Firefly is priced as a rocket company. Last quarter rockets were 8% of revenue. The other 92% is a lunar franchise holding the only clean commercial Moon landing anyone has managed, and a defence-software business pointed at the fastest-growing line in the Pentagon budget — and the market can value neither separately, because the company reports a single segment. The shares are 66% below their May high and almost none of that decline is the operating business. Our band is $11–29 against $21.12.
The marketed order book is 2.7 times the enforceable one, and only the enforceable one has fallen. Company-defined backlog was $1.5bn at 30 June, a record and up about 15% sequentially. Remaining performance obligations under GAAP, measured at the identical date, were $563.7m — down for a fourth consecutive quarter from a $723.1m peak. No reconciliation between the two is published. The market appears to have started pricing the smaller number while the sell side still quotes the larger.
The decline is mechanical, and the business improved right through it. A secondary priced at $48.00 two days after the all-time high, an 11.1m-share resale registration, a lock-up expiring 12 August, and roughly 61.8m registered shares against a $103m-a-day tape. Over the same window the EBITDA loss narrowed twice, guidance held at 95% booked, customer concentration more than halved as the largest account fell from 83.8% to 34.4%, and Lockheed Martin extended its launch agreement to 2031 after losing a satellite on Alpha. Every element of the fall is identifiable and none of it is an operating result.
The price already pays for the base case, which is why this is not a mispricing. At $21.12 the shares sit within 1% of our $21.00 central estimate and inside a $11–29 band. The discounted cash flow says far less than that; the peer tape says rather more; and the reason we publish a band 2.6 times wide is that those two disagree by a factor of nearly two and neither can be dismissed.
Start with the one number this company publishes that nobody can argue with. At 30 June 2026 Firefly disclosed remaining performance obligations — the GAAP measure of revenue it is contractually entitled to and has not yet recognised — of $563.7m. At the same date it marketed a backlog of $1.5bn. The difference is $936m of indefinite-delivery ceiling, unexercised options and multi-launch capacity that obliges no customer to order anything.
Neither figure is improper and the gap is normal in government contracting. What is not normal is the direction. The marketed number set a record while the enforceable number fell for a fourth consecutive quarter, from a $723.1m peak in the third quarter of 2025. The two halves of the order book are moving in opposite directions, no reconciliation between them is published, and the definition of “backlog” is not disclosed anywhere.
That matters most for 2027. Roughly $237m of the RPO is scheduled to be recognised within twelve months, on a 42 / 21 / 37 split across the first year, the second and everything after. This year’s guidance, which the company states is 95% booked, consumes about $215m of it. Sell-side FY2027 revenue of around $686m therefore rests on somewhere between $81m and $140m of enforceable support. The street is underwriting roughly $550m of revenue that is not under contract. That is not unreasonable given the setup in Section 09, but it is an assumption rather than a backlog, and it is the first thing to break if appropriations slip.
What the enforceable half contains matters as much as its size, and it is good. Six contracted lunar missions with five in parallel build. A multi-launch agreement Lockheed Martin lengthened by two years in August, to 2031, having lost a satellite on Alpha Flight 6 and therefore knowing more about the vehicle’s reliability than any other customer. An accredited missile-warning ground-processing incumbency carrying a superior performance rating from the Department of Defense, aimed at a budget line requested at $71.3bn for FY2027 against roughly $32bn this year. Roughly 90% of contract value is collected before launch, so the working capital is customer-funded. A smaller enforceable book of that composition is worth more than a larger one of options nobody exercises, and the honest reading of this section is that the marketed figure flatters something that did not need flattering.
In the second quarter of 2026, launch produced $9.4m of $117.7m — 8.0% of revenue. The remaining 92% is Spacecraft Solutions, which contains two economically unrelated businesses: lunar landers and orbital vehicles on one side, and the national-security software acquired with SciTec on the other. Firefly reports as a single operating segment. No segment profitability is disclosed anywhere, and neither is SciTec’s standalone revenue.
Firefly reiterated a dollar revenue guide in the same quarter it cut the 2026 Alpha launch target from four flights to three. Arithmetically that shifts the mix further away from launch. The company then beat consensus revenue by 31% in a quarter with no Alpha flight at all. Whatever this business now is, its near-term revenue is not a function of how often the rocket flies — and investors who value it on launch cadence are pricing the 8%.
| Measure | Value | Note |
|---|---|---|
| Reference price, 31 Aug 2026 close | $21.12 | The last completed session. Every multiple on this page scales with it |
| Market capitalisation | ~$3.54bn | On 167.4m shares outstanding |
| Enterprise value | ~$2.93bn | Less $608.3m of net cash — $635.3m of cash and short-term investments against $27.0m of notes payable |
| EV / FY2026E revenue | 6.7x | On the $435m guidance midpoint. 10.2x on trailing revenue of $287.0m |
| EV / FY2027E revenue | 4.4x | On our $673m base case. 4.3x on the higher of the two contested consensus readings |
| EV / enforceable contracted value | 5.2x | Against 1.95x on the marketed $1.5bn. The single most important line in this table |
| Gross margin | 20.3% | Q2 2026, down from 21.6% in Q1 and 27.7% in Q4 2025 |
| Trailing free cash flow | −$327m | −$106.3m in the second quarter alone, worse than −$78.9m in the first |
| Net cash per share | $3.32 | 17.2% of the market capitalisation. No convertibles, no maturity wall |
| 52-week range | $16–$62 | 66% below the high, reached in May 2026 |
| Realised volatility, 30-day / 1-year | ~81% / ~110% | 47 sessions in the last year moved 8% or more. Position size matters more than entry price |
Derived inputs. Enterprise value, net cash and every multiple in this table are ours, computed from the Form 10-Q for the quarter ended 30 June 2026 and the 31 August closing price. The EV / enforceable contracted value figure divides enterprise value by the disclosed $563.7m of remaining performance obligations and is a ratio we construct, not a company measure. Trailing-twelve-month revenue and free cash flow aggregate four reported quarters, of which only the last two include SciTec.
Discounted cash flow $13.75 at 35% · peer multiple $25.39 at 45% · revalued at sector rates $23.32 at 20%
The bridge above is the cash-flow leg only. Implied terminal EBIT margin 13.4% · FY2026–35 revenue CAGR 23.4% · terminal value is 135% of enterprise value, because the explicit decade subtracts.
The headline number is the published conclusion, not one leg of it. This panel computes all three methods the report uses and weights them 35 / 45 / 20, which is why the presets land on the band in Section 03 rather than somewhere near it. The cash-flow leg is an unlevered discounted cash flow on the enterprise with $608.3m of net cash added back: nine explicit years from FY2027 to FY2035, with terminal value taken as an exit multiple on FY2035 EBITDA rather than a perpetuity, because a perpetuity describes a terminal-year business still growing near 10% badly. Gross margin ramps linearly from the 20.3% reported in the second quarter to whatever you set. Operating expense ramps linearly from the $480m annualised run-rate of the first half to whatever you set. Cash tax is zero through FY2031 and 21% after, shielded by roughly $2.0–2.3bn of accumulated losses.
The three legs disagree by a lot, and that is the honest state of this name rather than a defect. On base settings the cash flows say $13.75, the peer tape says $25.39, and the same forecast revalued at the rates and terminal margin the sector actually clears at says $23.32. We weight the cash flows least because they are the method that cannot reproduce the market price of a single comparable in the sector, including the profitable ones — when a method disagrees with every observation in a cohort, the mis-calibration is the more likely explanation. The weights are fixed and the reasoning is in Section 03; everything else on this panel moves.
Move the revenue sliders first, then operating expense. Because the peer leg carries 45%, FY2027 revenue is now the heaviest single input — $100m on it is $1.81 a share, and $500m on the FY2035 endpoint is $2.68. Operating expense is the heaviest input inside the cash flows: the base case holds FY2035 spend at $480m, the same dollar figure Firefly is spending this year against revenue 6.6 times larger, and $50m on that line is $0.65 blended, or $1.85 on the cash-flow leg alone. By comparison a hundred basis points on the discount rate is $0.37. On this name the argument is about revenue and cost discipline, not about the discount rate, and the model is built so that shows.
Reading the inputs. Terminal EBIT margin is an output here, not an input. It is whatever the terminal gross margin and the FY2035 operating-cost endpoint leave behind, and it is printed in the readout so you can see what your settings imply. We built it that way deliberately: a reader can form a view on “does gross margin reach 30%” and on “does $2.9bn of revenue really carry only $480m of research and overhead”, and those two questions are answerable from the filings. Nobody can form a view directly on a terminal EBIT margin, and asserting one hides the assumption that produces it. The base settings imply 13.4%, which sits at the top of the range large defence primes actually earn.
The number above is not a fact about Firefly. It is what our assumptions produce, and on this name the most consequential assumption is the operating-expense endpoint, because research and development grew 48% year on year in the most recent half and the company has never guided it, never reduced it, and has not yet reached the most expensive phase of its largest programme. We hold it flat in dollars for a decade. That is a real assumption doing real work, it is exposed on a slider rather than buried in a cost build, and if you think $480m is the wrong number for 2035 then move it and use what falls out.
Same model, different inputs. Click any card to load it. Each card prints the weighted value across all three methods, so the spread from bear to bull is 2.5 times. Inside the cash-flow leg on its own it is twenty times, and the distance between those two figures is precisely why that leg carries only 35% of the weight. On a company where more than all of the value sits in a terminal state nine years out, a spread of this width is the honest shape of the distribution rather than a failure of analysis.
Each card loads into the model above and prints the weighted value across all three methods, which is why they land on the band rather than near it: $11.43, $20.90 and $28.19. Weighting those 30 / 45 / 25 gives $19.88 — below the base case, which on a company with a cash floor and an open-ended upside is the reverse of what you would expect. The reason is that the cash-flow leg of the bear case is close to nothing, and a business that never covers its cost base is worth little more than the cash on its balance sheet. That is a real possibility here and we do not want to smooth it away, which is why the floor of the band is the bear card rather than an average.
| Assumption | Bear | Base | Bull |
|---|---|---|---|
| FY2027 revenue | $545m | $673m | $760m |
| FY2035 revenue | $1.6bn | $2.9bn | $3.8bn |
| FY2026–35 revenue CAGR | 15.5% | 23.4% | 27.1% |
| Terminal gross margin | 24% | 30% | 36% |
| FY2035 operating expense | $270m | $480m | $685m |
| Implied terminal EBIT margin | 7.1% | 13.4% | 18.0% |
| Capex, % of revenue | 8% | 7% | 6% |
| Cash tax | Zero through FY2031, then 21% shielded by ~$2.0–2.3bn of losses | ||
| Discount rate | 14.0% | 12.5% | 11.5% |
| Exit multiple on FY2035 EBITDA | 9.0x | 12.5x | 13.0x |
| Diluted shares | 200m | 183m | 183m |
| Enterprise value, cash-flow leg | −$338m | $1,909m | $4,332m |
| Cash-flow leg, per share | $1.35 | $13.75 | $27.00 |
| Peer multiple leg, per share | $19.39 | $25.39 | $28.24 |
| Sector-rate revaluation, per share | $11.16 | $23.32 | $30.18 |
| Weighted value per share | $11.43 | $20.90 | $28.19 |
The base growth rate is deliberately below what the company is demonstrating. The reported +659% year-on-year is acquisition arithmetic and not the number. On Firefly’s own pro-forma disclosure, first-half 2025 revenue including SciTec was $159.3m against $198.6m actual in the first half of 2026; extending the implied SciTec run-rate across the ten pre-consolidation months of 2025 gives pro-forma FY2025 revenue of roughly $306m, against which the $435m guidance midpoint is +42%. Our base decays that to 23.4% over ten years.
FY2027 at $673m sits below the higher of two contested consensus readings. Reputable sources carry $584m and $686m for the same year — a 17% spread, which is itself worth knowing. We sit between them and nearer the top, because this year is 95% booked and the programme setup in Section 09 is genuinely favourable, but we do not take the higher figure, because roughly $550m of it is not under contract.
The operating-expense endpoint is where the argument actually is. First-half 2026 gross profit of $41.3m funded 17.8% of a $232.2m operating cost base. At a 20.3% gross margin, covering that base annualised needs roughly $1.14bn of revenue — 2.6 times this year’s guidance. So the company either grows into its costs or lifts the margin, and the base case assumes both, holding costs flat in dollars while gross margin rises ten points. If you think research and development instead keeps compounding, the model has an answer: $600m of FY2035 operating expense is $9.30 a share, $700m is $5.59, and $900m is nothing at all.
A 12.5% base discount rate is above what the market appears to apply. The risk-free rate was 4.75% on 31 August, and a size and execution premium on a thirteen-month reporting history with 110% realised volatility argues for the top of any reasonable range. We also carry a 1.5 point premium inside it for the share-supply overhang described in Section 07. Against that, the price itself implies something nearer 10.5–11.5% for this asset class, and on the evidence of the whole listed complex the market’s rate is the prevailing one and ours is the outlier. That is why the discounted cash flow carries a third of the weight below rather than the half it would otherwise get.
For a pre-profit company the discount rate is not the sensitive variable. Terminal margin and terminal revenue are. This grid holds every other base parameter fixed — 12.5% WACC, 12.5x exit, 30% terminal gross margin, 183m shares — and varies the implied terminal EBIT margin against FY2035 revenue.
| Terminal EBIT ↓ / FY2035 revenue → | $1.6bn | $2.1bn | $2.6bn | $3.1bn | $3.8bn |
|---|---|---|---|---|---|
| 5.0% bear-like | $2.49 | $3.38 | $4.20 | $4.97 | $5.98 |
| 8.0% | $4.27 | $5.71 | $7.09 | $8.42 | $10.21 |
| 11.0% prime range | $6.05 | $8.05 | $9.99 | $11.87 | $14.44 |
| 13.5% base | $7.53 | $10.00 | $12.40 | $14.74 | $17.96 |
| 15.0% | $8.42 | $11.17 | $13.84 | $16.47 | $20.08 |
| 18.0% bull-like | $10.20 | $13.50 | $16.74 | $19.92 | $24.31 |
One cell of thirty exceeds the $21.12 share price — the top-right corner, which needs an 18% terminal EBIT margin and $3.8bn of FY2035 revenue at the same time. On a conventional discounted cash flow the current price sits at or above almost the entire plausible parameter space, and we are not going to soften that.
But the grid should not be read as evidence that Firefly specifically is mispriced, because the same machinery at the same discount rate cannot reproduce the market price of a single name in the peer table below — including the profitable ones. When a method disagrees with every observation in a cohort, the honest inference is that the method is mis-calibrated for the cohort, not that the whole cohort is wrong. That is why the discounted cash flow carries 35% of the weight in the conclusion and the peer tape carries 45%, and it is the most important methodological choice on this page.
Base-case cash flows, with the exit multiple replaced by a perpetuity on FY2035 free cash flow. We publish this precisely because it disagrees: it sits roughly a third below the exit-multiple result, and the conservative reading of the model is therefore nearer $9 than $14. A perpetuity poorly describes a terminal-year business still growing near 10%, which is why the exit multiple is the primary method, but a reader is entitled to see the other answer.
| Terminal g ↓ / WACC → | 11% | 12% | 13% | 14% | 15% | 16% |
|---|---|---|---|---|---|---|
| 2.0% | $9.16 | $7.56 | $6.30 | $5.30 | $4.50 | $3.84 |
| 2.5% | $9.77 | $8.01 | $6.65 | $5.57 | $4.71 | $4.01 |
| 3.0% | $10.45 | $8.52 | $7.03 | $5.87 | $4.94 | $4.20 |
| 3.5% | $11.22 | $9.08 | $7.45 | $6.19 | $5.20 | $4.40 |
| 4.0% | $12.10 | $9.71 | $7.92 | $6.55 | $5.47 | $4.61 |
More useful than arguing about what the shares are worth is inverting the model, because the answer then depends on no view of the company at all. Solving for the FY2026–35 revenue CAGR that sets the discounted value equal to $21.12, across four capital-market assumption sets:
| Assumption set | Required CAGR | Implied FY2035 revenue | Interpretation |
|---|---|---|---|
| 14% WACC / 12.5% terminal EBIT / 9x exit | 38.9% | $8.4bn | Implausible. Twenty-nine times trailing revenue |
| 12% / 15% / 12x | 27.9% | $4.0bn | Demanding but not absurd |
| 11% / 18% / 14x approximating sell-side practice | 22.0% | $2.6bn | Below our base case of 23.4%. At the rates this sector actually clears at, the price asks for less than we already model |
| 10% / 20% / 16x | 18.0% | $1.9bn | Well below the base case — the price would be conservative |
Not anything Firefly has disclosed. The entire gap between our $13.75 and the $21.12 price is the cost of capital and the terminal multiple, and on both the market’s numbers are the ones the whole listed complex clears at. Second, the market does not discount the J-curve at all: the sell side capitalises FY2027 revenue at 5–9 times and structurally ignores the $659m of present value destroyed getting there. Third, and most interestingly, the market appears to disbelieve the headline backlog while the sell side still cites it. At 1.95x the marketed $1.5bn but 5.2x the $563.7m GAAP obligation, the price is far closer to a fair multiple on enforceable contracted value than on the marketed number. The tape corroborates it: four contract announcements between late June and early July moved the shares down.
This is the 45% of the answer. Every listed comparable at the same close, on the same basis.
| Company | Market cap | EV | TTM revenue | Gross margin | EV/sales |
|---|---|---|---|---|---|
| Firefly (FLY) | $3,536m | $2,927m | $287m | 20.3% | 10.2x |
| Rocket Lab (RKLB) | $38,250m | $36,000m | $769m | 37.3% | 46.8x |
| Intuitive Machines (LUNR) | $3,520m | $3,630m | $490m | 16.4% | 7.4x |
| Redwire (RDW) | $2,660m | $2,200m | $426m | 22.7% | 5.1x |
| Karman (KRMN) | $5,490m | $6,320m | $590m | 41.7% | 10.9x |
| Planet Labs (PL) | $7,070m | $6,830m | $336m | 55.6% | 20.4x |
| Voyager (VOYG) | $2,050m | $2,170m | $174m | 10.9% | 12.5x |
| AST SpaceMobile (ASTS) | $23,000m | $23,710m | $115m | 38.9% | 202.6x |
| L3Harris (LHX) | $49,580m | $59,060m | $22,930m | 25.5% | 2.6x |
Applying trailing peer multiples to Firefly’s forward revenue is the commonest way to make this stock look cheap, and it is wrong. Put every name on the same footing — enterprise value divided by annualised second-quarter revenue — and Firefly is at 6.2x against Intuitive Machines at 4.4x and Redwire at 4.7x. Redwire reported $117.1m in the same quarter Firefly reported $117.7m, at a similar gross margin, and trades at half the multiple. On a consistent basis Firefly carries a 32–41% premium to its two closest analogues, not a discount.
Regressing EV/sales on gross margin across the cohort, excluding AST SpaceMobile as an outlier the fit cannot accommodate, gives 2.55 + 0.415 x gross margin, which puts Firefly 7% below the line. But that result is carried almost entirely by Rocket Lab. Remove that one name as well and Firefly sits 24% above the fitted line while the explanatory power roughly doubles, from an R-squared of 0.19 to 0.37. An R-squared of 0.19 is itself the most important output here: across this cohort, sales multiples are not being set by fundamentals at all, which is exactly why we take the peer evidence as a description of clearing prices rather than of value.
The relative-value arguments that survive scrutiny are the balance sheet — $608m of net cash against Intuitive Machines’ net debt — and the landing record. The multiple is not one of them. Anyone buying Firefly because it is cheap versus Rocket Lab is making a bet on one comparable whose own rating no cash-flow model can reproduce.
A risk charged in the cash flows and again in the discount rate is charged twice, and double-counting is how a forensic review turns into a bear case by accident. This table states where each one sits.
| Risk | Priced in | How, and only there |
|---|---|---|
| Backlog does not convert; FY2027 undershoots | Cash flows | FY2027 base revenue set at $673m, below the higher consensus reading, with the growth path decaying from there. Not also charged in the discount rate |
| Operating expense never stops growing | Scenario weights | The bull case explicitly allows the cost base to reach $685m. The bear holds gross margin down instead. This is the model’s load-bearing input and it is on a slider |
| Gross margin stays near 20% | Cash flows | Carried as a 24 / 30 / 36% distribution across the three scenarios rather than a point. Not also charged in the exit multiple |
| Alpha or Block II failure | Scenario weights | A failure and a stand-down sit inside the bear scenario’s revenue and cadence path. Not also charged as an execution premium in the WACC |
| Eclipse slips again | Cash flows | Base assumes first flight in 2028, a year later than the company’s no-earlier-than-2027, with cadence deferred to 2030 |
| Cash burn and the 2027 raise | Cash flows | The full burn is in the explicit forecast. Only the bear case inflates the share count, to 200m. Charging both the burn and the dilution in every scenario would charge the same event twice |
| Employee dilution | Cash flows | Stock-based compensation of roughly $59m a year is inside the operating-expense line and never added back, which makes this model $17.0m a quarter more conservative than the company’s own adjusted EBITDA |
| AE Industrial and SciTec share supply | Discount rate | A single 1.5 point premium covers both blocks. These are existing shares, not dilution, so they appear nowhere in the share count |
| Customer concentration | Discount rate | A reduced premium, reflecting a 41.3% largest customer — but without a deteriorating framing, since year-on-year concentration has more than halved. See Section 07 |
| FY2027 appropriations risk | Cash flows | Base assumes an FY2027 outcome below the President’s request. Not also charged in the terminal multiple |
| Securities class action | Already in the price | $10–40m order of magnitude against a $3.5bn market capitalisation. No separate deduction |
| Goodwill impairment | Not priced — explained | Non-cash, and the underlying economics already sit in the revenue path. Charging it would be a double count |
| Thin margins, unproven cadence, sub-scale | Already in the price | These are what the discount to Rocket Lab and Planet Labs is compensating for. Not charged again here |
Three methods, weighted 35 / 45 / 20, run across the three scenarios. Every figure below is reproducible in the model at the top of this page — load a card and read it off.
| Method | Weight | Bear | Base | Bull | What it can and cannot establish |
|---|---|---|---|---|---|
| Discounted cash flow | 35% | $1.35 | $13.75 | $27.00 | The only method that tests whether the terminal economics are reachable — and demonstrably mis-calibrated for this sector, which caps its weight |
| Peer multiple | 45% | $19.39 | $25.39 | $28.24 | 6.0x FY2027E revenue plus net cash, inside a peer forward range of 4.4x to 7.9x. Anchored to observable clearing prices — in a cohort that is itself well off its highs |
| Revalued at sector rates | 20% | $11.16 | $23.32 | $30.18 | The same forecast at an 11% WACC, an 18% terminal EBIT margin and a 14x exit — what this sector actually clears at. Neither bullish nor bearish; a consistency check |
| Weighted | 100% | $11.43 | $20.90 | $28.19 | The bear and bull ends set the band; the base sets the central estimate |
At $21.12 the shares sit inside that band, at the 56th percentile, within 1% of the $21.00 central estimate. The band is the model run across a reasonable input range, not a separate opinion, and you can reproduce all three ends of it in the panel above. Holding every method between 20% and 50% of the weight instead of 35 / 45 / 20 moves the central estimate only between $18.95 and $22.44, so the conclusion does not depend on the weighting being exactly right.
The band is 2.6 times wide from floor to ceiling. That is uncomfortable and it is honest: more than all of this company’s value sits in a terminal state nine years away, informed by four quarters of public reporting, and no narrower band would be truthful in either direction. The floor is low because a business that never covers its cost base is worth little more than its cash; the ceiling is high because a demonstrated lunar landing, an accredited defence-software incumbency and a prime-backed medium-lift vehicle are each worth a great deal if they land. On the arithmetic, we rate the shares fairly valued. At the low the implied enterprise value is 2.1 times FY2027E revenue, the central estimate 4.8 times, the high 7.0 times; the current price is 4.8 times and the $39.50 sell-side mean is 9.8 times.
Firefly released second-quarter results and filed its Form 10-Q on 11 August 2026. Revenue was the largest in the company’s history and the first above $100m. The shares closed at $26.36 that day, peaked at $27.18 on 17 August, and have fallen in almost every session since, to $21.12 — a 22% decline from the post-earnings high in ten sessions.
| Q2 2026 | Result | Q/Q | Comment |
|---|---|---|---|
| Revenue | $117.7m | +45% | Against consensus of $89.6m — a 31% beat. Up 659% year on year as reported, which is almost entirely acquisition arithmetic |
| — Launch | $9.4m | −29% | 8.0% of revenue, recognised in a quarter with no Alpha flight. See Section 07 |
| — Spacecraft Solutions | $108.3m | +60% | 92.0% of revenue. Lunar hardware and defence software, reported as one line |
| Gross margin | 20.3% | −130bp | Down from 21.6% in Q1 and 27.7% in Q4 2025, as the acquired business rose to roughly 45% of revenue |
| Adjusted EBITDA | −$61.2m | — | Narrowed from −$64.7m. A second consecutive sequential improvement, and the only current evidence of operating leverage |
| Net loss | −$92.3m | — | Against −$96.7m in Q1. GAAP EPS −$0.57 against non-GAAP −$0.42; the gap is stock compensation, which is a real cost |
| Free cash flow | −$106.3m | — | The number that genuinely deteriorated. Worse than −$78.9m in Q1. Burn accelerated in the record quarter |
| Backlog, company-defined | $1.5bn | +15% | Record |
| Remaining performance obligations | $563.7m | falling | Down for a fourth consecutive quarter from a $723.1m peak in Q3 2025 |
| Cash and short-term investments | $635.3m | — | Plus a $305m undrawn revolver maturing 8 August 2028. Deferred revenue $209.5m |
Firefly guides one metric and one metric only: annual revenue. It has never issued EBITDA, capex, free-cash-flow, margin or quarterly guidance. On that single metric the record since the IPO is clean — one raise, one beat of the raised range, and now three consecutive reiterations of $420–450m for FY2026, with booked coverage improving from about 80% at initiation to about 95% of the midpoint.
Operational guidance was a different story. The 2026 Alpha launch target was cut from four flights to three, the first explicit operational reduction since the IPO, and Alpha Flight 8 — the first Block II vehicle — moved from the third quarter into the fourth. That a dollar guide could be reiterated while the flight count fell is the clearest possible statement of how little of this revenue now depends on the rocket.
All eight covering brokers refreshed within 48 hours of a 31% revenue beat, and they cut their price targets while maintaining their ratings. B. Riley went from $60 to $50, Roth from $60 to $45, and Jefferies to $35 from the $52 it had set on 2 June. Wells Fargo initiated at Equal Weight with a $25 target, the low mark. Goldman Sachs held at $34 and Morgan Stanley at $37. The mean sits at $39.50, and every published target is above the current price.
When every analyst keeps the thesis and cuts the multiple, the market is not re-rating the business. It is re-rating what it will pay for the business. That is a different problem, with a different cure, and it does not respond to good news. Firefly announced a $144.2m NASA lunar award, an Esrange milestone, an acquisition and a $13m Mars subcontract between late June and early July, and the shares fell anyway. When contract wins stop moving a stock, the market has stopped underwriting the backlog — and Section 01 explains why it may be right to.
One figure in this table is contested. Consensus non-GAAP EPS is reported at −$0.50 by the aggregator whose prior surprise history we can trace, which makes the −$0.42 result a 16% beat; a second source carries −$0.22, which would make it a large miss. We use the better-documented figure and print the disagreement rather than choosing silently. Sequential growth rates, the launch and spacecraft split and the margin comparisons are ours, computed from the reported quarters.
Firefly Aerospace was incorporated in Delaware in January 2017, rebuilt from the assets of the bankrupt Firefly Space Systems, recapitalised by Noosphere Ventures, and sold under CFIUS pressure to AE Industrial Partners in February 2022. AE Industrial remains the controlling shareholder and Firefly is a controlled company under Nasdaq rules. It listed on 7 August 2025 at $45.00 a share. It is headquartered in Cedar Park, Texas, recognised no revenue outside the United States in any period presented, and reports as a single operating segment with the chief executive as the chief operating decision maker evaluating consolidated net loss. No segment profitability is disclosed anywhere. That one choice is the central analytical obstacle in this name and it recurs in almost every section of this report.
| Engine A — national-security software | Engine B — launch, lunar and space systems | |
|---|---|---|
| Contents | SciTec-derived: FORGE/OPIR missile warning, ground radar digitisation, AI algorithms, Golden Dome interceptor work | Alpha, Eclipse, Blue Ghost, Elytra, Ocula |
| Est. FY2026 revenue | ~$220m | ~$218m |
| Economics | Cost-plus and fixed-price with hardware pass-through. Development is customer-funded, so company research spend is around 3% of revenue | Fixed-price deliveries carrying essentially the entire ~$290m research burden, with zero Eclipse revenue |
| Comparable terminal multiple | 2.0–2.5x sales on government software. Firefly itself paid 3.36x for this asset in an arm’s-length deal ten months ago | Wide. Redwire clears at 5.1x, Intuitive Machines at 7.4x |
| Status of the split | Unverifiable. Neither engine’s margin is disclosed. The only observable evidence runs against the software business being the high-margin one: consolidated gross margin fell 7.4 points as SciTec’s weight rose from zero to about 45%. We carry the split as an estimate and never rely on it in the model | |
| Period | Revenue | Gross margin | Net loss | Interpretation |
|---|---|---|---|---|
| FY2023 | $55.2m | 48.2% | −$135.5m | Best margin in the record, on a launch-weighted mix. Not comparable to today |
| FY2024 | $60.8m | −18.7% | −$231.1m | Negative gross margin on contract loss provisions. The low point |
| FY2025 | $159.9m | 19.2% | −$298.3m | +163%, of which two months is SciTec. Blue Ghost Mission 1 landed in March |
| Q1 2026 | $80.9m | 21.6% | −$96.7m | First full quarter with SciTec consolidated |
| Q2 2026 | $117.7m | 20.3% | −$92.3m | Record revenue, margin down sequentially, EBITDA loss narrower for a second quarter |
| Trailing twelve months | $287.0m | ~20.8% | −$363m | The base from which every multiple in this report is struck |
Essentially all revenue is US government or prime-contractor derived: NASA under the Commercial Lunar Payload Services programme, the Space Force and Space Systems Command, the Missile Defense Agency, AFRL, the Defense Innovation Unit, and primes including Lockheed Martin, Northrop Grumman and Blue Origin. In the second quarter the three largest customers were 41.3%, 16.2% and 14.6% of revenue, disclosed in the credit-risk note rather than the revenue note.
In the first half of 2026 Firefly produced $41.3m of gross profit against $139.0m of research and development and $93.2m of selling, general and administrative expense. Gross profit funded 17.8% of the operating cost base. Research and development grew 48% year on year and is heading higher, because the Eclipse medium-lift vehicle — the largest single use of cash — has not yet reached its most expensive phase. At a 20.3% gross margin, covering that annualised cost base needs roughly $1.14bn of revenue, 2.6 times this year’s guidance. Either the margin rises substantially or the revenue base becomes very large, or both, and Section 02 is largely an argument about which.
| Dimension | Score | Assessment |
|---|---|---|
| Regulation & accreditation | 9/10 | Cleared facilities, 475+ cleared staff, accredited missile-warning ground systems, CLPS vendor status, FAA launch licences and multiple range agreements. The hardest thing on this list to replicate and the least visible from outside |
| Intellectual property & technology | 8/10 | Four engine families in-house, carbon-composite structures, and a demonstrated soft lunar landing. Vertical integration on the SpaceX model, which is unusual outside SpaceX |
| Switching costs | 8/10 | Mission-specific lunar integration and accredited defence ground software both carry very high switching costs once a programme is running. Recompete incumbency in ground processing is sticky |
| Distribution & relationships | 8/10 | Northrop exclusivity plus equity, Lockheed extended to 2031, NASA CLPS incumbency, onboarded to NITE-STAR and MDA SHIELD vehicles. Access here is granted, not bought |
| Customer loyalty | 7/10 | Lockheed extending its agreement after losing a satellite on Alpha Flight 6 is the strongest single loyalty datapoint available anywhere in this file. Against it, VICTUS HAZE was redirected to a competitor — a disclosed loss |
| Brand | 6/10 | Strong and specific inside the customer set: the one that landed, and the one that did VICTUS NOX. Weak with generalist investors, who cannot see the defence business at all |
| Scale | 5/10 | $287m of trailing revenue against primes at $23bn. Manufacturing capacity has been built ahead of revenue, which is the right sequence and is currently pure cost |
| Data | 5/10 | Flight and lunar-surface telemetry plus SciTec’s algorithm libraries have genuine value, but there is no compounding data loop of the kind that scores highly here |
| Cost advantage | 4/10 | Alpha at roughly $14,500 a kilogram against Falcon 9 rideshare under $6,000. Vertical integration should deliver a cost advantage eventually; at three flights a year it does not |
| Network effects | 2/10 | Essentially absent. This is not a business with network effects and it is not going to become one |
Expanding. SciTec made Firefly an incumbent on Department of Defense missile-warning ground processing at exactly the moment that account is being asked to double, and incumbency on accredited ground systems survives recompetes. In lunar, Blue Ghost Mission 1 created a credential nobody else holds cleanly, and NASA is enlarging the programme — the CLPS ceiling is rising from $2.6bn to $4.2bn against up to 30 landings from 2027. A bigger programme awarded to a short, credentialed list is moat widening, not competition.
Shrinking. Dedicated small launch is the weakest position in the industry and it is getting weaker. Falcon 9 rideshare structurally undercuts Alpha for anything that tolerates a standard orbit, and Electron has far greater cadence. In medium lift Firefly is arriving last: Neutron, Terran R and Nova all target debut in the same window as Eclipse, which has already slipped from 2026 to no earlier than 2027.
The durable advantages are the ones a competitor cannot buy — clearances, accreditations, programme incumbency, a demonstrated landing. The weakest are the ones well-funded competitors are actively replicating on a two-to-three-year horizon: vehicle technology and launch cost. That is why a launch failure damages sentiment far more than it damages the moat.
| Competitor | Where they are stronger | Where Firefly is stronger |
|---|---|---|
| SpaceX | Everything to do with launch: cost, cadence, reuse, scale. Now a listed mega-cap absorbing the sector’s generalist bid | Nothing material. Firefly does not compete with SpaceX; it competes for the work SpaceX does not want or is not permitted to monopolise |
| Rocket Lab | Cadence, gross margin at 37.3% against 20.3%, scale at $769m of trailing revenue, a components business, and a vastly higher market rating | Lunar landing credential, defence-software depth, a net cash position, and a prime as exclusive medium-lift partner |
| Intuitive Machines | Nothing structural. Comparable scale and a similar CLPS position | Landed cleanly where IM-2 tipped over. Higher gross margin, $608m of net cash against net debt, and five landers in parallel build against a single-lander cadence |
| Astrobotic, Relativity, Stoke | Relativity’s private mark exceeds Firefly’s entire enterprise value on a vehicle that has never flown | Firefly has revenue, a flight-proven orbital rocket and a landed spacecraft. If these names list, the read-across is more likely to lift Firefly’s multiple than compress it |
| L3Harris, Northrop, Leidos, RTX | Scale, profitability, cash generation — and they bid the same missile-warning ground recompetes SciTec depends on | Growth rate, and a lunar and launch franchise the primes have chosen to partner with rather than build |
We ran a deliberately hostile read of the filings, the balance sheet and the registration statements, and then ran the same exercise in the other direction, because a review that finds only problems is not forensic. This is not a fraud pattern and we are not implying one. There is no accounting aggression, no debt problem, no going-concern issue, no miss on the one metric the company guides, and no published short thesis. What the concerns describe is a business that has not yet reached the scale its cost base requires, inside a share register that is currently setting the price.
The advantages below are, on the evidence, the stronger half of the ledger. A completed lunar landing, a lengthened prime relationship, an accredited defence incumbency, $608m of net cash and customer-funded working capital are verified facts about what the company has already done. Several of the concerns — the 2027 raise, the Block II record, the appropriations outcome — are forecasts about what has not happened yet. We do not net them off against each other, and neither should a reader, but the asymmetry in evidential quality is real and it is why the rating is not lower.
| Concern | Sev. | Evidence, and why it costs a shareholder money |
|---|---|---|
| Marketed backlog is 2.7x enforceable backlog, and the two move in opposite directions | 8/10 | $1.5bn company-defined against $563.7m of GAAP obligations at the identical date, the latter down four consecutive quarters from a $723.1m peak, with no published reconciliation and no published definition. Value the enterprise against enforceable revenue and the multiple is 5.2x, not the 1.95x the marketed figure implies |
| FY2027 consensus rests on revenue that is not under contract | 8/10 | Roughly $237m of RPO falls within twelve months and this year’s 95%-booked guidance consumes about $215m of it, leaving on the order of $81–140m of enforceable support for a ~$686m consensus. The street is underwriting roughly $550m of not-yet-contracted revenue, and it is the first thing to break if appropriations slip |
| Free cash burn accelerated in the record quarter | 8/10 | −$106.3m against −$78.9m in Q1. Running the disclosed path forward leaves roughly $430m at end-2026 and $43m at end-2027, and the revolver matures in August 2028 — before peak burn. The next equity raise is a 2027 event, not a 2028 one, and at current prices it happens near $21 rather than a recovered level. Sized to hold a $200m working balance, that is $600–700m |
| Gross margin is 20.3% and fell as the acquired business scaled | 6/10 | 27.7% in Q4 2025 with SciTec in for two months of three, then 21.6%, then 20.3% as its weight rose to about 45%. Explained as pass-through hardware; no margin target has ever been given. Twenty per cent on a services-and-hardware mix is government-contractor economics, and those are worth four to five times sales, not ten |
| 61.8m registered shares against a $103m-a-day tape | 6/10 | AE Industrial retains about 50.7m shares — 30% of the company and 47% of the float — having sold 8.0m at $48.00 on 28 May 2026. A separate 11,111,116-share resale registration for the SciTec sellers went effective 11 August 2026 and all lock-ups expired 12 August. This is not dilution, the shares already exist, but it is a mechanical price-setter — the stock peaked five days after the lock-up lapsed and has declined since |
| Alpha has three clean successes in seven attempts, and Block II is unflown | 6/10 | Two of six pre-IPO flights were clean. Flight 6 lost a Lockheed Martin satellite to plume-induced flow separation, an aerothermal design issue requiring vehicle changes. A ground test destroyed a booster on 29 September 2025. Flight 7 succeeded on 11 March 2026. Block II is a materially changed vehicle — stretched tanks, new avionics, new thermal coating — so its first flights carry new-vehicle risk rather than heritage |
| Launch revenue was recognised in a quarter with no launch | 5/10 | $9.4m of launch-segment revenue against a stated point-in-time recognition policy and zero Alpha flights; Q1’s $13.3m is consistent with one flight. Not an allegation of impropriety — the segment almost certainly contains responsive-space task orders, mission management or Eclipse engineering. But it means an outside model cannot infer Alpha’s unit economics from segment revenue, which is a real analytical loss |
| Goodwill and intangibles are 51% of book equity | 4/10 | $619.3m, of which $436.3m of goodwill from SciTec. The deal was announced at an $855m headline and booked at $550.3m, the gap being the collapse in Firefly’s own share price between signing and the 31 October 2025 close. The stock consideration was effectively issued at roughly a third of the IPO price, which is real permanent dilution the headline hides. Impairment itself is non-cash and is not charged in the valuation |
| Securities class action over IPO-period statements | 4/10 | Diamond v. Firefly Aerospace Inc. et al., W.D. Tex. 1:25-cv-01812, filed 11 November 2025, class period 7 August to 29 September 2025, alleging overstatement of Spacecraft Solutions demand and of Alpha’s reliability. Section 11 carries near-strict liability on the registration statement. Direct exposure is modest, but discovery into pre-IPO reliability assessments is a distraction during a Block II campaign |
Single-customer concentration of 41.3% in the second quarter reads as a deteriorating risk, and we initially treated it as one. The year-on-year disclosure says otherwise. In the first half of 2025 a single customer was 83.8% of revenue with only two material customers; in the first half of 2026 the largest is 34.4% with four. Concentration has more than halved and the base has broadened — because of the acquisition that is elsewhere treated as adding risk. The absolute level is still high and US government contracts are terminable for convenience, so a premium is warranted. The deteriorating framing is not, and we have removed it.
A forensic review that finds only problems is not forensic, it is one-sided. These are the advantages we think the market is not paying for, and why.
| Advantage | Evidence | Why the market may not be paying for it |
|---|---|---|
| The only fully successful commercial lunar landing, with five more contracted | Blue Ghost Mission 1 landed upright in Mare Crisium on 2 March 2025 and operated 346 hours with all ten NASA payloads functioning — the same period in which a competitor’s lander tipped over. Six contracted lunar missions, five in parallel build in quadrupled cleanroom space | The landing is eighteen months old and the stock has been repriced almost entirely off launch failures and share supply. NASA is raising the CLPS ceiling from $2.6bn to $4.2bn and has signalled up to 30 robotic landings from 2027, against thirteen eligible vendors and one that has done the job cleanly |
| A defence-software business aimed at the fastest-growing line in the Pentagon | SciTec holds FORGE/OPIR with a superior performance rating from the Department of Defense, and has added a $109m FORGE change proposal, $94m of Ground-Based Radar Digitization and an AFRL algorithm contract since closing. The FY2027 Space Force request is $71.3bn against roughly $32bn in FY2026 | Investors bucket Firefly with launch peers and value it on launch cadence, so a business that is 92% not-launch is valued on the 8%. SciTec has no separate revenue line, so the market cannot see it — and Firefly paid 3.36x revenue for it ten months ago in an arm’s-length deal |
| The customer with the best information on Alpha lengthened its commitment | Lockheed Martin lost a satellite on Alpha Flight 6. On 11 August 2026, with full knowledge of the record, it extended the multi-launch agreement by two years, allowing orders of up to 25 Block II missions through 2031 | The market read the 2026 cadence cut as evidence Alpha is failing commercially. The best-informed counterparty in the world on that question reached the opposite conclusion in the same week, and it is in no published model |
| A clean balance sheet with customer-funded working capital | $608.3m of net cash, 17.2% of the market capitalisation, against $27.0m of notes payable. No convertible, no maturity wall, no going-concern qualification; term loans repaid in full at the IPO. Roughly 90% of contract value is collected before launch and deferred revenue is $209.5m | The market is anchored on the burn rate and has not credited the financing structure. Growth here is funded by customers rather than by the balance sheet, and the absence of converts means no forced dilution at depressed prices |
| Elytra is a services franchise being contracted one award at a time, valued at roughly zero | Disclosed Elytra-attributable awards total roughly $88m — $75m MoonFall through 2028 and $13m SkyFall — plus a DIU/SDA deorbit-services preliminary design review won 13 August 2026 with the right to compete for execution, and Project Sinequone targeted for 2027 | No separate revenue line, no separate backlog, and each award is individually small enough to ignore. The pattern — deorbit, space domain awareness, lunar transfer — is a repeat-demand services business with an explicit orbital-debris policy tailwind |
| Northrop Grumman is an equity holder, exclusive partner and captive first customer for Eclipse | A $50m equity investment on 29 May 2025 plus exclusivity. Firefly builds the first stage for both Eclipse and Antares 330. Northrop has an ISS resupply obligation and no domestic first stage since the Russian supply chain ended | Eclipse contributes zero revenue and its slip to no earlier than 2027 was read purely as bad news. A prime that needs you to succeed is not a discretionary customer, and that structural fact carries no value in any current model |
Three of these are load-bearing and the rest are context. The backlog gap, the accelerating burn into a 2027 raise, and the share-supply overhang are the whole of it, and only the first two are about the business. Not higher than six, because there is no accounting aggression, no debt problem, no going-concern issue and no guidance miss on the metric the company guides. If you own this, the two things to watch are the Q3 RPO figure and any Form 4 activity showing whether AE Industrial is still selling.
Set against that, the advantages score seven, and that is the half more likely to be under-weighted. A demonstrated lunar landing, a defence-software incumbency pointed at a doubling budget line, a prime relationship that lengthened after a failure, $608m of net cash and an unmodelled services franchise are not consolation prizes. Note also what separates the two halves: the advantages are things that have already happened, while the heaviest concerns — the 2027 raise, the Block II record, the appropriations outcome — are forecasts that resolve one way or the other within nine months.
| Area | Score | Assessment |
|---|---|---|
| Guidance accuracy | 4/5 | On revenue: one raise, one beat of the raised range, three clean reiterations and no cut since the IPO. On operations: the 2026 Alpha target was cut from four flights to three, and the Flight 7 return to flight slipped three to five months against its original date |
| CEO track recordJason Kim | 3.5/5 | Chief executive since August 2024, from the US Air Force Academy via Raytheon and Northrop. In two years: the first fully successful commercial lunar landing, an IPO, a $550m acquisition, an Alpha return to flight and a follow-on. Tenure is short and the public record is four quarters long |
| Capital allocation | 3/5 | The SciTec acquisition was strategically sound and arguably the best thing that has happened to the business — but it was announced at an $855m headline and booked at $550.3m because the stock consideration collapsed between signing and closing. Paying in equity at a third of the IPO price is expensive money. The June 2026 primary raise at $48.00 was, by contrast, opportunistically well timed |
| CFO credibilityDarren Ma | 2.5/5 | Consistent, and consistently narrow. Annual revenue is the only guided metric; no EBITDA, capex, cash-flow or margin guidance has ever been issued, no margin target has ever been published, and Blue Ghost gross margin has not been broken out in any filing. Consistency is a virtue. This much opacity is not |
| Dilution management | 2.5/5 | Share count rose 13.4% in nine months. Stock-based compensation is scaling from a near-zero base to roughly $59m a year, about 1.7% of annual dilution from compensation alone before any raise |
| Board quality | 2.5/5 | A classified board in three staggered classes, plurality voting, cumulative voting prohibited, and controlled-company exemptions from a majority-independent board and from fully independent compensation and nominating committees. Kevin McAllister, formerly president of Boeing Commercial Airplanes, is a genuinely strong appointment on substance |
| Insider ownership and alignment | 2/5 | AE Industrial held 36.7% of the economics at 30 April 2026 and, with affiliates, more than 50% of voting power for director elections under a Director Nomination Agreement. It sold 8.0m shares at $48.00 two days after the all-time high. That is normal post-lock-up mechanics and it is also a signal, and both things are true |
| Compensation | 2/5 | FY2025 totals of $38.2m for the chief executive, $11.3m for the chief financial officer and $11.3m for the chief operating officer — roughly $60.9m for three officers at a company with $159.9m of revenue and a $298.3m net loss. Largely non-recurring IPO-year equity grants, but they are the source of the compensation charge now running through the income statement |
| Transparency | 2/5 | The weakest score, and the most consequential. Single-segment reporting on two economically distinct businesses; no reconciliation between $1.5bn of marketed backlog and $563.7m of GAAP obligations; no disclosure of SciTec’s organic contribution; no published definition of backlog. None of this is improper. All of it makes the company harder to value than it needs to be, and the market has responded by discounting the number it cannot verify |
Reporting SciTec as a separate segment with its own revenue and margin. It would let the market value an accredited defence-software business on defence-software multiples instead of valuing the whole company on launch cadence, and it would settle the single largest unknowable in this report — the organic growth rate. It is entirely within management’s gift and it costs nothing.
On whether they act like long-term owners: the operating team does. The controlling shareholder does not have to, and its incentives are those of a private-equity fund approaching the end of a hold period rather than those of a permanent owner. That is the honest reading of a sponsor that sold into the top of the range while retaining 30% of the company, and we price it as a discount-rate premium rather than treat it as a character judgement.
| Event | Timing | Upside | Downside |
|---|---|---|---|
| FY2027 appropriations, or a continuing resolution | US fiscal year begins 1 Oct 2026 | An enacted Space Force appropriation near the $59.2bn base request unlocks new starts in resilient missile warning and ground infrastructure — SciTec’s exact aperture — and funds 31 national-security launches against 11 in FY2026 | A full-year continuing resolution at FY2026 levels freezes new starts and slows obligations into the second-half recognition window. The sector de-rates on the headline regardless of company exposure |
| AE Industrial and SciTec share supply | Ongoing since mid-Aug 2026 | The block clears, by marketed secondary, LP distribution or open-market sales, and the mechanical pressure that took the shares from $27.18 to $21.12 ends. Overhang removal is typically followed by re-rating where fundamentals have not deteriorated, and here they have not | A second marketed secondary at a discount to an already-depressed price resets the reference lower and confirms the sponsor will sell at any level. ~61.8m registered shares against $103m a day of volume |
| CLPS 2.0 contracting vehicle | NASA has said a new vehicle is needed by 30 Sep 2026 | An on-ramp with Firefly’s submitted multi-tonne lander design, against up to 30 landings and a ceiling raised to $4.2bn. A 25–30% share is $1.0–1.3bn of award value — 35–44% of today’s enterprise value from one programme | Slippage past FY2026 is likely on procurement history. An on-ramp admitting SpaceX, Blue Origin and Lockheed on equal terms commoditises lander delivery |
| Q3 2026 results | Early-to-mid Nov 2026 | Revenue of $110–130m with gross margin recovering above 21% and a third consecutive narrowing of the EBITDA loss. Two data points is noise; three is a trend. The Q3 RPO figure is the single most informative number this company publishes | Consensus already models a sequential decline to about $104m. A guidance cut from a 95%-booked year would be a severe credibility event, and burn worse than −$106.3m pulls the raise forward |
| Alpha Flight 8 — Block II debut | No earlier than Q4 2026 | A clean flight takes Alpha to four successes in eight and validates the stretched tanks, new avionics and thermal coating that specifically address the Flight 6 failure mode. Two clean flights in one quarter would be the strongest cadence signal in the company’s history | A failure resets the record to three in eight, triggers an FAA mishap investigation and a six-to-twelve-month stand-down, and feeds the class action’s reliability allegations directly |
| Eclipse first-stage delivery and Miranda qualification | Stage delivery no earlier than 2027 | Formal engine qualification converts Eclipse from promise into hardware and validates the Northrop exclusivity. It is also the gate for Antares 330 return to flight — a captive customer independent of the merchant market | A further slip while Neutron, Terran R and Nova fly leaves Firefly entering medium lift last with the least reusable architecture, having spent the most. Research spend is already $139m per half and rising |
| FY2026 results and initial FY2027 guidance | Mid-March 2027 | FY2027 guided at or above $686m with RPO materially above $563.7m — the number that would genuinely change this analysis — or a move to reportable segments, which would let the market value the software business properly | FY2027 below $650m, margin guided flat near 20%, and a capital raise announced alongside. That combination confirms the bear framing of a sub-scale contractor on a growth multiple |
| Blue Ghost Mission 2 — far-side landing | No earlier than 2027, slipped from Dec 2026 | A world first for a commercial operator, validating the two-stage Blue Ghost, Elytra as an independent transfer stage, and parallel multi-mission build at once. It would put Firefly two clean landings ahead of every competitor | A landing failure erases the single differentiator justifying the premium to Intuitive Machines, triggers estimate revisions across five parallel lunar builds, and hits Elytra at the same time. The highest-consequence single event in the file, in both directions |
| Eclipse first launch from Wallops | No earlier than 2027 | A 16.3-tonne-to-LEO vehicle opens NSSL Lane 1 and Lane 2, ISS resupply via Antares 330, and merchant medium lift. At $50–60m a flight and eight to ten flights a year that is $400–600m of annual revenue — more than the entire company earns today | A first-flight failure or a slip to 2028 while three competitors fly strands several hundred million of research and capital spending and forces a raise into a broken narrative |
BULLYou are being offered four assets and paying for about one. At a $2.93bn enterprise value you get a demonstrated lunar lander, an accredited defence-software incumbent, a medium-lift vehicle with a prime as equity partner, and an in-space services franchise. Plus $608m of net cash, which is 17% of the market capitalisation.
BEARIt is a rocket company that stopped being one and hopes you will not notice. Launch was $9.4m of $117.7m. The 659% headline is bought, not earned, and the organic split has never been disclosed. You cannot tell me the growth rate of the thing you are asking me to buy.
BULLThe lunar credential is close to unique. Blue Ghost landed upright and ran 346 hours with all ten payloads working, in the same window a competitor tipped over. Thirteen vendors are eligible for CLPS and one has done it cleanly. NASA is raising the ceiling from $2.6bn to $4.2bn and signalling up to 30 landings from 2027.
BEARThe backlog is the tell. Management markets $1.5bn. The GAAP obligation at the identical date is $563.7m and it has fallen four consecutive quarters. The $936m difference binds no customer. On enforceable value the multiple is 5.2 times, not the 1.95 the pitch implies, and after this year’s 95%-booked guidance consumes the twelve-month RPO there is $81–140m of contracted support under a $686m consensus.
BULLThe market cannot see Engine A at all. SciTec sits inside Spacecraft Solutions with no separate line. It holds FORGE/OPIR with a superior performance rating and has added $109m and $94m awards since closing, pointed at a Space Force request of $71.3bn against $32bn. Firefly paid 3.36 times revenue for it ten months ago in an arm’s-length deal. You are getting it at zero.
BEARThe economics do not work at scale. Gross margin is 20.3% and it fell in the record quarter, as the software business you are praising scaled to 45% of revenue. First-half gross profit of $41.3m funded 17.8% of a $232.2m cost base. Breakeven needs roughly $1.14bn of revenue while research spend grows 48% a year.
BULLThe best-informed counterparty voted with its order book. Lockheed Martin lost a satellite on Alpha Flight 6 and then, in August 2026, extended the launch agreement to 2031 for up to 25 Block II missions. The market read the cadence cut as Alpha failing commercially. The one party with full information reached the opposite conclusion in the same week.
BEARAnd the people who know most are selling. AE Industrial sold 8.0m shares at $48.00 two days after the all-time high and still holds about 50.7m — 30% of the company, 47% of the float. The SciTec sellers’ 11.1m shares were registered on 11 August with no lock-up. That is 61.8m registered shares against a $103m-a-day tape, and the burn is accelerating into a 2027 raise at around today’s price.
BULLWhich is exactly my point, and it is mechanical rather than fundamental. A $48 secondary on 28 May, a resale registration, a lock-up expiring 12 August, a mega-cap competitor listing in June and a nineteen-year high in long rates. Over the same period backlog rose, guidance held at 95% booked, and the EBITDA loss narrowed twice. The late-August decline came on twenty-day volume of $47.8m against a $102.7m ninety-day average, short interest fell from 13.19m shares to 11.13m, and the 29 July low of $17.84 was never re-tested. That is a buyers’ strike, not distribution.
BEARThe DCF is unforgiving whatever the tape says. One cell in thirty reaches the current price, and only with an 18% terminal margin and $3.8bn of 2035 revenue simultaneously. Terminal value is 135% of enterprise value because the explicit decade destroys $659m before the business you are buying exists. And the sell side cut targets into a 31% beat — B. Riley minus ten, Roth minus fifteen, Jefferies minus seventeen.
BULLThey cut the multiple and kept the rating, which is the whole argument. Every professional model still sits above the price. The most bearish, at $25, implies 18% upside. And your DCF cannot reproduce the market price of a single name in the peer table, including the profitable ones.
CHAIRBoth hold, and that is why the rating is fairly valued rather than a fudge between two weak cases. The bear’s RPO analysis is the best-evidenced fact in this report and the bull cannot rebut it: $563.7m against a marketed $1.5bn, four consecutive declines, no reconciliation, a 42 / 21 / 37 recognition schedule. Setting FY2027 below the higher consensus follows directly and should be held with more conviction, not less. The margin arithmetic and the accelerating burn are equally solid. What the bull proves and the bear cannot rebut is that the 2026 round trip is not new information about the business. Every element of the decline is identifiable and mechanical while the disclosed operating measures improved. The net cash, the landing record and the Lockheed extension are facts, not narrative. Where the bear overreaches: treating a 41.3% largest customer as deteriorating concentration when it has fallen from 83.8% with the material customer count doubling, and leaning on a discounted cash flow that is mis-calibrated for the entire cohort. Where the bull overreaches: the claim that Firefly is cheap on forward metrics — on a consistent basis it trades at a 32–41% premium to Intuitive Machines and Redwire, and the peer-regression discount is entirely a Rocket Lab artefact.
Three things, none of which further work resolves from public filings. SciTec’s standalone revenue and margin. Organic book-to-bill — remaining performance obligations rose $20.9m over four quarters while $287.0m of revenue was recognised, but the acquired contribution is undisclosed, so the organic figure could be anywhere from about 1.07 times to roughly 0.55. And the terminal gross-margin path, on which the observed 7.4 point decline as SciTec consolidated is a warning rather than proof. Anyone quoting a point target on this name is expressing a view on the first of those and calling it a valuation.
One. The Q3 remaining-performance-obligation figure — the single most informative number this company publishes, and the one that would move our FY2027 assumption in either direction. Two. Whether Alpha Block II flies cleanly in the fourth quarter. Three. Whether FY2027 appropriations pass or a continuing resolution bites. Four. Any Form 4 activity showing whether AE Industrial is still selling, since the sponsor overhang is currently the dominant price-setting mechanism.
Firefly builds rockets and spacecraft. It makes a small rocket called Alpha that carries about a tonne into orbit, and it is building a much bigger one called Eclipse with Northrop Grumman. It also builds a lunar lander called Blue Ghost, which in March 2025 became the first commercial spacecraft to land on the Moon upright and complete a full surface mission — something only a handful of national space programmes had managed before.
In late 2025 Firefly bought a company called SciTec for $550m. SciTec writes software for the US military that processes satellite data to detect missile launches. That acquisition changed what Firefly is. Today more than nine of every ten dollars of revenue comes from spacecraft and defence software, and fewer than one from actually launching rockets.
Almost entirely from the US government and from big defence contractors. NASA pays it to deliver payloads to the Moon. The Space Force pays it for missile-warning software. Lockheed Martin and Northrop Grumman pay it for launches and rocket stages. Contracts are large, long, and paid in instalments as work is completed, which means revenue arrives in lumps and a single customer decision can move a whole quarter.
Firefly tells investors its order book is $1.5bn. Its own accounts say the part customers are actually obliged to pay for is $563.7m. Neither number is dishonest — the bigger one counts contracts that let customers order more if they want to, and the smaller one counts only what has been ordered. But the big number keeps setting records while the small one has fallen four quarters in a row, and the company does not publish anything explaining the gap. If you take one thing from this report, take that.
Right: the upgraded Block II rocket flies cleanly twice in the fourth quarter, NASA adds Firefly to an enlarged Moon programme, Congress passes the defence budget it has requested, and the private-equity owner finishes selling so the share supply pressing on the price disappears. Any two of those together would change the story materially.
Wrong: the rocket fails, triggering an investigation and a six-to-twelve-month stand-down. Congress passes a stopgap budget instead of a real one, freezing new programme starts. The company raises money in 2027 at a depressed price. Or the $1.5bn order book converts much more slowly than the accounts suggest.
One closing thought. A stock falling 66% while its revenue grows 42% is not automatically a bargain, and it is not automatically broken either. Here the fall has an identifiable mechanical cause and the business improved through it, which argues for the first reading — and the price still requires a decade of compounding that has not happened yet, which argues for caution. This is not a starter holding, and position size matters far more here than entry price. If a further halving would be intolerable, the answer is not to wait for a better price. It is to own less of it, or none.
| Status | What it covers |
|---|---|
| Verified from Firefly filings | All revenue, gross profit, operating expense, net loss, cash, short-term investments, notes payable, equity, deferred revenue, goodwill and intangibles, remaining performance obligations and their 42 / 21 / 37 recognition schedule, share counts, customer-concentration percentages, acquisition consideration and pro-forma figures, guidance ranges and named contract values — taken from the Form 10-Q for the quarter ended 30 June 2026 (filed 11 August 2026), the Form 10-Q for the quarter ended 31 March 2026, the Form 10-K for FY2025 (filed 20 March 2026), the Form 10-Q for the quarter ended 30 September 2025, the DEF 14A of 17 April 2026, the Form 424B4 of 1 June 2026 and the Form 424B3 of 11 August 2026, all for CIK 0001860160. Figures were checked against SEC XBRL company facts rather than taken from secondary summaries. |
| Verified from market data | The $21.12 closing price of 31 August 2026 and the daily bar series behind it; the $16–$62 52-week range; the $27.18 post-earnings high of 17 August and the $17.84 low of 29 July; realised volatility of roughly 81% over thirty days and 110% over one year; traded volume; short interest; and every peer market capitalisation, enterprise value, trailing revenue and gross margin in the Section 03 table. Analyst price targets, the $39.50 mean and consensus revenue and earnings estimates are from market-data aggregators as at 31 August 2026 and are attributed as such. |
| Our estimate or judgement | Enterprise value, net cash and every multiple in Section 01; the pro-forma FY2025 revenue of roughly $306m and the +42% like-for-like growth rate derived from it; the EV-to-enforceable-contracted-value ratio; the entire nine-year cash-flow path, the scenario settings, the sensitivity grid, the Gordon cross-check, the reverse test and the $11–29 band with its $21.00 central estimate; the 35 / 45 / 20 method weights and the 30 / 45 / 25 scenario weights; the split of revenue between the two engines; the peer regression and its exclusions; and all scores in Sections 06, 07 and 08. |
| Still open | SciTec’s standalone revenue, gross margin and contribution to remaining performance obligations are not disclosed anywhere, so the organic-versus-acquired split of growth is an estimate and organic book-to-bill cannot be computed. No reconciliation between the $1.5bn company-defined backlog and the $563.7m GAAP obligation exists, and the definition of backlog is not published. Segment profitability is not disclosed. FY2027 consensus revenue is contested between reputable sources at $584m and $686m, a 17% spread, and this report sits between them. Section 382 limitations on approximately $2.0–2.3bn of net operating losses are unverified, and the model’s tax shield assumes they do not bind. The terms of the $305m revolving facility beyond its August 2028 maturity were not obtained. The second-quarter earnings call transcript was not obtained, so nothing on this page is drawn from it — no management quotation and no characterisation of analyst questions appears here, and the quarter is described from the release and the Form 10-Q only. |
Method. Unlevered discounted cash flow, methodology version 1.2, in the J-curve form described in Section 02. Nine explicit years from FY2027 with terminal value by exit multiple on FY2035 EBITDA, discounted at 12.5% in the base case, against net cash of $608.3m measured at the 30 June 2026 balance-sheet date. Gross margin ramps linearly from the 20.3% reported in the second quarter; operating expense ramps linearly from the $480m first-half annualised run-rate; depreciation is held at 7% of revenue; cash tax is zero through FY2031 and 21% thereafter against an assumed $2.15bn of loss carryforwards. Stock-based compensation of roughly $59m a year sits inside the operating-expense line and is never added back, which makes this model $17.0m a quarter more conservative than the company’s own adjusted EBITDA. Share counts are 167.4m outstanding today, with terminal counts of 183m in the base and bull and 200m in the bear. The reference price used throughout is the $21.12 close of 31 August 2026, the last completed session before publication, and every multiple in this report scales directly with it.