeled):
Per --watchlist rules, I am NOT logging a forecast.ts Brier forecast (skip the create step in the sweep). If promoted to a thesis, the scoreable line would be: "STEM FY26 adjusted EBITDA ≥ $10M" (p≈0.60) and "STEM FY26 revenue ≥ $150M" (p≈0.55).
Bull case. Stem is the survivor of a sector apocalypse, re-based as a real software business and priced like roadkill. The pivot is no longer a promise — it is in the numbers: GAAP gross margin (8)% → 38%, non-GAAP 52% (record), four straight positive-EBITDA quarters, cash opex −30% "permanently," ARR $61M growing, 37.5 GW of solar AUM, and a utility-scale pipeline management calls "the strongest we've ever seen." Crucially, Stem's asset-light, hardware-agnostic model walked through the exact tariff/ITC storm that bankrupted Powin and battered Fluence — it doesn't carry the China-battery inventory risk. At a $67M equity market cap on ~$61M of ARR and ~$165M of guided revenue (0.44x P/S), the equity is a cheap call option: if management hits the FY26 guide (positive OCF, $65–70M ARR) and then refinances or equitizes the debt at a higher price, the equity re-rates multiples. The June-2025 debt exchange proves the team can do balance-sheet surgery. Contrarian kicker: the market is so fixated on the revenue optics of a deliberate hardware exit that it is ignoring a software business inflecting to cash-generative.
Bear case (the 2–3 things that permanently impair).
Pre-mortem (18 months out, thesis broke): Battery resales slipped again, utility bookings converted slower than guided, FY26 revenue printed ~$135M, cash hit ~$15M by year-end, and Stem did an emergency raise / debt-for-equity swap that diluted holders 60–80%. The 52% gross margin was always real — and always irrelevant — because the enterprise value was the debt, and the equity was the residual.
Are multiples too high? On equity, no (0.44x P/S). On enterprise value, the question is wrong — EV/Sales 2.62x is not "expensive software," it is "the debt is most of the enterprise." The honest framing: the equity is fairly priced as a distressed option — cheap if the recap is benign, near-zero if it isn't.
Contrarian view (what the market refuses to see): that this is no longer a clean-energy hardware story at all — it is a distressed-credit / capital-structure story wearing an AI-software costume. The operating turnaround is genuine and largely done; the unresolved variable is entirely on the right side of the balance sheet. Anyone trading the EBITDA beat is trading the wrong instrument.
I am short STEM. Here is how the bull thesis dies.
Short's verdict: the operating story is true and the equity is still a trap, because the equity isn't the operating story — it's the leftover after a $369M debt stack. Borrow cost and squeeze risk (52-week high $32) are the short's real enemy, not the fundamentals.
A regulated-utility levered call on the Georgia data-center build-out — the cleanest large-cap way to own AI power demand, but priced as if the affordability politics and equity dilution won't bite; own the growth, respect the ~24x multiple.
A 90%-regulated New Jersey wires utility wearing a merchant-nuclear data-center costume — you are paid ~17x for a 6–8% regulated compounder, and the AI-power optionality is real but unpriced *and* unproven; own it as a rate-base bond with a free nuclear call option, not as the next Constellation.
A regulated Florida-utility crown jewel (FPL) bolted to the world's largest renewables developer (NEER), trading at a 22x premium that prices the AI-power supercycle as a sure thing while the OBBBA tax-credit cliff and a $95B debt stack sit unpriced — own the moat, but the multiple already pays for the catalyst.