Up 31.6% to US$151.21, and DOWNGRADED to SELL on all three horizons. The prior report said HOLD[14]. Be precise about why that changed, because it is not the price. August scored Business Quality 68, which falls in the High band, and the High-quality row of the Decision Matrix does not reach a sell. This run's published build returns 60, which falls in the Medium band, and that row does. The valuation band reads Expensive in both runs. The entire signal change is the Quality band, which is why that build is published in full below, with what it would take to move it.
The provider's forward price-to-earnings ratio for this stock is wrong, and this report does not use it. It shows 97.6×. Reconciling the consensus earnings-per-share strip against the consensus net income strip on the actual 13.57bn shares fails in every single year, by a factor of about four; dividing each year's net income by its earnings per share implies a share base near 3bn, which is the pre-IPO count[4]. Restated to the true count, the forward multiple is roughly 396×, not 98×. The anchor here is therefore built on revenue and EBITDA, which are consensus lines struck in dollars and cannot be corrupted by a share count.
So the anchor is built on EBITDA, and it says Expensive by a wide margin. Warranted forward enterprise-value-to-EBITDA works out at 15.06× on a two-stage discount at 11.28% with growth capped at the framework's top secular bucket; the actual is 60.47× on FY2027 consensus EBITDA, a ratio of 4.02× against a 1.40× Expensive line[4]. That holds at every growth assumption tested. One lens disagrees and is shown rather than dropped: on RAW FY2027 consensus revenue the name trades at 19.14×, below the 20× line that marks a technology name rich — only after the framework's mandatory 25% growth haircut does it cross. A reader should know one of the five lenses says this is not expensive.
An error in an earlier draft of this report, recorded rather than quietly fixed. I first set the warranted multiple to the 20× revenue guardrail line and read a ratio off it. That is a category error: the guardrail is a rich-line cap used to test an actual multiple, not a warranted level, and the band thresholds are defined against a discounted-cash-flow output. Sitting at 1.15× a rich line is not the same as sitting at 1.15× a fair value, and treating them alike understated this name badly. The renderer's own consistency check caught it.
Two things genuinely got better and are recorded as such. The lockup overhang the August report flagged has now passed three tranches — about 911.5m shares on 6 August, ~319m from 12 August and a further release on 10 September, taking the free float to around 17% — and the stock rose through all three[6]. Supply arriving without a price break is information. And the tape is much stronger: above both real moving averages with MACD positive, which lifts Timing up ten.
The dated event to watch is three weeks away. The US$60bn all-stock acquisition of Cursor's parent CLOSED ON 14 AUGUST, at 2.88% dilution[5]. The August report carried it in prose but recorded it in no machine field, so nothing downstream could see it; that is fixed here.
SpaceX designs and flies reusable rockets, and that reusability is the thing everything else rests on: it made launch cheap enough that the company could afford to become its own biggest customer. Starlink, the satellite broadband constellation it launches for itself, ended June with 12 million subscribers and is now the largest revenue line. In early 2026 the company absorbed xAI — Grok, the Colossus data centre and X — and it has agreed to buy Cursor's parent, so it is now also an artificial-intelligence business selling cloud capacity to Google and Anthropic. It listed on Nasdaq on 12 June 2026. It is filed as an Industrial, which describes where it started rather than what it now is.
What it is. A launch operator, a satellite broadband network and, since early 2026, an artificial-intelligence business — xAI, Grok, the Colossus data centre and X were folded in during the year, and the Cursor acquisition added a fourth in August[5]. Revenue grew 92% year on year to US$7.81bn in Q2 with adjusted EBITDA of US$3.5bn, up 191%, and a net loss narrowing to US$541m[2]. It is scored as an Industrials name on returns-versus-cost-of-capital and backlog.
| Component | Score | Weight | Contribution |
|---|---|---|---|
| Universal sub-signals (average of four) | 63.0 | 40% | 25.20 |
| Competitive moat (re-derived below) | 73.6 | 25% | 18.40 |
| ROIC & capital allocation | 41.2 | 20% | 8.24 |
| Industry benchmark | 52 | 15% | 7.80 |
| Business Quality | 59.64 → 60 | ||
The four universal sub-signals: revenue trajectory 92 — +92% year on year, and the company guides to US$100bn of annualised recurring revenue by December on contracted cloud deals with Google and Anthropic[3]; profitability versus peers 42 — still a loss, with an operating margin of −1.8% against −41.4% in Q1, so improving very fast from a negative base; cash generation 30 — free cash flow is deeply negative on roughly US$18.4bn of quarterly capital expenditure, and the framework is explicit that persistent negative free cash flow is a red flag at any growth rate; and balance-sheet health 88 — US$100.0bn of cash against US$39.7bn of debt is US$60.3bn net cash, which is as strong as this gets[1].
Who it competes with. On launch: Blue Origin, Rocket Lab, ULA, Arianespace and a Chinese state complex. On broadband: Amazon's Kuiper, OneWeb/Eutelsat and the terrestrial carriers Starlink has said it will target directly with a mobile service. In AI cloud it now sells to, and competes with, the hyperscalers.
Share trajectory: gaining, decisively. On the Q2 call management put SpaceX at 80-90% of total Earth mass to orbit per year[3], and Starlink ended the quarter at 12 million subscribers, roughly double a year earlier[2]. No new competitive disclosure has been made since that release.
Business Quality 60 — down eight against August.
Start with what cannot be used, because it rules out the obvious anchor. This company is loss-making — a US$541m net loss in the June quarter and US$4,276m in March[2] — so the data provider returns a null trailing price-to-earnings ratio[1], and any vendor figure struck on a negative base would be an artifact. More importantly the forward earnings consensus is defective: reconciled against the consensus net-income strip on the actual 13,572,821,625 shares it fails by roughly four times in every year, implying a share base near 3bn — the pre-IPO count[4]. The provider's enterprise-value field is broken too; it exceeds market capitalisation plus all debt, which is impossible for a company holding US$100bn of cash, so enterprise value is reconstructed from its components at US$1,992bn[1].
So the anchor is EBITDA, and the choice is deliberate. It is the only consensus line on this name struck in dollars, which makes it immune to the share-count defect, and it is the stream a capital-intensive infrastructure business is actually valued on. Warranted forward EV/EBITDA is 15.06×, from a two-stage discount at 11.28% — the 9 September ten-year of 4.78%, plus the framework's fixed 4.5% equity risk premium, plus a 2.0-point risk add-on that a derived beta of 3.40 earns several times over[9] — with near-term growth at the top secular bucket of 15%, used rather than the 10% cyclical cap because the growth is contracted rather than forecast. Actual is 60.47× on FY2027 consensus EBITDA of US$32.94bn. The ratio is 4.02×, against a 1.40× Expensive line.
The lens that disagrees, shown rather than dropped. On forward EV/Revenue the picture is genuinely different: 19.14× on raw FY2027 consensus revenue of US$104.09bn, which is below the 20× line that marks a technology name rich. Apply the framework's mandatory 25% haircut to consensus growth and it becomes 23.76×, above the line. On trailing revenue it is 86.45×. One of the five lenses says this is not an expensive stock, and a reader is entitled to weigh that themselves rather than be shown only the four that agree.
No point fair value is published, deliberately: with FY2027 consensus revenue spanning US$69.5bn to US$139.9bn, a single number would imply a precision that does not exist[4]. The scenario band is the honest form of the same answer. Two of the five lenses are likewise declared unavailable with their reasons rather than silently skipped — own-history, because the 12 June listing date leaves only about 63 sessions and no pre-IPO market price from which a multiple series could be built[1]; and PEG, because trailing earnings are negative and the forward strip is the defective one[4].
The sector-median lens, for completeness. Aerospace and defence peers trade around 2-3× EV/Revenue and 12-18× EV/EBITDA; large-cap technology trades at roughly 15-25× forward EV/EBITDA[12]. At 60.47× EV/EBITDA this name is two to four times the technology band and a multiple of the defence band — which is the arithmetic behind saying it is priced as something other than an industrial.
Valuation 34 — up ten. The price rose 31.6% while trailing revenue did not, and the analyst upside compressed from +85% to +45.9%.
Two drivers, pulling opposite ways.
For: Starlink's subscriber base roughly doubled year on year to 12 million and connectivity revenue rose 66%[2]; the company has contracted AI cloud revenue with Google and Anthropic that begins ramping in October, plus a further US$6.7bn signed in the first weeks of Q3[3]. That is about as concrete as a growth driver gets — it is signed rather than forecast.
Against: the macro report keeps AI & Productivity at HIGH dominance but describes the leadership as wobbling, with breadth narrowing — equal-weight below its 50-day while the cap-weighted index is above it, which it calls the concentration tail[10]. A US$2tn company that has just bought an AI lab, a data centre and a coding tool is squarely in that cohort. Separately, NATO Rearmament was downgraded to BACKGROUND with the defence ETF down 11.7% on the month, so the adjacent defence bid has stopped paying for now.
Underlying Drivers 48 — mixed, with the company-specific side clearly positive and the cohort-level side clearly not. At 48 the driver sits well above the 35 line, so it cannot amplify this SELL to a STRONG SELL.
Industrials (XLI) is Neutral on all three horizons[10]. SPCX is not carried in the macro report's watchlist, so the sector map is the whole of the mechanical read.
The August report recorded this as a Headwind with a Contrarian stance. The sector call has firmed since, which is a genuine improvement and is recorded as one. Stance Neutral, pressure Neutral, conviction 45. Mechanically this matters in one place: amplification to STRONG SELL needs a driver at or below 35 and Headwind pressure, and neither holds.
The honest caveat on a sector label for this name. Calling a company that launches rockets, sells satellite broadband and runs an AI lab an ‘Industrial’ is a filing convention, not an economic description. Its actual macro exposure is closer to the technology complex, which the same report has Underweight across every horizon. The sector map is used because the framework says to use it; the mismatch is disclosed because a reader should weigh it.
Source: sector map (GICS Industrials → XLI, N/N/N) · Macro report 2026-09-09
Most of the multi-timeframe table on this stock is fiction, so start there. SPCX listed on 12 June 2026 — about 63 trading sessions ago — yet the data provider returns 60 monthly bars, 149 weekly bars, and a daily 200-day moving average of 60.87 against a US$151.21 price[8]. Those series are back-filled through a period when this security did not trade. They are discarded, exactly as the August report discarded them. The DAILY frame returns 204 bars against 63 real sessions, so it is partly back-filled too — which is why its 200-day average of US$60.87 is unusable (it sits below the 52-week low) while its 20- and 50-day averages, which fall entirely inside the live window, are not. Only those two are real, and the trend term is scored on them alone — which is why it reads 62 rather than the 90-plus a five-timeframe ‘strongly bullish’ confluence reading would otherwise produce. Hourly and 15-minute volume ratios of 0.08 and 0.06 are Saturday artefacts and were excluded too.
On the real data the tape is genuinely strong. US$151.21 sits above both the 20-day (US$142.99) and the 50-day (US$135.07), RSI is 60.27, and MACD at 5.287 is above its 4.906 signal. The 52-week range position is 38.4%. That is a real improvement on August.
What holds Timing back is risk-reward, and it is a volatility problem. The nearest meaningful swing-low shelf is 130.39, so a stop that can actually be held sits at US$129.00 — US$22.21 away, or 2.89 ATR, beyond the 2.5-ATR line the framework calls unfavourable. This name's derived beta is 3.40 against the S&P with 93.4% annualised volatility, 7.65 times the index[9]. You have to risk a great deal to find out whether you are right.
timing = mtf 62 x 0.30 + risk-reward 32 x 0.20 + macro 38 x 0.15 + sentiment 72 x 0.18 + catalyst 60 x 0.17 = 53.86 -> 54.
Entry/Exit Timing 54 — up ten. Neutral, and it does not enter the matrix row that decides this name.
| Date | Event | Impact | Forecast | Previous | Relevant? | Why |
|---|---|---|---|---|---|---|
| 2026-09-16 | FOMC rate decision | Medium | hold | — | ⚠ Medium | Hike-versus-hold. Matters to a beta-3.4 name as volatility rather than as thesis; Industrials is medium macro-sensitivity. |
| 2026-09-18 | Bank of Japan decision | Medium | hike to 1.25% priced at 97% | — | ⚠ Medium | The carry unwind hits the highest-beta names hardest, and this is one of them. |
| 2026-08-14 | Cursor acquisition CLOSED (already happened) | High | — | 391,041,680 shares issued | ✅ Yes | Recorded because an earlier draft of this report treated it as still pending. It is in the share count now. |
| 2026-09-26 | Next scheduled Donatien review | Company-specific | — | — | ✅ Yes | Standing +14-day ceiling — no company-specific dated event remains inside the window. |
| 2026-10-01 | AI cloud revenue begins ramping | High | — | — | ✅ Yes | The Google and Anthropic contracts plus US$6.7bn signed in Q3 begin ramping in October. The single most important operational date on this name. |
| 2026-11-05 | SpaceX Q3 FY2026 results (approximate) | High | — | — | ✅ Yes | First new fundamental information since 4 August, and the first test of whether the October ramp is real. |
| Timeframe | Trend | Direction | RSI | MACD | Key S/R | Breakout | Vol |
|---|---|---|---|---|---|---|---|
| Monthly | BACK-FILLED — not scored | n/a | n/a | n/a | not usable | Pre-IPO synthetic | n/a |
| Weekly | BACK-FILLED — not scored | n/a | n/a | n/a | not usable | Pre-IPO synthetic | n/a |
| Daily | Strong uptrend | Bullish | 60.3 | + | S: 130.39 R: 155.00 | Above 20- and 50-DMA | 0.99x |
| Hourly | Strong uptrend | Bullish | 54.1 | + | S: 145.92 R: 155.00 | None | n/a — Sat |
| 15-min | Strong uptrend | Neutral | 49.5 | flat | S: 148.41 R: 151.85 | None | n/a — Sat |
The US$100bn annualised-recurring-revenue exit rate lands as guided, the contracted Google and Anthropic cloud revenue ramps from October as described, Cursor closes and adds its US$4bn, and the market keeps paying above 20 times revenue for the combination. A Tesla combination, still unannounced, would be additional optionality this case does not need.
FY2027 revenue arrives near the consensus centre and the multiple compresses modestly toward the guardrail as the loss narrows but profitability stays out of reach. That is roughly flat from here, and flat is the point: at this multiple a year of solid execution buys no return, because the price contains it already.
Revenue lands nearer the consensus low of US$69.5bn — a two-to-one spread means the low end is not a tail, it is a forecast someone credible is holding — and the multiple compresses to the 20x line. The cohort-level risk sits here too: if the concentration tail the macro report is watching does turn, a US$2tn AI-levered name de-rates with it regardless of its own delivery. The falsification for that leg is breadth broadening, equal-weight catching the cap-weighted index.
Probability-weighted US$151.50 against US$151.21 — +0.2%, essentially flat. That is the honest shape of this name and not a hedge: the framework says the price is ahead of the value, and the distribution says the optionality is real enough to offset it. A SELL here is a statement about price, not a forecast of collapse. The analyst panel is far more bullish at US$220.68[1], and the gap is the anchor — the panel is valuing the revenue ramp, this framework is discounting it 25% before capitalising it.
Forecast: Fundamental: cannot open while the revenue consensus spans a two-to-one range, because no honest point fair value can be struck over that dispersion. Forecast confidence: Low until the Q3 print narrows it. Technical: the reachable one, and it is close — price is already above the 50-day and MACD is positive; what is missing is volume, and volume on this name arrives with news. Forecast confidence: Moderate within 4-8 weeks, though the merger closing that might have supplied it is already behind us. Catalyst: opens with the November print. The practical read: with the merger already closed, only one decisive date remains — the Q3 print in November, which carries the first post-merger share count and the first evidence on whether the October cloud ramp is real.
Where this stops being a SELL — and it is a long way from here. The published band is set by the ratio arm, so the test is the warranted ratio falling to 1.40×. That means forward EV/EBITDA of 21.08×, an enterprise value near US$695bn, and a share price around US$55.63 — roughly 63% below today. Or, holding the price, FY2027 consensus EBITDA would have to roughly treble. An earlier draft of this report answered this question off the revenue guardrail instead and gave ‘a price near US$131’, which is not the arm the verdict is published on and materially understated this report's own conclusion.
What would make it worse. The revenue ramp slipping — the single most important number on this company is whether the October cloud ramp happens on schedule; a de-rating in the AI cohort, for which the tell is breadth, not this company's own results; or a close below US$129.00.
What is actually dated. The Cursor close is behind us, on 14 August. Ahead: next-update trigger is the standing +14-day ceiling. The AI-cloud ramp begins in October, then the Q3 print in November carries the first post-merger share count. The FOMC on 16 September and the Bank of Japan on 18 September both matter to a beta-3.4 name
The honest summary. This is an extraordinary business with a genuine moat — reusability is a real, unreplicated cost advantage — carrying US$60bn of net cash, growing 92%, and holding signed contracts for much of what it has promised. It is also loss-making, burning about US$18.4bn a quarter in capital expenditure, and priced at a multiple that is below the Expensive line on raw consensus and above it after the framework's mandatory haircut. The SELL is a price call resting on a convention, and this report says so rather than dressing it as certainty. It stays Stopped and visible, and it reactivates automatically as soon as a later report upgrades any horizon.
You would be risking about 15% to a holdable stop for a probability-weighted return of +0.2%.
What you are risking. A stop at US$129.00 is 2.89 average daily ranges away, and this name's derived beta is 3.40 with 93.4% annualised volatility[9] — roughly a fortnight of ordinary movement. The bear case is US$105.00, -31%, and it is not exotic: it is consensus revenue landing at the low end of a range someone credible is already forecasting.
What you are gaining. The best cost advantage in the industry, US$60.3bn of net cash, 92% revenue growth, and signed cloud contracts rather than forecast ones. Plus unpriced optionality on a Tesla combination that remains genuinely possible and genuinely unannounced.
The honest framing. On raw consensus this is not an expensive stock. After the framework's mandatory 25% growth haircut it is, narrowly. Buying here is a bet that the revenue ramp lands closer to consensus than to the haircut — which is a reasonable bet, and is not the one this framework is built to take.
Selling at US$151.21 means giving up a 31.6% gain since August and accepting that the framework's discipline disagrees with a 46%-upside analyst panel.
What you give up. An unreplicated moat, 92% growth, a guided US$100bn ARR exit rate, and a consensus target of US$220.68 from nineteen analysts of whom 85% are bullish[1]. This report is deliberately more cautious than the street, and the street may well be right.
What you avoid. A forward EV/EBITDA of 60.47× — two to four times large-cap technology, and the one measure here that depends on neither the growth haircut nor a stale share count. A 2.88% dilution that has already happened and is now in the denominator rather than ahead of it. And a beta-3.4 name in a cohort whose breadth the macro report is watching narrow.
What would bring it back. Not a lower price so much as a bigger denominator: haircut forward revenue above roughly US$96.6bn takes this under the 20× line by itself. It stays Stopped and visible and reactivates as soon as a later report upgrades any horizon.
{
"ticker": "SPCX",
"date": "2026-09-12",
"version": "v6",
"company": "Space Exploration Technologies Corp.",
"brand": "SpaceX",
"currency": "USD",
"exchange": "NASDAQ",
"exchange_ticker": "NASDAQ:SPCX",
"isin": "US84615Q1031",
"api_ticker": "SPCX",
"finder_ticker": "SPCX",
"finder_exchange": "NASDAQ",
"sector": "Industrials \u2014 Aerospace & Defense (launch, satellite broadband, AI)",
"gics_sector": "Industrials",
"prior_report": "calibration-SPCX-20260807-0752.json",
"report_file": "SPCX_Signal_v6_20260912_1700.html",
"analysis_status": "stopped",
"analysis_status_note": "Remains Stopped under SKILL Step -2 B3: this refresh returns no BUY on any horizon (short, medium and long all SELL). VISIBLE on the watchlist and auto-reactivates to 'donatien-pick' as soon as a later report upgrades any horizon.",
"price_at_rating": 151.21,
"price_at_rating_note": "Settled close of FRIDAY 2026-09-11. The run date 2026-09-12 is a SATURDAY. Confirmed by get_technical_indicators and get_multi_timeframe_analysis, both of which return 2026-09-11 as the latest daily bar at 151.21.",
"signal_short": "SELL",
"signal_medium": "SELL",
"signal_long": "SELL",
"primary_signal": "SELL",
"short_entry_confirmed": false,
"quality_score": 60,
"valuation_score": 34,
"timing_score": 54,
"driver_score": 48,
"driver_label": "Mixed",
"composite_short": 50,
"composite_medium": 49,
"composite_long": 51,
"lifecycle_stage": "high-growth",
"moat_score": 73,
"overall_confidence": 45,
"quality_confidence": 50,
"valuation_confidence": 45,
"timing_confidence": 45,
"economic_alignment_stance": "Neutral",
"economic_alignment_conviction": 45,
"economic_alignment_pressure": "Neutral",
"economic_alignment_source": "sector map (GICS Industrials -> XLI, N/N/N)",
"macro_report_date": "2026-09-09",
"shares_outstanding": 13572821625,
"shares_outstanding_provider": 13181779945,
"shares_reconciliation_note": "THE PROVIDER SHARE COUNT IS STALE AND THE CROSS-CHECK DOES NOT CATCH IT \u2014 recorded in full because this is the subtlest form of a trap already on this ticker's record. get_ticker_details returns 13,181,779,945 and provider market cap $1,992,068,956,160 divided by the $151.21 close gives 13,174,187,925 \u2014 a 0.06% difference, so shares x price reconciles to market cap and LOOKS clean. It is not. BOTH fields are stale in the SAME direction, so they agree with each other and disagree with reality. The 7 August calibration recorded the identical ~13.18bn count; it has not moved across a 391,041,680-share issuance on 14 AUGUST, when the Cursor merger closed (8-K, 14 Aug 2026: 389,289,254 shares for Cursor common and preferred plus 1,752,426 for vested restricted stock units). A reconciliation between two fields from the same stale snapshot proves only that the snapshot is internally consistent. THE CORRECTED COUNT IS 13,572,821,625, and every per-share and enterprise-value figure in this report is struck on it. A further 29,128,326 unvested Cursor restricted stock units and options were ASSUMED rather than issued, so they are dilution to come, not dilution already in the count. The income statement's weightedAverageShsOut of 5,864,000,000 (Q2) and 3,884,000,000 (Q1) are weighted averages spanning the pre-IPO period and are not a share count at all.",
"enterprise_value_usd": 1992045357212.25,
"enterprise_value_note": "RECONSTRUCTED on the CORRECTED share count: 13,572,821,625 x $151.21 = $2,052,346,357,916 market cap, + total debt $39,708,000,256 - cash $100,009,000,960 = $1,992,045,357,212. On the stale provider market cap the same reconstruction gives $1,931,767,955,456, which is $60.3bn too low. The provider's own enterpriseValue field of $3,926,132,785,152 is separately BROKEN and discarded: it exceeds market cap plus all debt, which is arithmetically impossible for a company holding $100bn of cash. The 7 August report discarded it on the same test.",
"val_multiple_basis": "warranted FORWARD EV/EBITDA, two-stage \u2014 the stream chosen deliberately. This company is pre-profit, so there is no P/E anchor; its forward EPS consensus is defective on share count; and its forward REVENUE multiple swings across the Expensive line depending on the growth haircut. EBITDA is the one consensus line on this name struck in DOLLARS and therefore immune to the share-count defect, and it is the stream a capital-intensive infrastructure business is actually valued on. BOTH SIDES FORWARD: warranted is the forward form of the two-stage sum, actual is enterprise value over FY2027 consensus EBITDA. Neither side is trailing and the two are never mixed.",
"warranted_multiple": 15.06,
"actual_multiple": 60.47,
"warranted_ratio": 4.02,
"val_band": "expensive",
"threshold_fragility_na_reason": "derive_fragility.py is a P/NAV instrument \u2014 it expects a per-share NAV and a band ladder in NAV units, and returns the distance to the next band. This report's anchor is a forward EV/EBITDA multiple, so there is no NAV to move and the script has no defined output (it raises on nav_to_better). It was run and it failed; it was NOT coerced by inventing NAV-shaped inputs, which would have produced a derived-looking number about a model that does not exist. The sensitivity that matters is published instead, and every figure in it is computed rather than typed \u2014 see threshold_fragility_derived.",
"threshold_fragility_derived": {
"quantity": "warranted ratio (actual forward EV/EBITDA / warranted forward EV/EBITDA)",
"published": 4.015,
"expensive_boundary": 1.4,
"distance_to_boundary_pct": 186.8,
"price_at_boundary_usd": 55.63,
"by_growth_bucket": {
"g_near 10%": {
"warranted": 15.064,
"ratio": 4.014,
"band": "expensive"
},
"g_near 15%": {
"warranted": 17.554,
"ratio": 3.445,
"band": "expensive"
},
"g_near 20%": {
"warranted": 20.371,
"ratio": 2.968,
"band": "expensive"
}
},
"by_ebitda_range": {
"consensus low": {
"ebitda_usd_bn": 22.0,
"ratio": 6.012,
"band": "expensive"
},
"consensus mean": {
"ebitda_usd_bn": 32.94,
"ratio": 4.014,
"band": "expensive"
},
"consensus high": {
"ebitda_usd_bn": 44.28,
"ratio": 2.987,
"band": "expensive"
}
},
"reading": "The published ratio is 4.02x against a 1.40x boundary, so it would have to fall 65% to leave the Expensive band \u2014 a share price near $55.63. It stays Expensive at EVERY growth bucket the framework offers and at every point of the FY2027 EBITDA consensus range INCLUDING THE HIGH END, where the ratio is still 2.99x. The valuation band is therefore NOT the fragile part of this report. QUALITY IS: at 60 it is 5 points below the High band, and at 65 the matrix row would not reach a sell. That is where a reader should push."
},
"anchor_inputs_note": "r = 11.28% = the 9 Sep macro report's 4.78% ten-year + the framework's fixed 4.5% equity risk premium + a 2.0-point risk add-on, which the DERIVED beta of 3.40 earns several times over. g_near = 10%, the secular-growth bucket cap \u2014 the top bucket is used rather than the 10% cyclical one because the near-term growth is CONTRACTED rather than forecast (signed Google and Anthropic cloud agreements plus $6.7bn booked in Q3). g_term = 3%. Two-stage warranted trailing 16.571x, forward 15.064x. SENSITIVITY, and the conclusion is invariant: at a 10% g_near the warranted multiple is 15.06x and the ratio 4.01x; at the most generous 20% it is 20.37x and the ratio 2.97x. Expensive on every one.",
"val_band_basis_note": "Expensive on the RATIO arm: 60.47x actual against 15.06x warranted is 4.02x, far above the 1.40 line. This replaces an earlier draft of this analysis that set warranted_multiple to the 20x EV/Revenue guardrail LINE and read the ratio off that. That was a category error and it is recorded so it is not repeated: the guardrail is a rich-line CAP used to test an actual multiple, not a warranted level, and the band thresholds (1.40 Expensive, 1.20 Full) are defined against a DCF output. Sitting at 1.15x a RICH line is not the same thing as sitting at 1.15x a FAIR value, and treating them alike understated the name by a wide margin.",
"forward_ev_revenue_raw": 19.14,
"forward_ev_revenue_haircut": 23.76,
"forward_ev_revenue_consensus_low": 28.66,
"trailing_ev_revenue": 86.45,
"forward_ev_ebitda": 60.47,
"forward_ev_ebitda_note": "THE CROSS-CHECK THAT DOES NOT DEPEND ON THE HAIRCUT, and the reason the Expensive call is not fragile. FY2027 consensus EBITDA of $32.94bn (n=16) gives 60.47x forward EV/EBITDA. EBITDA is a consensus line struck in dollars, so unlike the EPS strip it cannot be corrupted by a wrong share count. At a 31.7% implied margin, 19.14x revenue and 60.47x EBITDA are the SAME statement \u2014 the 20x revenue guardrail is calibrated for high-margin software, where 20x revenue is roughly 50x EBITDA, so this name sits right at the boundary the guardrail describes and above it on the earnings measure.",
"forward_eps_defect_note": "THE PROVIDER'S FORWARD P/E IS WRONG AND IS NOT USED. get_yahoo_quote returns eps_forward 1.54984 and forward_pe 97.56. Reconciling the consensus EPS strip against the consensus NET INCOME strip on the actual 13,572,821,625 share count fails in every year: FY2027 epsAvg 1.63 x shares = $21.5bn against a netIncomeAvg of $5.03bn; FY2028 4.19864 x shares = $55.3bn against $11.54bn; FY2025 -2.94468 x shares = -$38.8bn against -$9.36bn. Dividing each year's net income by its EPS implies a share base of 3.09bn, 2.75bn and 3.18bn respectively \u2014 i.e. the EPS consensus is struck on a PRE-IPO share count of roughly 3bn, not on 13.18bn. Restating FY2027 to the true count gives EPS $0.3708 and a forward P/E of 408x, against the provider's 97.56x. The name is about four times more expensive on earnings than the provider field says. This is the recorded ABXX.TO defect class caught before publication, and it is why no P/E-based figure appears in this report's anchor.",
"eps_trailing": -1.1,
"trailing_pe": null,
"trailing_pe_na_reason": "No trailing P/E exists: the company is loss-making (Q2 FY2026 net loss $541m, Q1 FY2026 net loss $4,276m). Any vendor P/E on a negative earnings base is an artifact and is not quoted.",
"clean_pe": null,
"clean_peg": null,
"clean_pe_na_reason": "Negative trailing earnings; no clean P/E or PEG is computable.",
"nonop_pct_of_net_income": 47,
"nonop_note": "Q2 non-operating lines are LOSSES, not inflating gains (nonOperatingIncomeExcludingInterest -$254m, net interest -$289m), so reported and clean earnings are both negative and the earnings-quality inflation adjustment does not apply. Recorded because the arm must be adjudicated, not skipped.",
"valuation_lenses": [
{
"lens": "Warranted forward EV/EBITDA (the anchor)",
"weight": 0.4,
"score": 14,
"state": "available",
"value": "60.47x actual vs 15.06x warranted = 4.02x",
"note": "Two-stage, both sides forward, on the only consensus line immune to the share-count defect. Sub-score mapped as 40 x (1.40 / ratio), so a ratio of exactly 1.40 scores 40 \u2014 the top of the Expensive band."
},
{
"lens": "Forward EV/Revenue vs the sector guardrail",
"weight": 0.2,
"score": 45,
"state": "available",
"value": "19.14x raw / 23.76x after the 25% growth haircut, vs a 20x line",
"note": "Scored at 45 \u2014 a FAIR reading, and the one lens on this page that disagrees with the verdict, which is why it is shown rather than dropped. On raw FY2027 consensus revenue the name sits BELOW the 20x line; only after the framework's mandatory 25% haircut does it cross. A reader should know one of the five lenses says this is not expensive."
},
{
"lens": "Own-history decile",
"weight": 0.15,
"score": null,
"state": "unavailable",
"value": "n/a \u2014 3 months of listed history",
"note": "DECLARED UNAVAILABLE WITH ITS REASON rather than silently skipped. SPCX listed on 12 June 2026, so there are about 63 trading sessions and no pre-IPO market price. A multiple series cannot be constructed, and a 3-month window that consists of one drawdown and one rally carries no information about whether today's multiple is high or low for this name."
},
{
"lens": "PEG",
"weight": 0.1,
"score": null,
"state": "unavailable",
"value": "n/a \u2014 no positive earnings, and the forward EPS strip is defective",
"note": "DECLARED UNAVAILABLE WITH ITS REASON. Trailing earnings are negative, so no trailing PEG exists; and the forward EPS consensus fails a share-count reconciliation by a factor of about four (see forward_eps_defect_note), so a forward PEG built on it would be wrong in the flattering direction."
},
{
"lens": "Analyst consensus",
"weight": 0.15,
"score": 72,
"state": "available",
"value": "mean target $220.68 (+45.9%), 80.6% bullish on the n=19 target panel, consensus Buy",
"note": "Genuinely positive and the strongest bull input on the page. Two caveats recorded rather than buried: the upside has COMPRESSED from +85% at the last report to +46%, and this is the same panel whose EPS strip fails the share-count reconciliation \u2014 which is why valuation_confidence is 45 rather than higher."
}
],
"valuation_blend_note": "Two of the five lenses are unavailable and both are declared with reasons, so the blend runs over the available 0.75 of weight rescaled to 1.0: 33.87 -> 34. Nothing was silently skipped. SENSITIVITY: re-run the anchor at the most generous near-term growth the framework allows (20% rather than 15%) and the sub-score rises to 18.9, taking the blend to 36.5 \u2014 still inside the Expensive band, so the pillar band does not turn on the growth cap. The one genuinely dissenting input is the EV/Revenue lens, scored 45: on RAW FY2027 consensus revenue this name sits BELOW the 20x guardrail line and only crosses after the framework's mandatory 25% growth haircut. That lens carries 20% of the weight and it is published at its dissenting value rather than nudged into line.",
"quality_build": {
"weights": {
"universal": 0.4,
"moat": 0.25,
"roic_capital_allocation": 0.2,
"industry_benchmark": 0.15
},
"universal_sub_signals": {
"Revenue trajectory": 92,
"Profitability vs peers": 42,
"Cash generation": 30,
"Balance-sheet health": 88
},
"universal_average": 63.0,
"moat_dimensions": {
"Pricing power": 58,
"Network effects": 60,
"Switching costs": 70,
"Cost advantage": 95,
"Intangible assets": 85
},
"moat_score": 73.6,
"roic_sub_signal": {
"roic_percentile_vs_peers": 22,
"capital_allocation": 48,
"management_skin_in_game": 60,
"weights": {
"roic": 0.4,
"capital_allocation": 0.3,
"skin_in_game": 0.3
},
"score": 41.2
},
"industry_benchmark_score": 52,
"weighted_sum": 59.64,
"score": 60,
"max_component": 73.6,
"moat_derivation_note": "RE-DERIVED from the five dimensions this run, and the re-derivation CHANGED it \u2014 which it did not in an earlier draft of this report, and that was the defect. That draft reproduced the inherited 78.0 exactly while the quarter's own pricing metric moved against it, unmentioned: STARLINK ARPU FELL 22.4% YEAR ON YEAR, from about $85 to about $66 (10-Q, Q2 FY2026). Pricing power is cut 80 -> 58. The 80-100 band requires proven repeated price increases absorbed by the market, and the only evidence offered for it was LAUNCH pricing \u2014 the Space segment, $962m of $7,814m, or 12% of revenue. Connectivity is $4,291m, 55%, and its per-user pricing is falling. A pricing-power score cannot rest on 12% of revenue while the majority line moves the other way. Network effects 60 (Starlink improves with constellation density, which is scale rather than a true network effect, though X/Grok carries a real one); switching costs 70 (multi-year government and enterprise launch contracts, consumer hardware cost); cost advantage 95 (reusability, and nobody has replicated it at scale \u2014 this is the moat); intangible assets 85 (spectrum and orbital-slot allocations plus NASA/DoD certification are genuine regulatory barriers). Average 73.6, against the 78.0 the August report carried; Quality falls with it, and the verdict does not move.",
"capital_allocation_note": "capital_allocation = 48 is a NEW discretionary input this run; the 7 August calibration published roic_percentile and skin_in_game but no capital-allocation score, so it is justified here rather than assumed. Against: sustained reinvestment at a negative ROIC (capex around $18.4bn a quarter), and the xAI acquisition was a RELATED-PARTY purchase from the chief executive, which is a governance mark against. For: the June IPO was well timed and well priced, the balance sheet carries $60.3bn of net cash, and the Cursor deal at roughly 15x its $4bn annualised revenue is struck BELOW this company's own revenue multiple, which makes it accretive on that measure even while dilutive on share count. Net: mixed, marginally below average. Sensitivity: at 45 the pillar reads 59.46, at 55 it reads 60.06 \u2014 Medium on both, so this input does not decide the verdict.",
"sensitivity_note": "QUALITY IS VERDICT-DECIDING HERE and, unlike a name whose components all sit below the band, the boundary IS reachable. That is stated up front. At 60 the matrix reads Medium x Expensive -> SELL; at 65 it would read High x Expensive -> HOLD. The largest component is the MOAT at 73.6, so the supremum over all weightings is 73.6 and weightings clearing 65 plainly exist. NO ROBUSTNESS CLAIM IS MADE. What reaching 65 actually requires: shifting about 16.5 percentage points of weight out of ROIC-and-capital-allocation into the moat, which would leave the ROIC leg at roughly 3.5% \u2014 not a defensible weighting, since it would discard almost all of the returns evidence on a company that does not yet earn a positive return. Within the SKILL's own sanctioned latitude (industry benchmark 15-20%) the score moves by under a point. The WITHIN-component sensitivity is safe in both directions: profitability 42 -> 55 gives 60.94 and cash generation 30 -> 45 gives 61.14, both still Medium."
},
"quality_detail": {
"industry_benchmark_name": "ROIC vs WACC + backlog growth (Industrials)",
"industry_benchmark_value": "ROIC still negative; Q2 operating margin -1.8% against -41.6% in Q1",
"industry_benchmark_score": 52,
"moat_score": 73.6,
"roic_percentile_vs_peers": 22,
"capital_allocation": 48,
"management_skin_in_game": 60,
"band_note": "Quality 60 is 5 points below the High band; see quality_build.sensitivity_note."
},
"revenue_q2_usd": 7814000000,
"revenue_q1_usd": 4694000000,
"revenue_growth_yoy_pct": 91.9,
"operating_income_q2_usd": -143000000,
"net_income_q2_usd": -541000000,
"ebitda_q2_reported_usd": 111000000,
"adjusted_ebitda_q2_usd": 3500000000,
"operations_note": "Q2 FY2026, the latest reported quarter and the same one the 7 August report used: revenue $7.814bn (+92% YoY), adjusted EBITDA $3.5bn (+191%), net loss narrowed to $541m, connectivity revenue $4.3bn (+66%) on a Starlink base that ended the quarter at 12 million subscribers. Operating margin -1.8% against -41.6% in Q1 \u2014 the loss is narrowing fast, but it is still a loss.",
"guidance_note": "On the 4 August call the CFO said the company is on pace for $100bn of ANNUALISED RECURRING REVENUE by December, driven largely by AI cloud-services agreements including closed deals with Google and Anthropic that begin ramping in Q4, and disclosed a further $6.7bn of cloud-services revenue contracted in the first weeks of Q3 over a six-month period beginning in October. This matters for an honest reading of the consensus: an earlier draft of this analysis judged the FY2026 consensus revenue of $43.3bn implausible against $12.5bn of H1 actuals. That judgement was WRONG and is withdrawn \u2014 the H2 ramp is management's own guidance on contracted revenue, and an ARR exit rate is not the same quantity as full-year recognised revenue.",
"cursor_acquisition_usd": 60000000000,
"cursor_shares_issued": 391041680,
"cursor_rsus_assumed": 29128326,
"cursor_dilution_pct": 2.88,
"cursor_note": "CLOSED, NOT PENDING \u2014 and an earlier draft of this report had it the other way round, which is the single worst error this run made. The all-stock acquisition of Anysphere, the parent of the Cursor AI coding tool, at a $60bn valuation BECAME EFFECTIVE ON 14 AUGUST 2026 per the company's 8-K: 'on August 14, 2026 (the Effective Time), the Merger became effective'. It issued 389,289,254 Class A shares for Cursor's common and preferred stock, struck on a seven-day volume-weighted average before closing, plus 1,752,426 shares for vested restricted stock units \u2014 391,041,680 shares in total, or 2.88% of the corrected count. A further 29,128,326 unvested restricted stock units and options were assumed, which is dilution still to come. The earlier draft described the deal as 'expected to close in Q3, subject to regulatory approval', derived a share count from the 12 September price rather than the August VWAP, and built its next-update date around a closing that had already happened four weeks earlier. It was caught by the independent source auditor reading the 8-K. Cursor's annualised revenue was about $4bn by early June, so the deal was struck near 15x revenue \u2014 below this company's own multiple on every basis in this report.",
"lockup_note": "The staggered post-IPO lockup has now largely run, and the record is more mixed than an earlier draft of this report claimed. Tranches: about 911.5m shares on 6 AUGUST (day 55), a second release on 20 AUGUST (day 70), then 319.0m on 9 SEPTEMBER (day 90) and a further 59.1m of Rule 144 affiliate stock on 10 SEPTEMBER (day 91). That leaves the free float near 17%, which only reconciles once the 9 September tranche is counted. CORRECTION TO THE EARLIER DRAFT: it recorded the second tranche as 12 August, recorded one September release rather than two, and said the stock 'rose through all three'. Reporting has the stock CLOSING ABOUT 4% LOWER on the second expiry day, briefly below its $135 IPO price, before rebounding the next session. AND THE INTER-EXPIRY GAINS DO NOT CHECK OUT EITHER: an earlier draft said roughly 29% and 14.5% 'do check out', but on this report's own price source they are +16.6% (6 Aug close 114.92 -> 20 Aug 134.00) and +10.1% (20 Aug -> 9 Sep 147.55, or +10.6% to 10 Sep 148.18). The aggregator figures are contradicted by the tape, and an assertion that a check was performed is the one thing that must not be wrong. The honest read is that supply has been absorbed without a durable break, not that every expiry was shrugged off. Sourcing on the tranche calendar is aggregator-grade rather than a filed schedule.",
"tesla_merger_note": "STILL RUMOUR, and re-verified this run rather than carried. On the 22 July Tesla earnings call Musk declined to confirm a combination, saying there would be 'more and more overlap' but that 'we can't talk about combining companies and that kind of thing on an earnings call \u2014 it's got to be done with the appropriate process'. The SpaceX president had hinted at it around the June IPO. There is NO announced transaction, no agreed terms and no dated decision, so Gate 5's takeover machinery does not apply and val_band is computed normally. It remains a Binary-Event caution and scenario optionality, exactly as the prior report framed it.",
"val_guardrail_na_reason": "The per-sector guardrail line is a P/E-family test and this name is anchored on EV/EBITDA, so the guardrail arm genuinely does not apply to the anchor \u2014 a CONFIRMED inapplicability, not an undeclared skip. It is not simply dropped either: the framework's EV/Revenue line for a pre-profit technology name (20x) IS tested, as the 20%-weighted cross-check lens, and the name reads 19.14x on raw FY2027 consensus revenue \u2014 BELOW the line. So the guardrail floor does NOT fire on this name on any basis, and no part of the Expensive call rests on it. AN EARLIER DRAFT OF THIS REPORT APPLIED THE 25% GROWTH HAIRCUT TO THE MULTIPLE'S DENOMINATOR and tested 23.76x against the line. That is withdrawn: the framework's haircut disciplines g INSIDE the warranted multiple, while the guardrail floor tests the ACTUAL multiple, and a haircut actual multiple is not a framework object. The haircut now sits where it belongs, in g_near. Note also that the sector string resolves to the Industrials 23.0x P/E line, which would be the wrong yardstick for either an EBITDA or a revenue multiple.",
"relative_strength_vs_spy": -9.1,
"relative_strength_vs_sector": -3.9,
"relative_strength_note": "Measured from the 12 June listing, because there is no earlier price: SPCX -6.05% against SPY +3.04% (-9.1pp) and XLI -2.16% (-3.9pp) \u2014 BEHIND both since the IPO. Over one month the picture inverts sharply: SPCX +13.44% against SPY -0.81% (+14.3pp) and XLI -7.18% (+20.6pp), so it is a strong recent leader inside a weak since-listing record. Both windows are given because either alone misleads. Raw unadjusted closes; SPCX pays no dividend.",
"relative_strength_1m_vs_spy": 14.3,
"relative_strength_1m_vs_sector": 20.6,
"trailing_pe_reported_note": "trailing_pe, clean_pe and clean_peg are ALL NULL and each carries its own reason field. The company is loss-making, so every one of them would be an artifact. Recorded explicitly because a null and a not-computed must not read the same.",
"range_position_52w_pct": 38.4,
"swing_lows": [
104.83,
105.11,
107.01,
130.39,
138.17
],
"realized_beta_vs_spy": 3.4,
"realized_beta_note": "DERIVED, because the provider returns a null beta for a name with three months of history. Regressing 62 daily SPCX returns against SPY from 12 Jun to 11 Sep 2026 gives a realized beta of 3.40, with annualised volatility of 93.4% against SPY's 12.2% \u2014 a 7.65x volatility ratio. Recorded as the evidence for treating this as a top-risk-tier name; it also means the standard risk add-on tier would apply comfortably if a discount-rate anchor were being computed.",
"timing_detail": {
"mtf_confluence": 62,
"mtf_confluence_label": "Bullish on the only real timeframe",
"risk_reward_score": 32,
"macro_score": 38,
"sentiment_score": 72,
"catalyst_clustering_score": 60,
"dynamic_macro_weight": 0.15,
"sentiment_weight": 0.18,
"catalyst_weight": 0.17,
"rsi_daily": 60.27,
"atr_daily": 7.696,
"sma20_daily": 142.99,
"sma50_daily": 135.07,
"macd_daily": 5.287,
"macd_signal_daily": 4.906,
"stop_distance_usd": 22.21,
"stop_distance_atr": 2.89,
"formula": "timing = mtf 62 x 0.30 + risk-reward 32 x 0.20 + macro 38 x 0.15 + sentiment 72 x 0.18 + catalyst 60 x 0.17 = 53.86 -> 54.",
"synthetic_series_note": "MOST OF THE MULTI-TIMEFRAME TABLE IS BACK-FILLED AND WAS NOT SCORED. SPCX listed on 12 June 2026, about 63 trading sessions ago, yet the provider returns 60 MONTHLY bars and 149 WEEKLY bars, and a daily 200-day average of 60.87 against a 151.21 price. Those series cannot be real. Only the daily timeframe and its 20- and 50-day averages are usable, and the trend term is scored on those alone \u2014 which is why it reads 62 rather than the 90-plus a five-timeframe 'strongly bullish' confluence reading would imply. The 7 August report reached the same conclusion and discarded the same series. Hourly and 15-minute volume ratios of 0.08 and 0.06 are SATURDAY artefacts and were also excluded.",
"note": "On the real data the tape is genuinely strong: $151.21 sits above the 20-day $142.99 and the 50-day $135.07, RSI is 60.27, and MACD at 5.287 is above its 4.906 signal. That is a marked improvement on the prior report's 44. It is held back by risk-reward: the nearest meaningful swing-low shelf is 130.39, so a logical stop sits at $129.0, which is $22.21 away or 2.89 ATR \u2014 beyond the 2.5-ATR line the framework calls unfavourable. On a name with a 93% annualised volatility, a stop you can actually hold is a long way down."
},
"analyst_consensus_target": 220.68,
"analyst_target_high": 450.0,
"analyst_target_low": 117.0,
"analyst_target_median": 220.0,
"analyst_target_upside_pct": 45.9,
"analyst_grades_consensus": "Buy",
"analyst_bullish_pct": 80.6,
"analyst_coverage_count": 19,
"analyst_panel_note": "THE TWO PANELS DISAGREE AND BOTH ARE REPORTED, because an earlier draft published '85% bullish' which appears in NO declared source and matches the 7 August calibration \u2014 i.e. it was carried, not re-pulled. The figures that do exist: get_yahoo_analyst_targets returns n=19 with a distribution of 6 strong buy / 23 buy / 5 hold / 2 sell (36 actions, 80.6% bullish) and a mean target of $220.68; get_grades_consensus returns a separate, smaller panel of 9 buy / 1 hold (90% bullish, consensus Buy). 85% is reachable only by dropping the two sells from the Yahoo distribution. The headline bullish percentage here is the Yahoo one at 80.6%, and the coverage count of 19 is the target panel.",
"recent_upgrades_30d": 0,
"recent_downgrades_30d": 0,
"fmp_rating": "C-",
"fmp_overall_score": 1,
"fair_value_est": null,
"fair_value_na_reason": "No point fair value is published. The company is loss-making, the forward EPS consensus is defective on share count, and the revenue consensus spans $69.5bn to $139.9bn for FY2027 \u2014 a better-than-two-to-one spread. Publishing a single number over that dispersion would imply a precision that does not exist. The scenario band below is the honest form of the same answer.",
"stop_loss": 129.0,
"target_price": 150.0,
"scenario_bull_target": 210.0,
"scenario_base_target": 150.0,
"scenario_bear_target": 105.0,
"scenario_probabilities": {
"bull": 25,
"base": 45,
"bear": 30
},
"scenario_weighted_target": 151.5,
"scenario_note": "Probability-weighted $151.50 against $151.21 is +0.2% \u2014 essentially flat, and that is the honest shape of this name rather than a hedge. The framework says the price is ahead of the value; the distribution says the optionality is real enough to offset it. A SELL here is a statement about price, not a forecast of collapse.",
"entry_groups_met": 0,
"entry_conviction": "Wait",
"exit_groups_live": 0,
"exit_action": "Hold",
"ladder_delta_note": "The entry conviction and entry groups met both sit where they sat in August: Wait and 0 of 3. The Fundamental group cannot open because no fair value is published; the Technical group fails its volume condition; the Catalyst group needs a post-earnings move and the next print is in November.",
"hard_gate_state": "caution",
"gates_triggered": [
"Valuation Ceiling (Gate 3) \u2014 TRIGGERED ON THE RATIO ARM. Forward EV/EBITDA of 60.47x against a warranted 15.06x is 4.02x, far above the 1.40x Expensive line. The GUARDRAIL arm does NOT fire and is not claimed: actual forward EV/REVENUE is 19.14x, BELOW the 20x line, and the guardrail is in any case a P/E-family test that cannot apply to an EBITDA anchor. Gates cap and never raise, so this leaves the SELL standing."
],
"gates_caution": [
"Dilution \u2014 CARRIED FORWARD from the 7 August report, which triggered an Accounting/Dilution gate. Re-adjudicated on this run's facts and DOWNGRADED to a caution rather than dropped. The $60bn all-stock Cursor acquisition has ALREADY CLOSED, on 14 August, issuing 391,041,680 shares or 2.88% \u2014 so this is dilution that has HAPPENED and is now in the denominator of every multiple in this report, not dilution to come. A further 29,128,326 assumed restricted stock units are still ahead. The xAI combination earlier in 2026 was also stock-funded. But the framework's Gate-4 test is share-count growth above 5% a year for two or more years, and a single 3% issue does not meet it; nor does it cap a signal that is already SELL. The lockup tranches are NOT dilution \u2014 they release existing shares to trade, they do not create new ones, and conflating the two would double-count the supply story.",
"Binary event \u2014 a SpaceX/Tesla combination remains rumoured and unannounced (see tesla_merger_note), and a nationalisation or national-security-control tail exists for any dominant launch provider. Neither is dated, so Gate 5 does not fire, but both are real sources of gap risk in a name with a 93% annualised volatility.",
"Pre-profit with heavy capital intensity \u2014 the company is loss-making on a GAAP basis and reinvesting at roughly $18.4bn of capex a quarter. Financial distress does NOT fire: cash is $100.0bn against $39.7bn of total debt, so the position is $60.3bn net cash and liquidity is not the question. The question is when the reinvestment turns into a return, and the answer is not yet visible in the reported numbers.",
"Data quality \u2014 two provider fields on this name are demonstrably broken and both are discarded with their arithmetic shown: the enterprise-value field exceeds market cap plus all debt, and the forward EPS consensus is struck on a roughly 3bn pre-IPO share base rather than the corrected 13.573bn. A third, the multi-timeframe monthly, weekly AND daily series, every one of which is back-filled through a period before the June listing (60, 149 and 204 bars against 63 real sessions). Recorded as a gate-level caution because a reader checking this name against a data terminal will see different numbers, and should know which ones to trust."
],
"do_not_buy_triggers": [],
"dnb_adjudication": "TRIGGER 2, ABSOLUTE ARM, ADJUDICATED IN FULL \u2014 this is the closest call in the report and it turns on the carve-out, not on the arithmetic. ARM (a) deep-expensive fires at 2.0x the warranted multiple with NO exceptional, proven, durable growth. The arithmetic bar IS CLEARED: the ratio is 4.02x, well above 2.0x, and it stays above on every growth assumption tested (4.01x at a 10% g_near, 2.97x at 20%). So the whole question is the carve-out, and the settled reading of that clause is growth IN THE EARNINGS STREAM THE MULTIPLE CAPITALISES \u2014 the growth that excuses a breach must be the same quantity the breached multiple was built from. This anchor is built on EBITDA. SpaceX's adjusted EBITDA grew +191% YEAR ON YEAR to $3.5bn in Q2 \u2014 REPORTED, not forecast \u2014 and the consensus strip carries EBITDA from -$0.71bn in FY2025 to $16.53bn in FY2026 to $32.94bn in FY2027. That is exceptional, it is proven in the reported quarter, and it is durable in the sense the clause requires: the near-term step is CONTRACTED, through signed Google and Anthropic agreements plus $6.7bn booked in the first weeks of Q3. THE CARVE-OUT APPLIES and arm (a) does NOT fire. This places the name on the FIX / STRL / PLTR side of the precedent line, where growth converts into the capitalised stream, and not on the RGTI / POWL / SE side, where it does not \u2014 SE failed this test at +7% earnings growth while its revenue ran +48%; the opposite shape to this. Recorded openly: the carve-out has no magnitude bound, which is a known framework gap rather than a finding about this company, and it is doing real work here at 4.02x. It is not load-bearing for the published signal, because a Do-Not-Buy would only harden a SELL that already stands. THE GUARDRAIL ROUTE into arm (a) also fails: it needs the actual multiple at 1.5x the sector line, and on the EV/Revenue cross-check the name is 19.14x on raw consensus \u2014 BELOW the 20x line entirely \u2014 and 23.76x after the haircut, which is 1.15x the line, nowhere near 1.5x. ARM (b) expensive-plus-live-catalyst does not fire either: it requires a currently ARMED and TRIGGERING systemic tail, and the 9 September macro report describes AI leadership as 'wobbling rather than breaking', with QQQ and XLK both still above their 200-day averages. Narrowing breadth is a watch item and belongs in the bear case, which is where this report puts it; it is not a live de-rating catalyst, and the standing precedent caps Expensive AI-cohort names at HOLD on a not-triggering tail rather than issuing a DO NOT BUY. Trigger 1 leverage: no \u2014 $60.3bn NET CASH. Trigger 3 negative revisions: no \u2014 estimates are rising sharply. Trigger 4 insider selling: no reported C-suite cluster, and lockup expiries are not insider sales. Trigger 5 structural threat: none identified. NO TRIGGER FIRES.",
"signal_derivation": "Base Decision Matrix: Quality 60 = Medium (40-64), Valuation 34 = Expensive (<40) -> the 'Medium | Expensive | Any' row -> SELL at every horizon; Timing 54 (Neutral) does not enter that row. Amplification: a SELL intensifies to STRONG SELL only when the driver is <=35 AND economic pressure is Headwind \u2014 the driver is 48 and pressure is Neutral, so neither condition holds. The Short technical-confirmation cap and the quality-starter override are unreachable from a SELL base. Hard gates cap and never raise, so the Valuation Ceiling leaves SELL standing. No Do-Not-Buy trigger fires (see dnb_adjudication). THIS IS A DOWNGRADE FROM THE PRIOR RUN, and the cause is specific and mechanical: August scored Quality 68, which falls in the High band, whose matrix row does not reach a sell; this run's published build returns 60, which falls in the Medium band, whose row does. The valuation band reads Expensive in both runs. So the entire signal change is the Quality band, and the Quality build is published in full with its sensitivity because of that.",
"next_update_date": "2026-09-26",
"next_update_basis": "The standing +14-day ceiling, because NO company-specific dated event remains inside the window. An earlier draft set this to 30 September on the expectation that the Cursor acquisition would close in Q3 \u2014 it had already closed on 14 August, so that trigger was spent before it was written. What is left inside two weeks is macro rather than company: the FOMC on 16 September and the Bank of Japan on 18 September, and Industrials is a medium-macro-sensitivity sector, so neither earns an event-driven date under the scheduling rule. The genuinely decisive dates sit outside the ceiling: the AI-cloud revenue ramp beginning in October, and the Q3 print in early November which will carry the first post-merger share count and the first test of that ramp.",
"deltas_vs_prior": {
"prior_date": "2026-08-07",
"prior_price": 114.92,
"price_change_pct": 31.6,
"quality": "68 -> 60 (-8) \u2014 NOT A LIKE-FOR-LIKE DELTA",
"quality_delta_note": "NOT DECOMPOSABLE, and not to be read as a deterioration. The 7 August calibration published NO four-component build \u2014 only quality_detail, carrying moat 78, ROIC percentile 22, skin-in-game 60 and benchmark 52, every one IDENTICAL to this run. Under this run's build those inputs cannot reach 68 at any capital-allocation value (maximum about 63.9); reproducing 68 would need a universal-sub-signal average near 73-81 against this run's 63.0, on the same Q2 data. The gap is a change of METHOD, not of business fact, and no reported figure deteriorated. Two points of it ARE a genuine re-derivation \u2014 moat 78 -> 73.6 on the Starlink ARPU fall. The rest is that a build is now published where none was before.",
"valuation": "24 -> 34 (+10)",
"timing": "44 -> 54 (+10)",
"signal_short": "HOLD -> SELL",
"signal_medium": "HOLD -> SELL",
"signal_long": "HOLD -> SELL",
"val_band": "expensive -> expensive",
"analysis_status": "stopped -> stopped",
"gates_triggered": "none -> Gate 3 (Valuation Ceiling). NEWLY TRIGGERED this run and named here because the next run diffs against this field. The August report recorded its triggered gates as 'Valuation Ceiling' and 'Accounting/Dilution' in prose but not as numbered gates, so the machine-readable transition reads none -> [3]. The dilution gate is re-adjudicated this run and DOWNGRADED to a caution rather than dropped \u2014 see gates_caution.",
"note": "Price +31.6% to $151.21, and the signal is DOWNGRADED to SELL on all three horizons. The whole of that change is the Quality BAND: 68 falls in High, whose matrix row does not reach SELL, while this run's published build returns 60, which falls in Medium, whose row does. Valuation is Expensive in both runs. Timing IMPROVED +10 on a genuinely better tape. Two things got better and are recorded as such: the lockup overhang has passed three tranches with the stock rising through all of them, and the macro sector call for Industrials is Neutral rather than the prior run's Headwind read. Two things got worse: the price rose 31.6% while trailing revenue did not, and the analyst upside compressed from +85% to +46%."
},
"carried_verbatim": [
"competitive_share_trajectory",
"economic_study"
],
"competitive_share_trajectory": "gaining",
"competitive_threat_level": "moderate",
"competitive_note": "Gaining, and not marginally: SpaceX flies the large majority of global orbital mass to orbit, and Starlink ended Q2 with 12 million subscribers, roughly double a year earlier. Direct rivals are Blue Origin, Rocket Lab, ULA, Arianespace and a Chinese state complex on the launch side, and Amazon's Kuiper, OneWeb/Eutelsat and terrestrial carriers on broadband. In AI cloud it now competes with the hyperscalers it also sells to. Unchanged this run: no new competitive disclosure since the Q2 release.",
"sections": {
"data_sources": {
"items": [
{
"id": "quote",
"name": "get_yahoo_quote + get_ticker_details (SPCX), settled close Fri 11 Sep 2026"
},
{
"id": "q2",
"name": "SpaceX Q2 FY2026 results, filed 4 Aug 2026 (get_income_statement)"
},
{
"id": "call",
"name": "SpaceX Q2 FY2026 earnings call, 4 Aug 2026 \u2014 CFO Bret Johnsen and CEO Elon Musk"
},
{
"id": "estimates",
"name": "get_analyst_estimates (SPCX, annual) \u2014 FY2025 to FY2028 consensus strip"
},
{
"id": "cursor",
"name": "SpaceX / Anysphere (Cursor) $60bn all-stock acquisition announcement, June 2026"
},
{
"id": "lockup",
"name": "SPCX post-IPO lockup schedule and outcomes, 6 Aug / 12 Aug / 10 Sep 2026"
},
{
"id": "tesla",
"name": "Tesla Q2 2026 earnings call, 22 Jul 2026 \u2014 Musk on a possible combination"
},
{
"id": "mtf",
"name": "get_multi_timeframe_analysis + get_technical_indicators (SPCX), 11 Sep 2026"
},
{
"id": "beta",
"name": "Derived realized beta: 62 daily SPCX vs SPY returns, 12 Jun - 11 Sep 2026"
},
{
"id": "macro",
"name": "Donatien Macro-Economic report, 9 Sep 2026 (MacroDriver-state-20260909.json)"
},
{
"id": "filings",
"name": "SEC EDGAR filing index for CIK 0001181412, 7 Aug - 12 Sep 2026"
},
{
"id": "prior",
"name": "Donatien prior calibration, calibration-SPCX-20260807-0752.json (7 Aug 2026)"
}
]
}
},
"post_audit_fixes_round1": "BLOCKER: the Cursor acquisition was reported as pending and guided to close by 30 September; it had CLOSED on 14 August, issuing 391,041,680 shares. MAJOR: consequently the share count was stale at 13,181,779,945 and my reconciliation did not catch it, because the provider market cap is stale in the SAME direction so the two fields agreed with each other while disagreeing with reality. Corrected to 13,572,821,625 and every multiple restated. MAJOR: the filings note claimed no company-originated corporate action in the window; the merger 8-K sits inside it. MAJOR: lockup calendar corrected (second tranche 20 Aug not 12 Aug; two September releases; the stock FELL about 4 percent on the second expiry, it did not rise through all three). MAJOR: '85 percent bullish' appears in no declared source and was carried from August; replaced with the sourced 80.6 percent. MINORS: mass-to-orbit re-cited to the call; daily MTF frame disclosed as partly back-filled; ratio stabilised against rounding drift; Cursor share figures restated from the August VWAP count.",
"audit_route_note": "Round 1: a FRESH general-purpose agent (never a fork) \u2014 Layer 0.5 source auditor \u2014 census-only, never saw the report.. Verdict FAIL (1 blocker, 4 majors, 4 minors). Round 2: a FRESH general-purpose agent (never a fork) \u2014 Layer 2 independent auditor \u2014 a fresh agent that read the rendered report and re-ran every high-risk verification from scratch. It was told mid-run that the artifact had changed under it and re-read from disk.. Verdict FAIL (8 blockers, 6 majors, 2 minors).",
"delta_vs_prior": "vs calibration-SPCX-20260807-0752.json. signal_short HOLD -> SELL. signal_medium HOLD -> SELL. signal_long HOLD -> SELL. quality 68 -> 60. valuation 24 -> 34. timing 44 -> 54. driver set to 48 (absent in the prior report). economic alignment 42 -> 45. price 114.92 -> 151.21. warranted_ratio set to 4.02 (absent in the prior report). short_entry_confirmed set to false (absent in the prior report). next_update 2026-08-21 -> 2026-09-26. base target 120 -> 150. bull target 225 -> 210. bear target 70 -> 105. relative strength vs SPY negative but stabilising (bounced off new ATL $104.83) -> -9.1. relative strength vs sector set to -3.9 (absent in the prior report). moat 78 -> 73. overall confidence set to 45 (absent in the prior report). scenario weights set to {'bull': 25, 'base': 45, 'bear': 30} (absent in the prior report). gates_triggered 2 -> 1 (added Valuation Ceiling (Gate 3) \u2014 TRIGGERED ON THE RATIO ARM. Forward EV/EBITDA of 60.47x against a warranted 15.06x is 4.02x, far above the 1.40x Expensive line. The GUARDRAIL arm does NOT fire and is not claimed: actual forward EV/REVENUE is 19.14x, BELOW the 20x line, and the guardrail is in any case a P/E-family test that cannot apply to an EBITDA anchor. Gates cap and never raise, so this leaves the SELL standing.; removed Valuation Ceiling; Accounting/Dilution). gates_caution 2 -> 4 (added Dilution \u2014 CARRIED FORWARD from the 7 August report, which triggered an Accounting/Dilution gate. Re-adjudicated on this run's facts and DOWNGRADED to a caution rather than dropped. The $60bn all-stock Cursor acquisition has ALREADY CLOSED, on 14 August, issuing 391,041,680 shares or 2.88% \u2014 so this is dilution that has HAPPENED and is now in the denominator of every multiple in this report, not dilution to come. A further 29,128,326 assumed restricted stock units are still ahead. The xAI combination earlier in 2026 was also stock-funded. But the framework's Gate-4 test is share-count growth above 5% a year for two or more years, and a single 3% issue does not meet it; nor does it cap a signal that is already SELL. The lockup tranches are NOT dilution \u2014 they release existing shares to trade, they do not create new ones, and conflating the two would double-count the supply story.; Binary event \u2014 a SpaceX/Tesla combination remains rumoured and unannounced (see tesla_merger_note), and a nationalisation or national-security-control tail exists for any dominant launch provider. Neither is dated, so Gate 5 does not fire, but both are real sources of gap risk in a name with a 93% annualised volatility.; Pre-profit with heavy capital intensity \u2014 the company is loss-making on a GAAP basis and reinvesting at roughly $18.4bn of capex a quarter. Financial distress does NOT fire: cash is $100.0bn against $39.7bn of total debt, so the position is $60.3bn net cash and liquidity is not the question. The question is when the reinvestment turns into a return, and the answer is not yet visible in the reported numbers.; Data quality \u2014 two provider fields on this name are demonstrably broken and both are discarded with their arithmetic shown: the enterprise-value field exceeds market cap plus all debt, and the forward EPS consensus is struck on a roughly 3bn pre-IPO share base rather than the corrected 13.573bn. A third, the multi-timeframe monthly, weekly AND daily series, every one of which is back-filled through a period before the June listing (60, 149 and 204 bars against 63 real sessions). Recorded as a gate-level caution because a reader checking this name against a data terminal will see different numbers, and should know which ones to trust.; removed Binary Event (SpaceX/Tesla merger talks + staggered lockup overhang + nationalisation tail); Financial Distress (large capex burn, but ample liquidity)). Carried without change: val_band, entry_conviction, entry_groups_met, exit_action, hard_gate_state, analysis_status, industry benchmark score, competitive share trajectory, exit_groups_live. NOTE: Quality 61 to 60 and Valuation 35 to 34 on the two substantive re-scorings; Timing 54 and all three signals unchanged. The name is MORE expensive on every correction, not less. Composites 50/49/51.",
"post_audit_fixes_round2": "The verdict, the anchor arithmetic, the Quality build and the DNB withholding were all SUSTAINED. What failed was the published justification, and most of it was residue from an anchor design I abandoned mid-authoring. BLOCKERS: Gate 3's prose still said the ratio was 'below the 1.20 Full line' and that the Expensive call 'rests on the guardrail' \u2014 both written when the ratio was 1.15 on the abandoned guardrail-as-warranted design; Gate 3 is triggered on the RATIO arm and the guardrail arm does NOT fire, because actual forward EV/Revenue of 19.14x is BELOW the 20x line. The EV/EBITDA figure was labelled 'forward EV/Revenue' in five places including two machine fields. val_guardrail_na_reason carried pre-correction figures. The quality delta presented a METHOD change as a like-for-like fall: August published no four-component build, so 68 and 61 are not comparable and the difference is not decomposable. The framework's pre-profit anchor-SKIP clause was never adjudicated even though taking it produces a HOLD \u2014 now answered on the page, with the direction of the choice stated, since taking the anchor gives SELL and skipping it does not. The withdrawn lockup claim survived in the Changes box and the delta note. And the claim that the 29 percent and 14.5 percent inter-expiry gains 'do check out' was false \u2014 on this report's own price source they are +16.6 percent and +10.1 percent. SUBSTANTIVE RE-SCORINGS: moat pricing power cut 80 to 58 because the 80-100 band needs proven price increases and the only evidence was LAUNCH pricing, 12 percent of revenue, while Starlink ARPU \u2014 55 percent of revenue \u2014 FELL 22.4 percent year on year, unmentioned; moat 78 to 73.6, Quality 61 to 60. And g_near moved from the 15 percent secular bucket to the 10 percent cyclical bucket, because the name is classified Industrials and 'contracted rather than forecast' is not the bucket's criterion; warranted 17.55 to 15.06, ratio 3.45 to 4.01, i.e. MORE expensive. Also: the 25 percent haircut was being applied to the actual multiple's denominator, which is not a framework object \u2014 withdrawn, the haircut now sits in g_near where it belongs; the exit condition was keyed to the wrong arm and understated itself badly (about 55.63 dollars a share, not 131); threshold fragility is now published as a DERIVED block with an explicit reason why derive_fragility.py cannot apply to an EV/EBITDA anchor.",
"lint": {
"exit": 0,
"assertions": 71,
"errors": 0,
"warnings": 5,
"could_not_run": [
"CCY-PE-SKIPPED"
],
"at": "2026-09-12T16:44",
"report": "SPCX_Signal_v6_20260912_1700.html"
},
"probability_weighted_target": 151.5
}