Summary

Initiate at Hold ($123.66 target, −9.5%): heavy capex caps compute upside

InitiationBy Penmark Team
SPCXHold$123.66-9%
Page 1 of 29
Equity research | Initiation of coverage | United States
Technology & Aerospace | Coverage: Initiated | Published 15 November 2026, 07:00 ET
Penmark Team | research@penmarkresearch.com
Space Exploration Technologies Corp. (Nasdaq: SPCX)

Initiate at Hold ($123.66 target, −9.5%): heavy capex caps compute upside

Rating (Low conviction)
Initiation
Hold
12-month target
$123.66
Price, 13 Nov
$136.63
Price return*
−9.5%
Total return
−9.5%
Bull case
$233.82
Bear case
$10.97

We start coverage at Hold with a $123.66 target, a 9.5% drop from the $136.63 share price. SpaceX pairs launch and Starlink with Super Intelligence (SI), its artificial intelligence (AI) unit spanning graphics processing unit (GPU) cloud, Grok, Cursor and X. SI revenue per watt is the yearly sales earned per watt (W) of powered compute capacity. Shares price peak rents near $35/W and ignore nearly $429bn of two-year capital spending.

  • SI quadruples FY27E sales to about $196bn, nearly 9% above consensus. Expanding capacity to 8 gigawatts (GW) drives negative free cash flow (FCF) of about $110bn.
  • SI rents fade from about $31/W to $26/W as new GPU supply arrives. Fixed power costs leave FY28E adjusted earnings before interest, taxes, depreciation and amortisation (EBITDA) about 8% below consensus.
  • Space & Connectivity ($190.1bn discounted cash flow (DCF)) and a 25% bull weight prevent Sell. Connectivity reaches nearly $28bn at a 63% margin as third-generation (V3) Starship adds mobile capacity.

What the market is missing. The current price implies a weighted average cost of capital (WACC) near 9% or FY30E SI rents near $35/W, while 90-day contracts allow repricing.

Catalysts. Initial public offering (IPO) lock-up expiry on 8 Dec 2026 and Flight 15 come first. Late-February results test $100bn annualised recurring revenue (ARR) ahead of mid-2027 renewals.

Valuation and target. Sum-of-the-parts (SOTP) DCF gives an $88.49 base value today, near our $80.75 enterprise value (EV) check. Scenario weights of 10%, 65% and 25% yield our target.

Risks. Hold sits 0.5 percentage points (pp) above Sell, and five of six stress tests trigger Sell. Higher rates, a 1 GW delay or softer rents tip the call; chief executive officer (CEO) bull builds add about $15 per share.

Exhibit 1: Rapid AI expansion still leaves shares at 32.2x FY27E earnings per share (EPS)
Dec yr-end, $bnFY24AFY25AFY26EFY27EFY28E
Revenue14.0218.6749.41196.10343.80
Growth y/y34.9%33.2%164.6%296.9%75.3%
Adjusted EBITDA5.356.5826.40121.85210.06
Margin38.2%35.3%53.4%62.1%61.1%
Net income0.02(4.94)4.7260.4597.91
Diluted EPS ($)n.d.n.d.0.214.246.81
Growth y/yn.m.n.m.n.m.1,888%60.5%
Free cash flow(5.39)(13.95)(61.45)(110.37)(0.69)
At $136.63, 13 NovFY24AFY25AFY26EFY27EFY28E
P/En.m.n.m.639.9x32.2x20.1x
EV/EBITDA362.3x294.4x73.4x15.9x9.2x
P/Bn.d.n.d.10.4x7.3x5.1x
FCF yield−0.3%−0.7%−3.1%−5.6%0.0%
Dividend yield0.0%0.0%0.0%0.0%0.0%
ROEn.d.n.d.2.6%26.6%30.0%
Net debt / EBITDAn.d.n.d.−0.9x0.7x0.4x
P/E at $123.66 targetn.m.n.m.579.1x29.1x18.2x

Source: Form 424B4; Q3 FY26 Form 10-Q; Penmark Research model (cut-off 13 Nov 2026). Table shows price-to-earnings (P/E), price-to-book (P/B) and return on equity (ROE); pre-IPO rows marked not disclosed (n.d.) or not meaningful (n.m.).

Bull upside lifts our $98.63 base case to Hold, with 0.77x upside to downside

Exhibit 3: Weighted scenarios reach our $123.66 target (−9.5%), as only the bull case clears spot
$0$50$100$150$200$250Jul-26Jan-27Jul-27ActualEstimatesPrice $137Bull $234 (+71%)Base $99 (−28%)Bear $11 (−92%)Target $124 (−9%)

Source: yfinance; company filings; Penmark Research SOTP DCF model (11.46% cost of equity)

Scenarios and price target basis
Bull$233.82+71.1%25% prob.

Spot value is $209.78. Compute reaches 12.7 GW by end-FY27E as firm SI rents hold a 65% margin in FY30E.

Base$98.63−27.8%65% prob.

Rolls up from our spot base DCF. Compute reaches 8.0 GW in FY27E as SI rents fade toward $26/W.

Bear$10.97−92.0%10% prob.

Spot value is $9.84. Excess supply and buyer exits cut SI rents to $16/W as grid limits cap FY27E at 5.0 GW.

Price target (prob.-weighted)$123.66−9.5% return = Hold (spot $110.95)
Upside to downside0.77x+$97.19 up / −$125.66 down
Exhibit 4: Operating profit scales to $121.23bn by FY28E
Income stmt, $bnFY24AFY25AFY26EFY27EFY28E
Revenue14.0218.6749.41196.10343.80
Adjusted EBITDA5.356.5826.40121.85210.06
Ongoing SBC(0.78)(1.95)(3.73)(5.88)(8.59)
Acq. SBC & restr.(0.28)(0.53)(3.22)(7.50)(6.50)
Depr. & amort.(3.82)(6.70)(13.91)(42.87)(73.73)
Operating income0.47(2.59)5.5565.60121.23
Net fin. & tax(0.45)(2.35)(0.83)(5.15)(23.32)
Net income (GAAP)0.02(4.94)4.7260.4597.91
Exhibit 5: FCF troughs at −$110.37bn in FY27E
Cash flow, $bnFY24AFY25AFY26EFY27EFY28E
Operating cash flow5.786.7924.51127.30199.58
Capex & leases(11.16)(20.74)(85.57)(228.34)(200.26)
Spectrum cash paid0.000.00(0.39)(9.33)0.00
Free cash flow(5.39)(13.95)(61.45)(110.37)(0.69)
FCF ex prepaymentsn.d.n.d.(66.62)(116.61)(11.53)
GPU lease net drawn.d.n.d.0.0035.00(8.00)
New debt drawnn.d.n.d.0.0041.998.69
Buybacks & div.0.000.000.000.000.00
Exhibit 6: Gross debt & leases peak at $116.52bn
Balance sheet, $bnFY24AFY25AFY26EFY27EFY28E
Cash & securitiesn.d.n.d.63.3830.0030.00
PP&E, Space & Conn.n.d.n.d.29.3839.3451.20
PP&E, SI computen.d.n.d.92.80269.02384.39
Goodwill & spec.*n.d.n.d.57.2472.6169.93
Net assets incl. cashn.d.n.d.242.81410.97535.52
Gross debt & leasesn.d.n.d.38.84115.84116.52
Net cash / (debt)n.d.n.d.24.54(85.84)(86.52)
Shareholders’ equityn.d.n.d.185.01269.95382.96
Exhibit 7: SI compute scales to 12.5 GW by FY28E
Per share & KPIsFY24AFY25AFY26EFY27EFY28E
Diluted shares (m)n.d.n.d.14,11214,37414,374
Diluted EPS ($)n.d.n.d.0.214.246.81
FCF per share ($)n.d.n.d.(4.35)(7.75)(0.05)
Book value / sh ($)n.d.n.d.13.1118.7826.64
SI compute, end (GW)0.30.82.78.012.5
SI rev / watt ($/W)**n.d.n.d.27.530.529.3
Consumer subs (m)4.48.915.522.529.5
Consumer ARPU ($)91.081.067.564.061.0

Source: Form 424B4; Q3 FY26 Form 10-Q; Penmark Research model (cut-off 13 Nov 2026). Tables follow generally accepted accounting principles (GAAP) and show stock-based compensation (SBC), property, plant and equipment (PP&E), average revenue per user (ARPU) and *operating net working capital (NWC). **4Q26E run-rate.

Shares trade 78% above AI-compute peers despite a 62.1% FY27E margin

Exhibit 8: Adjusted EBITDA margin expands 26.9pp to 62.1% in FY27E
Growth & marginsFY24AFY25AFY26EFY27EFY28E
Growth y/y
Sales34.9%33.2%164.6%296.9%75.3%
Adjusted EBITDA40.0%23.1%301.0%361.5%72.4%
Operating incomen.m.n.m.n.m.1,083%84.8%
Diluted EPSn.d.n.d.n.m.1,888%60.5%
Dividend per share0.0%0.0%0.0%0.0%0.0%
Margins & coverage
GAAP gross mgn.*42.9%49.4%n.d.n.d.n.d.
Adj. EBITDA margin38.2%35.3%53.4%62.1%61.1%
EBIT margin3.3%−13.9%11.2%33.5%35.3%
FCF margin−38.4%−74.7%−124.4%−56.3%−0.2%
EBITDA / net interest4.4x4.5x105.4xn.m.34.8x
Exhibit 9: Return on equity reaches 26.6% in FY27E at 32.2x current earnings
Returns & valuationFY24AFY25AFY26EFY27EFY28E
Returns & leverage
ROE (avg. equity)n.d.n.d.2.6%26.6%30.0%
ROA (avg. net assets)n.d.n.d.2.0%18.5%20.7%
ROCE (pre-tax EBIT)n.d.n.d.2.5%21.5%27.4%
Net debt / equityn.d.n.d.−13.3%31.8%22.6%
Net debt / EBITDAn.d.n.d.−0.9x0.7x0.4x
Valuation at $136.63
P/En.m.n.m.639.9x32.2x20.1x
EV/EBITDA362.3x294.4x73.4x15.9x9.2x
FCF yield−0.3%−0.7%−3.1%−5.6%0.0%
Dividend yield0.0%0.0%0.0%0.0%0.0%
P/E at our targetn.m.n.m.579.1x29.1x18.2x

Source: Form 424B4; Form 10-Q; Penmark Research model. Tables show earnings before interest and taxes (EBIT), return on assets (ROA) and return on capital employed (ROCE). *Forecasts built at segment Adjusted EBITDA; FY27E net interest is +$108m.

Exhibit 10: Our FY27E estimates sit 8.7% above consensus sales and 5.3% above EPS
Penmark vs Street4Q26E1Q27EFY26EFY27E
Revenue ($bn)
Penmark estimate22.3333.8749.41196.10
Consensus*22.9035.0048.80180.37
Prior-year period4.964.6918.6749.41
Diff. vs consensus−2.5%−3.2%+1.2%+8.7%
Growth y/y (Penmark)350.1%621.6%164.6%296.9%
Diluted EPS ($; qtrs: Adj. EBITDA $bn)
Penmark estimate12.93n.d.0.214.24
Consensus*11.04n.d.0.224.03
Prior-year period1.471.13n.d.0.21
Diff. vs consensus+17.1%n.d.−3.8%+5.3%
Growth y/y (Penmark)778.9%n.d.n.m.1,888%

Source: MarketScreener; Yahoo Finance; Q3 call; Penmark Research model. *1Q27E shows $33–37bn guidance midpoint; quarterly bottom block shows Adjusted EBITDA ($bn) because quarterly consensus EPS is unpublished.

Exhibit 11: At 15.9x FY27E EBITDA, SpaceX trades 78% above AI-compute peers
FY27E comparablesSPCXAI med.Diff.S&C med.
Valuation (current prices)
EV/EBITDA (FY27E)15.9x8.9x+78.1%11.6x
P/E (FY27E)32.2xn.m.n.m.20.4x*
Enterprise val. ($bn)1,938.573.2+2,548%146.0
FCF yield (FY27E)−5.6%n.d.n.d.n.d.
Dividend yield0.0%0.0%0.0pp0.8%
Margins, risk & scale
EBITDA margin62.1%58.9%+3.2pp48.7%
Segment margin**64.0%58.9%+5.1pp63.0%
FY27E EBITDA ($bn)121.911.0+1,007%20.3
Levered beta1.50***2.33−0.83x0.68
Sales growth FY27E296.9%~120%+177ppn.d.

Source: MarketScreener; Multiples.vc; yfinance; Penmark model (13 Nov 2026). AI median = CoreWeave & Nebius; Space & Connectivity (S&C) median = T-Mobile & Iridium offer (+36.9% diff.). *T-Mobile fwd P/E. **SI 64.0%, Conn. 63.0%. ***Penmark beta (peer median 2.04).

Upcoming catalysts test our narrow 0.5pp Hold buffer above the Sell line

Exhibit 12: Four dated events over the next 16 months test our Hold thesis
  • Dec 26Lock-up and Flight 15: Up to 798m Class A shares unlock on 8 Dec alongside Starship dual catch.
  • Feb 27Fourth-quarter results: Late-February results test our 2.7 GW compute forecast and the $100bn ARR goal.
  • 1H 27SI renewals and build pace: Mid-year renewals test rents near $31/W alongside an 8 GW build pace.
  • Nov 27EchoStar spectrum close: Closing the $9.3bn cash deal clears early-2028 Starlink Mobile service.

Source: Form 424B4; Q3 FY26 Form 10-Q & call; Penmark Research.

Exhibit 13: Three downside shocks push our rating below the −10% Sell line
  • Higher funding costs. Financing heavy capex widens spreads; +50 basis points (bp) on WACC cuts $6.80/sh (Sell). 35% prob.
  • Power and permit delays. Grid limits in Texas and Tennessee slowing FY27E by 1 GW cut $3.27/sh (Sell). 30% prob.
  • GPU wave and 90-day repricing. Rival capacity or major buyer pushback cutting rents $3/W trims $3.08/sh (Sell). 40% prob.
  • Upside: CEO compute plan. Reaching 12.7 GW end-FY27E (+4.7 GW) lifts value toward the bull case (+$15.38/sh). 25% prob.

Source: Q3 FY26 Form 10-Q; Penmark Research model (spot $/sh impact).

Exhibit 14: Shares trade at 32.2x FY27E EPS, down from 49.8x in June
PricePrice at each multiple of FY27E EPS ($4.24)$100$125$150$175$200$225Jul-26Sep-26Nov-2625x30x40x50x$137 (32.2x)

Source: yfinance daily closes since 12 Jun 2026 IPO; Penmark Research FY27E EPS

Exhibit 15: At 15.9x FY27E EBITDA and a 62.1% margin, SpaceX tops AI peers
0x5x10x15x20x20%30%40%50%60%70%EV / FY27E EBITDAFY27E EBITDA marginSPCXCoreWeaveNebiusT-MobileIridiumMicrosoftAlphabetAmazonR² = 0.00, n=7

Source: MarketScreener; Multiples.vc; yfinance; Penmark Research estimates (13 Nov 2026)

Exhibit 16: Four operating and funding signposts confirm or break our Hold rating
Indicator (latest)Thesis confirmed if (supports Hold / upside)Thesis broken if (re-rate to Sell or Buy)
SI compute (1.7 GW, 3Q26)4Q26 ends at 2.5–3.0 GW and FY27E tracks ~8.0 GW (+0.44 GW/mo).<7.0 GW end-FY27E (−$3.27/sh; Sell) or ≥10.0 GW (+$6.54/sh; $128.40 target, still Hold).
SI rent ($/W; >$30/W)FY27 renewals clear ~$30.5/W and fade gradually toward $26.0/W by FY32E.Any −$0.15/W/yr shift below path (<$27.03/W FY30E) tips rating to Sell.
Starlink subs (13.6m, $69)Net adds hold ~7.0m/yr in FY27E–28E and V3 satellites enter service mid-2027.Starship V3 slips past 2027 or churn stalls adds below 5.0m/yr (bear path).
Funding & WACC (11.3%)$40bn Apollo GPU lease and ~$42bn FY27E debt fund near 6.25–7.0% cost.10-year US Treasury (UST) yields or spreads lift WACC +50bp (−$6.80/sh; Sell) or cut −50bp (+$7.71/sh).
What would change our view: At the current price, Buy (above +10%) needs a target above $150.29, or a share price below $112.41. Sell (below −10%) triggers below a $122.97 target, or above $137.40 on unchanged forecasts.

Source: Q3 FY26 Form 10-Q; Form 8-K (26 Oct 2026); Penmark Research model. In the peer scatter, R-squared (R², the share of multiple variation explained by margin) is 0.00 because capital intensity outweighs margin.

PART 1 · SUMMARY · 3 THE THESIS IN CHARTS

Rapid sales growth is priced in, leaving five of six stress tests at Sell

Exhibit 17: Third-quarter revenue rose 161% year on year (y/y) to $14.6bn
RevenueAdj. EBITDA margin (right)$0bn$5bn$10bn$15bn20%30%40%50%60%70%1Q252Q253Q254Q251Q262Q263Q26

Source: Form 424B4 prospectus and Q2-Q3 FY26 Form 10-Q filings

Exhibit 18: Our $196.1bn FY27E sales forecast sits 8.7% above consensus
Street rangeConsensus meanPenmark$0bn$100bn$200bn$300bn$400bn$500bnFY26EFY27EFY28E$49bn, +1%$196bn, +9%$344bn, +4%

Source: Yahoo Finance; 24/7 Wall St (37 estimates); Penmark Research model

Exhibit 19: The current price implies $34.6/W FY30E SI rents, $7.4/W above base
$15/W$20/W$25/W$30/W$35/W$40/WFY27EFY29EFY31EPrice-implied $33/WPenmark base $26/WPenmark bear $16/WCost of capital $21/W

Source: Penmark Research SOTP DCF model and reverse-DCF price check

Exhibit 20: Only our 25%-weighted bull scenario clears the current share price
$0$50$100$150$200$250$10$88$210Bear10%Base65%Bull25%Weighted $111Price $137

Source: Penmark Research model; spot values at 13 Nov 2026 before 12m roll-forward

Exhibit 21: At 32.2x FY27E P/E on 26.6% ROE, SpaceX tops cloud peers
15x20x25x30x35x20%30%40%50%Forward P/EROE (%)SPCXMicrosoftAlphabetAmazonOracleT-MobileR² = 0.17, n=5

Source: yfinance; Penmark Research FY27E estimates (13 Nov 2026)

Exhibit 22: Five of six stress tests push our target below the −10% Sell line
Beta 2.55 (Sell)Beta 2.04 (Sell)Base alone (Sell)WACC +1pt / -1ptTerminal RONIC 12.8%(Sell)Target $123.66$70.00$90.44$98.63$106.70$145.40$120.92

Source: Penmark Research model (Sell below $122.97)

How the six charts connect: Operating momentum is strong, with third-quarter sales up 161% y/y to nearly $15bn. Our FY27E revenue forecast sits nearly 9% above consensus. Even so, the current price implies FY30E SI rents more than $7/W above our base path, and only the bull case clears spot. Shares also trade richer than cloud peers on earnings, and cutting terminal return on new invested capital (RONIC) or raising beta flips Hold to Sell.
PART 1 · SUMMARY · 4 KEY DEBATES

Winning three operating debates adds $30.02 per share, still $18.12 below spot

We sit on the cautious side of consensus on three operating debates and on funding costs. Even if SpaceX wins all three operating debates, spot value rises about $30 per share to roughly $119, still short of the current price.

Exhibit 23: Four debates on compute pace, SI rents, margins and risk divide the market
DebateMarket viewPenmark viewEvidenceChanges our view ifResolves
2027 SI compute build10.0 GW end-FY27E (JPMorgan; CEO sees +10 to 20 GW).8.0 GW end-FY27E (+5.3 GW added; +$6.54/sh gap).~220k GB300s/mo (~0.44 GW/mo); FY26 adds 1.9 GW; $228bn capex.1Q or 2Q27 pace tops 0.65 GW/mo with firm grid power.Jul 2027
SI rent durabilityRents hold at $27–30.5/W; spot implies $34.59/W in FY30E.$30.5/W in FY27E fades to $27.18/W in FY30E and $26.0/W by FY32E (+$4.63/sh).90-day exit right; Customer B is 34.5% of Q3 sales; Nebius at $20–25/W.Multi-year non-cancellable renewals above $30/W in 1H27.2H 2027
Long-run SI marginSI margin holds 62%+ (consensus FY28E group margin 68.6%).SI Adjusted EBITDA margin fades from 64% in FY27E to 55% by FY32E (+$18.84/sh).Power, cooling and staff per watt stay fixed as rents ease; Nebius 54.6%.FY27–28 results show software and Grok mix offsets rent fade.Feb 2028
WACC & V3 mobile32 Buy ratings ($256.41 mean target) price low WACC and V3 upside.11.3% WACC (1.50 beta) and $190.1bn Space & Conn. DCF; peer beta gives $90.44 target (Sell).30-year SpaceX notes yield ~8%; 10-year UST is 5.10%; mobile starts early 2028.10-year UST falls to 4.6% (+$7.71/sh) and V3 flies weekly in 1H27.Jun 2027

Source: Q3 FY26 Form 10-Q & call; MarketScreener; Investing.com; Penmark Research model.

Exhibit 24: Granting the market view on three debates adds $30.02, still short of spot
Market viewPenmark baseCurrent price vs base DCFSI margin 62% vs 55%FY27E compute 10 vs 8GWSI rent $27/W vs $26/W$88.49$136.63$88.49$107.33$88.49$95.03$88.49$93.12

Source: Penmark Research model (spot value at 13 Nov 2026)

How the debates map to our thesis

Three thesis pillars test about $30 per share of operating upside. Our compute build analysis tests an 8 GW versus 10 GW ramp, worth nearly $7 per share. Our pricing analysis tests SI rent fade and long-run margin fade, worth over $23 combined. Our Starlink and valuation analysis tests the mobile floor and discount rates.

Why consensus multiples fall short

Headline earnings multiples ignore nearly $429bn of two-year capital spending. Adding all three bull operating shifts lifts our base value to roughly $119 per share. That still sits about $18 below the current price without lower rates or larger builds.

Exhibit 25: Five dated milestones from Dec 2026 to Feb 2028 test our core debates
Dec 2026Day-180 lock-up & Flight15Up to 798m Class Aunlock; dual catch testFeb 2027Q4 FY26 print & $100bnARRTests 2.7 GW exit & Q1$33-37bn guideJul 2027Q2 FY27 compute buildcheckTests 8 GW vs 10 GW pace2H 2027SI renewals & EchoStarcloseTests $/W rent & $9.3bnspectrum cashFeb 2028FY27 full-year marginprintTests SI margin & mobilelaunch

Source: Company filings; Q3 FY26 earnings call; Penmark Research calendar

PART 2 · THE BUSINESS · 5 THE STORY

A rocket pioneer now gets 83% of sales from AI compute on 90-day contracts

SpaceX leads orbital launch and satellite internet, yet its value now rests on gigawatt data centres. It must spend nearly $429bn over two years for cloud buyers who can reprice in three months.

24 yrsFounded 2002 · IPO Jun 2026
13.6mStarlink consumer subs, 3Q26
$49.41bnFY26E sales ($18.67bn FY25A)
~22,000Staff · ~9,600 satellites
$1,963.9bnMarket cap at $136.63

The scene

On 28 September 2026, Starship Flight 14 reached orbit and deployed 26 next-generation Starlink satellites. That same month, crews wired about 220,000 NVIDIA GB300 chips into SpaceX data halls. Investors still picture a rocket maker, yet in the third quarter its SI unit earned about $7.6bn. That topped rockets and Starlink combined.

What it actually does

SpaceX runs three businesses under one roof. Space flies reusable Falcon 9 boosters up to 34 times alongside Starship for governments, satellite firms and internal flights. Connectivity runs about three-quarters of Earth’s maneuverable satellites, selling broadband across the 80% of land that fibre misses. SI rents gigawatt GPU clusters and sells Grok, coding tool Cursor and X ads.

How the money comes in

SpaceX pairs sticky consumer subscriptions with short-fuse wholesale compute contracts. SI generates over half of FY26E sales, with Connectivity near two-fifths and Space under one-tenth. Connectivity bills households $69 a month, while Space draws on a $63bn launch backlog. Every dollar of sales yields $0.53 in Adjusted EBITDA, yet consumes $1.73 in capital spending.

Where the story turns, and why it might break

SI capacity doubled past 1.7 GW this year as SpaceX bought Cursor at a $60bn value. It also signed Anthropic for about 325,000 GPUs at $1.25bn a month, lifting SI past four-fifths of FY27E sales. Reusable rockets protect Starlink, yet cloud buyers can exit on 90 days’ notice. One buyer drove over a third of third-quarter sales just as two-year spending nears $429bn.

$34.6/WPrice-implied FY30E SI rent
The current price assumes FY30E SI rents sit 27% above our forecast and 65% above the cost-of-capital floor.Source: Penmark Research reverse DCF (cut-off 13 Nov 2026). Rents just $0.15/W/yr below our path tip Hold to Sell.
1 · HOW THEY SIGN
90 daysDirect cloud deals for scarce power; cancellable on 90 days’ notice after three months.
2 · WHAT THEY BUY
325,000 GPUsAnchor cluster scale for Anthropic; SpaceX installs ~220,000 GB300s a month.
3 · WHAT THEY PAY
$1.25bn/moAnchor fee through May 2029; fleet averages $30.5/W across 5.35 GW in FY27E.
4 · WHAT SPACEX KEEPS
$19.52/W64% FY27E SI EBITDA on $40/W build cost, before five-year server refresh.

Exhibit 26: One SI anchor contract earns $19.52/W before server refresh. Source: Form 424B4; Q3 FY26 Form 10-Q & call; Penmark Research model.

Why the story matters for the stock: Investors value SpaceX’s AI cloud as if Starlink’s orbital moat protects GPU pricing forever. We expect SI rents to fade toward $26/W and margins toward 55% as new industry supply arrives.
PART 2 · THE BUSINESS · 5 HOW THE COMPANY MAKES MONEY (1 OF 3)

Super Intelligence drives 86% of group FY27E Adjusted EBITDA

Exhibit 27: SI reaches $26.1bn in FY26E, lifting group Adjusted EBITDA to $26.4bn
AI labs & X ads$26.1bn15.5m subs & E&G$18.7bnGov & commercial$4.7bnSI compute & AI$26.1bnConnectivity$18.7bnSpace launch$4.7bnGroup revenue$49.4bnAdj. EBITDA (53%)$26.4bnCash costs (47%)$23.0bnCustomer groupsSegmentsFY26E revenueOperating split

Source: Penmark Research FY26E model (9M FY26A reported plus Q4 FY26E stub). E&G = enterprise and government.

2.7 GWFY26E SI compute · +1.9 GW y/y
15.5mFY26E consumer subs · +6.6m y/y
$67.5FY26E consumer ARPU/mo · −17%
$63.0bn3Q26 backlog · $17.5bn deferred
476 t3Q26 mass to orbit · 35 launches

What it sells, to whom, and why customers buy

SpaceX runs three distinct business models across consumer, corporate, government and AI-lab buyers. Space generates under a tenth of FY26E sales, flying Falcon and Starship missions for the National Aeronautics and Space Administration (NASA), commercial operators and the $13.7bn National Security Space Launch (NSSL) programme. It also flew two-thirds of third-quarter missions internally to deploy Starlink. Connectivity generates nearly two-fifths of FY26E sales, selling low Earth orbit (LEO) broadband to households and enterprise fleets, where third-quarter sales jumped 91% y/y. SI generates over half of FY26E sales and rises past four-fifths in FY27E by renting GPU clusters and selling Grok, Cursor and X ads.

Pricing, payment model and revenue equations

Each segment follows its own revenue equation. Space collects milestone payments against a $63bn backlog, with two-thirds recognised within a year. Connectivity charges monthly subscription fees near $69 per household alongside upfront dish sales. SI multiplies powered capacity by revenue per watt, splitting compute roughly evenly between internal AI tools and outside tenants paying over $30/W.

Cash versus profit: working capital, prepayments and capex

Upfront customer deposits cushion working capital, yet capital spending overwhelms operating cash. Deferred revenue reached $17.5bn at September, lifting FY26E operating cash flow to nearly $25bn. Even so, roughly $86bn of FY26E capital spending creates negative FCF past $61bn, deepening past $110bn in FY27E.

What this map shows for valuationSpace, Connectivity and SI carry very different capital and refresh needs, so a single group multiple misprices the stock. Our SOTP DCF values Space & Connectivity at about $190bn EV and SI at about $1,124bn EV. We deduct ongoing SBC separately and treat $17.5bn of spent customer prepayments as debt-like in our bridge.
PART 2 · THE BUSINESS · 5 HOW THE COMPANY MAKES MONEY (2 OF 3)

Operating costs trimmed 3Q26 gross margin from 62% to 19% operating profit

Exhibit 28: Third-quarter sales of $14.58bn yielded $2.79bn of GAAP operating profit
SI$7.56bnConnectivity$5.25bnSpace$1.77bnRevenue$14.58bnCost of revenue$5.48bnGross profit$9.10bnR&D$4.60bnSG&A & other$1.71bnOperating profit$2.79bn

Source: Q3 FY26 Form 10-Q (three months ended 30 Sep 2026)

Operating margins look wide at 60% Adjusted EBITDA in the third quarter, yet narrow to 19% GAAP EBIT. Roughly $14.6bn of sales left $9.1bn in gross profit. Research and development (R&D) of $4.6bn plus $1.7bn of selling, general and administrative (SG&A) costs trimmed operating profit to $2.8bn.

Why it matters: Ongoing SBC of about $3.7bn in FY26E and five-year GPU depreciation are real economic costs that our DCF deducts in full.
Exhibit 29: FY27E sales of $196.1bn rest on 5.35 GW of average compute plus Starlink
Avg compute5.35 GW (8.0 end)Rev per watt$30.5/W/yrSI segment$163.2bn (83%)Consumer19.0m avg x $64/moE&G, kits, mobile$13.2bn (+50%)Connectivity$27.8bn (14%)FY26E base$4.6bnLaunch growth+10.0%Space$5.1bn (3%)Group revenue FY27E$196.1bn (+297%)

Source: Penmark Research model; highlighted nodes are primary valuation drivers

Exhibit 30: SI and Connectivity both top 63% FY27E EBITDA margins as Space builds Starship
50%80%−20%30%0%600%FY27E EBITDA margin (%)FY27E revenue growth (%)Circle area: FY27E revenue ($bn) MatureHypergrowthLowmarginReinvestingSIConnectivitySpaceCoreWeaveT-Mobile

Source: Penmark Research model; peer consensus

Exhibit 31: Two customers accounted for 51% of 3Q26 revenue, led by Customer B at 35%
Each square is 1%Customer B (SI cloud)35%35Customer A (multi-seg)17%17All other customers48%48

Source: Q3 FY26 Form 10-Q Note on concentration (exact shares: 34.5%, 16.6%, 48.9%)

PART 2 · THE BUSINESS · 5 HOW THE COMPANY MAKES MONEY (3 OF 3)

Each SI watt earns $19.52 EBITDA on $40 capex before GPU refresh

Exhibit 32: One watt of SI compute earns $19.52 EBITDA on $40 of build cost
$0$10$20$30$40FY27E rent / WPower, staff & opexFY27E EBITDA / W5-yr refresh / yrPre-tax cash / W$30.50−$10.98$19.52−$4.80$14.72

Source: Penmark Research model ($40/W new build capex; $24/W 5-year GPU server refresh)

What drives volume and blended $/W price

Blended SI revenue per watt peaks near $30.5/W in FY27E across 5.35 GW of average compute. Management prices outside GPU rentals above $30 to $50/W while keeping roughly half of capacity for Grok, Cursor and X. As high-bandwidth memory (HBM) and grid bottlenecks ease, we fade blended rents toward $27/W in FY30E and $26/W by FY32E.

Fixed versus variable cost to deliver

Power, cooling and site operating costs per watt stay largely fixed near $11/W to $12/W. A $4.5/W drop in blended rents therefore trims SI Adjusted EBITDA margin by 9pp to 55%. That lands right alongside peer Nebius near 55%.

Capex intensity and the five-year GPU refresh

Every new gigawatt costs $40bn to build, and servers require a $24/W replacement five years later. New halls push group capital spending past $228bn in FY27E and $200bn in FY28E. Five years later, replacing FY27E servers triggers over $127bn of refresh spending in FY32E.

Reconciling management’s sub-year payback claimThe chief financial officer (CFO) cited under-one-year payback on marginal rented watts priced above $40/W. Across total capacity earning about $20/W in EBITDA on $40/W capex, pre-tax cash payback is roughly two years, and return on invested capital (ROIC) moderates after server refresh and SBC.
Exhibit 33: New SI halls push FY27E capex to $228.3bn ahead of server replacements
$0bn$50bn$100bn$150bn$200bn$250bnFY24AFY25AFY26EFY27EFY28EFY29EFY30EFY31EFY32EActualEstimatesSI new capacitySI 5-yr GPU refreshSpace & Connectivity

Source: Form 424B4; Q3 FY26 10-Q; Penmark Research model

PART 2 · THE BUSINESS · 6 THE MARKET

SpaceX operates 75.6% of active satellites as AI data centres reach $7tn

Exhibit 34: Addressable demand spans $77bn US space budgets to $7tn in AI data centres
AI data centres to 2030Sat-to-mobile 10Y GDPUS gov space (2024)SPCX FY26E revenueNSSL Phase 3 Lane 2$1bn$10bn$100bn$1tn$10tn$7.0tn$1.4tn$77bn$49.4bn$13.7bn

Source: Form 424B4 prospectus (Space Foundation, CTIA); Penmark Research model

Exhibit 35: SpaceX operates 75.6% of the world's 12,700 active maneuverable satellites
12,700Mar 2026 satsSpaceX Starlink76%9,600All other operators24%3,100

Source: Form 424B4 prospectus (as of 31 Mar 2026)

Bottom-up market sizing and demand cycle

Three structural demand curves support group growth. US government space spending reached about $77bn in 2024. Another $13.7bn of NSSL launch awards run through 2032. Land networks cover just a fifth of Earth while daily traffic tops 585 exabytes (EB), and direct-to-phone links could add $1.4tn to 10-year gross domestic product (GDP). Global data-centre spending is projected at $7tn through 2030, with generative AI taking roughly 70% of power.

Concentration and orbital share

Orbital launch and LEO broadband are near-monopolies, whereas AI cloud stays fragmented. SpaceX carried over four-fifths of roughly 2,600 tonnes of global payload in 2025. Falcon 9 flew 165 missions that year. It also runs about 9,600 of Earth’s 12,700 maneuverable satellites.

Value chain and who earns the margin

SpaceX owns launch and orbital spectrum, yet buys chips and grid power outside. Reusable boosters let Connectivity earn a 63% FY27E EBITDA margin without paying outside launch markups. In SI, NVIDIA and Taiwan Semiconductor Manufacturing Co. (TSMC) capture chip profits while turbines and grid hookups pace expansion.

Regulatory, spectrum and substitution threats

Permits and customer self-build remain the main constraints. Federal Aviation Administration (FAA) flight approvals govern Starship V3, while Federal Communications Commission (FCC) approval of EchoStar’s $9.3bn cash spectrum transfer gates 2028 mobile. State grid reviews risk a 1 GW build delay, and cloud tenants can self-build on 90 days’ notice.

Exhibit 36: Vertical integration across launch, 9,600+ satellites and AI apps captures core margin
Chips & grid powerNVIDIA, utilities,Megapack$40/W SI capexOrbital launchFalcon 9 (34x reuse),Starship>80% global upmassLEO & mobile9,600+ sats, EchoStarspectrum$18.7bn FY26E revAI compute clusters1.7 GW live; 8.0 GW FY27E$26.1bn FY26E SI revAI apps & socialGrok, Cursor, X (~550mMAU)$63bn group backlogVertically integrated SpaceX operations

Source: Form 424B4 prospectus; Q3 FY26 10-Q; Penmark Research

PART 2 · THE BUSINESS · 7 COMPETITIVE POSITION

SpaceX leads orbital launch and LEO broadband despite 90-day AI cloud contracts

Exhibit 37: SpaceX leads on launch cost and 9,600-satellite scale, with a narrower AI moat
SPCXLaunchpeersCarriersNeocloudsReusable launch costLEO broadband scaleSpectrum & mobileAI cluster build speedContract lock-inNoneWeakFairStrongLeading

Source: Form 424B4; Q3 FY26 10-Q; Penmark Research competitive assessment

Exhibit 38: SI (64.0%) and Connectivity (63.0%) top peer FY27E EBITDA margins
SI (SPCX FY27E)CoreWeaveConnectivity (SPCX)IridiumNebiusT-Mobile64.0%63.2%63.0%56.8%54.6%40.6%

Source: MarketScreener (Nov 2026); Penmark Research model

Exhibit 39: SpaceX ranks #1 in reusable launch and global LEO broadband
Segment & criterionSpaceX rank & proofDisclosed peersCustomer self-build / substitute
Space: $/kg & cadence#1 · 34x reuse; >80% massUnited Launch Alliance (ULA), Blue Origin, Rocket LabSovereign vehicles (high cost, low cadence)
Space: reliability#1 · >99% on ~650 flightsULA (#2 NSSL), Northrop GrummanPayload rideshare / mission deferral
Conn.: latency & reach#1 · 21ms, >400 Mbps, 700 TbpsViasat, SES, OneWeb, IridiumFibre & 5G fixed wireless access (FWA; 20% land)
Conn.: direct-to-cell#1 · ~650 D2D* sats, 7.4m dev.AST SpaceMobile, Globalstar, LynkTerrestrial tower densification / roaming
SI: time-to-power (GW)Top 3 · ~220,000 GB300/moCoreWeave, Nebius, Microsoft, AlphabetHyperscaler self-built GW campuses (2027–28)
SI: lock-in & modelsMixed · 90d cloud; Cursor $14bnOpenAI, Anthropic, MetaMulti-cloud portability & open-weight models

Source: Form 424B4; Q3 FY26 10-Q & call; Penmark. Mbps = megabits/sec; Tbps = terabits/sec. *Direct-to-device (D2D) links standard phones to satellites.

Why orbital moats endure while SI normalises

Physics and rocket reuse protect Connectivity, whereas commodity GPUs cap SI. Rivals need years of heavy-lift flights to match roughly 9,600 Starlink satellites and EchoStar spectrum, keeping Connectivity EBITDA margins above 60% through FY40E. By contrast, CoreWeave and Nebius buy the same NVIDIA chips, with recent Nebius deals near $20 to $25/W.

Comparable peers versus disclosed competitorsCoreWeave, Nebius, T-Mobile and Iridium serve as segment valuation benchmarks in our peer tables. Operationally, the prospectus names cloud giants Microsoft, Alphabet, Amazon and Meta alongside AI labs OpenAI and Anthropic as direct rivals, reinforcing why SI rents fade toward $26/W.
PART 2 · THE BUSINESS · 8 MANAGEMENT, GOVERNANCE AND CAPITAL ALLOCATION

Musk holds 84.4% of votes as three-year spending tops operating cash by $172.5bn

Execution under Elon Musk and chief operating officer (COO) Gwynne Shotwell built dominant launch and satellite moats. Minority holders face controlled voting, EBITDA-linked pay and ~$514bn of three-year spending.

Exhibit 40: Founder Elon Musk controls 84.4% of total shareholder voting power
DimensionDisclosed structure & alignment
Dual-class controlMusk holds 849.5m Class A + 5,569.1m Class B (10 votes/sh) = 84.4% vote (~82.4% alt. basis); Class B elects 51% of board.
Executive payMusk 302.1m Class B award needs $1.07tn–$6.57tn cap & 100 terawatts (TW) off-Earth; CFO options pay 371k/sh per $10bn EBITDA.
Board & insidersControlled Nasdaq exemption; Class B director D. Harrison resigned 4 Nov (no dispute); Shotwell sold 342k sh under Rule 10b5-1.
Related partiesSpaceX guarantees $20.21bn Valor leases ($1.92bn paid Jan–Apr 26); bought $653m FY25 / $307m 4M26 from Tesla (<1% holder).

Source: Form 424B4; Form 4 (22 Sep 2026); Form 8-K (6 Nov 2026).

Exhibit 41: Ten-vote Class B shares give Elon Musk 84.4% of votes on 44.7% stake
Elon MuskOther holdersVoting power84.4%15.6%Diluted economic44.7%55.3%

Source: Form 424B4 beneficial ownership table (6.42bn beneficial Musk shares vs 14.37bn diluted)

Exhibit 42: Three-year operating cash of $351.4bn falls $172.5bn short of capex and spectrum
−$200bn$0bn$200bn$400bnFY26E-28E OCFSI capexS&C capexEchoStar cash3Y cumulative FCF$351.4bn−$466.0bn−$48.2bn−$9.7bn−$172.5bn

Source: Penmark Research model (FY26E, FY27E and FY28E summed)

Exhibit 43: All three near-term Penmark forecasts sit inside management guidance bands
PenmarkGuidanceConsensus4Q26E compute (GW)4Q26E capex ($bn)1Q27E revenue ($bn)2.7 vs guide 2.5–3.0$37.5bn vs guide $35.0bn–$40.0bn$33.9bn vs guide $33.0bn–$37.0bn

Source: Q3 FY26 call guidance (20 Oct 2026); Penmark Research model

What governance and capital allocation mean for value: Shifting CFO option hurdles from FCF to Adjusted EBITDA rewards rapid gigawatt expansion over near-term cash flow. That incentive supports our FY27E revenue forecast sitting nearly 9% above consensus, while Class A holders with under 16% of votes cannot block TeraFab spending or a Tesla tie-up.
PART 3 · THE THESIS · 9 WHAT MOVES THE STOCK (1 OF 2)

Broad market and sector factors explain only 32.1% of daily share moves

SpaceX trades on forward SI capacity, compute rent durability, Starship V3 milestones and discount rates rather than trailing earnings. Across 108 trading days since listing in June 2026, each confirmed driver passed three or four tests. Dropped candidates passed at most one test, yet the stock still prices roughly $48 per share above our base case.

Exhibit 44: Event-day share moves exceed DCF per-share impact across all four drivers
DriverWhat moves SPCXTest countMarket reaction vs Penmark DCF valueKey debateNext number to watch
SI GW build & permitsMonthly GB300 installs & Texas/Tennessee grid approvals.Confirmed by four of four tests5 Nov grid fear cut SPCX −9.1% (~−$177bn diluted cap); DCF: −1 GW = −$3.27/sh, +2 GW = +$6.54/sh.Compute build pace4Q26 end GW (~25 Feb): we 2.7 GW (2.5–3.0 guide).
SI $/W rent & capexRented $/W yield vs quarterly capex burn.Confirmed by three of four tests2Q beat +9.4% reversed −13.6% next day on capex; 3Q rent guide added +7.8%. DCF: ±$3/W = ±$3.08/sh.Rent & margin fade1Q27E sales (~25 Feb): we $33.87bn vs $35bn mid.
Starship V3 & mobileOrbital V3 deployment & mobile spectrum.Confirmed by three of four testsFlight 14 orbit added +15.5% (2–5 Oct); space peers R² = 26.5%. DCF: Space & Conn. = $13.23/sh ($190.1bn).Starlink & bull skewFlight 15 catch (Dec 26) & 4Q subs (15.5m E).
WACC & bond spreads10Y UST yield & 30Y SpaceX note spreads.Confirmed by three of four testsP/E fell 49.8x to 32.2x as 10Y UST hit 5.10% & 30Y notes ~8%; Nasdaq-100 R² = 22.9%. DCF: ±50bp = −$6.80/+$7.71.Discount rate & skewDec FOMC & FY27 bond/lease spreads.

Source: Form 10-Q; Q3 call transcript; yfinance (108 trading days); Penmark Research model. bp = basis points; FOMC = Federal Open Market Committee.

Exhibit 45: Starship Flight 14 led post-IPO gains with a +15.5% two-day rally
Flight 14 orbit (2-5 Oct)2Q26 print day (4 Aug)Dir. exit / bounce (6 Nov)3Q26 beat T+1 (21 Oct)$40bn GPU lease (26 Oct)Midterm / AI selloff (5Nov)2Q26 capex digest (5 Aug)+15.5%+9.4%+8.6%+7.8%−0.4%−9.1%−13.6%

Source: yfinance daily closes; Penmark Research model (12 Jun to 13 Nov 2026)

Exhibit 46: Single-driver bull shifts still leave base value well below the share price
Share priceBase + catalyst+4.7 GW 2027 computeWACC -50bp (4.6% 10Y)SI rent +$3/W FY28-30EStarlink +1m/yr adds$103.87$136.63$96.20$136.63$91.57$136.63$88.69$136.63

Source: Penmark Research model (single-driver shifts on $88.49 spot base)

Post-listing return decomposition:Shares fell about 15% from the June listing close to mid-November, lagging the benchmark by nearly 17pp. The stock still sits 1.2% above its IPO price as consensus FY27E EPS rose roughly sevenfold. Yet the forward P/E compressed from nearly 50x at the June peak to about 32x, showing that capital needs and discount rates govern the multiple.
PART 3 · THE THESIS · 9 WHAT MOVES THE STOCK (2 OF 2)

Company-specific news drives 67.9% of daily return variance, led by SI power

Exhibit 47: Company-specific news explains 67.9% of SpaceX daily stock return variance
Idiosyncratic SPCXFull multi-factor R2Space & telecom peersNasdaq-100 aloneS&P 500 aloneAI neoclouds(CRWV/NBIS)67.9%32.1%26.5%22.9%20.8%17.5%

Source: yfinance daily return regressions, N = 107 trading days (12 Jun to 13 Nov 2026). CRWV = CoreWeave; NBIS = Nebius.

Exhibit 48: SI compute power, funding and TeraFab dominated 57% of Q3 call analyst questions
SI GW, power & fundingTeraFab & chip supplyStarlink Mobile & V3Internal vs rented GPUTesla combinationStarship pads & cadence28.6%28.6%14.3%14.3%7.1%7.1%

Source: Q3 FY26 earnings call transcript (20 Oct 2026), Penmark topic coding

SI COMPUTE BUILD & POWER PERMITS (COMPUTE PACE DEBATE)

Investors extrapolate power announcements beyond DCF impact. Grid permit fears cut the stock about 9% on 5 November. Meanwhile, +10 GW commentary dominated 29% of call questions. Adding 2 GW lifts spot value ~$6.5. A 1 GW delay cuts ~$3.3. Watch: Year-end GW in late February (~2.7 GW estimate; ±5% move).

SI RENT & CAPEX PAYBACK (RENT & MARGIN DEBATES)

Capex surprises punish peak-rent multiples. Shares fell nearly 14% after Q2 capex offset a revenue beat. They later rose nearly 8% on Q3 rent and ARR commentary. A $3/W rent shift moves value ~$3.1. Holding 62% margin adds nearly $19. Watch: First-quarter revenue guide in late February (~$33.9bn estimate).

STARSHIP V3 & MOBILE SPECTRUM (STARLINK & SKEW DEBATE)

Orbital milestones support the connectivity floor. Flight 14 sparked a 15.5% two-day rally. Space peers explain ~27% of daily moves, and orbital topics drew 21% of call questions. Space and Connectivity contributes ~$13 per share in our base case. Watch: Flight 15 dual catch in December 2026 (±6–10% move).

DISCOUNT RATE & BOND SPREADS (STARLINK & SKEW DEBATE)

Back-loaded cash flow makes SpaceX sensitive to rates. The Nasdaq-100 explains ~23% of daily moves. Long bond yields near 8% compressed forward P/E from nearly 50x to ~32x. A 50bp WACC shift moves spot value by roughly $7 to $8 per share. Watch: December rate decision and FY27 debt spreads (±5% move).

Exhibit 49: Three candidate drivers failed to pass two empirical stock tests
Candidate driverTest resultWhy it was dropped
Trailing GAAP EPSConfirmed by zero of four testsNine-month EPS was negative, and the Q3 earnings beat left the stock flat until call ARR and GW commentary arrived.
Tesla merger / TeraFabConfirmed by one of four tests (Q&A)Analysts raised Tesla and TeraFab on the Q3 call, but management hedged timelines and neither has a revenue model.
Board / insider filingsConfirmed by one of four tests (events)Shares rose 8.6% on 6 November with an AI rebound despite a director exit, and Shotwell sold under a pre-set plan.

Source: Securities and Exchange Commission (SEC) Form 4 & 8-K; Q3 FY26 transcript; yfinance (n = 107 daily returns, 15 Jun–13 Nov 2026); Penmark Research.

Sample & methodology limits:SpaceX has only five months of public trading across 108 sessions, two earnings releases and one call transcript. That brief history inflates raw daily equity beta to nearly 2.7 against the S&P 500. Our valuation therefore uses a 1.50 beta and stress-tests a 2.04 peer-rule beta in our valuation chapter.
PART 3 · THE THESIS · 10 SI COMPUTE BUILD AND CAPEX
1 · THE CLAIM · SI COMPUTE BUILD PACESI compute reaches 8.0 GW in FY27E at a $110bn free cash flow burn

What the market believes

Consensus FY27E revenue lags recent GPU installations, while bulls price management’s +10 to 20 GW ambition without its cash cost. Street capex implies roughly 6 GW at end-FY27E. Bullish analysts model 10 GW, while management cited $500bn+ of potential ARR.

The evidence (three independent sources)

  • Reported 10-Q KPIs & PP&E. SI compute rose from 0.8 GW in FY25 to 1.7 GW in Q3. Nine-month SI capex reached $40bn as servers and construction in progress (CIP) neared $67bn.
  • Q3 call run-rate vs physical grid limits. Installing roughly 220,000 GB300s monthly reaches 2.7 GW at year-end and 8 GW a year later. Texas and Tennessee turbine and permit queues cap faster growth.
  • Contracted lease & debt filings. October filings locked in $40bn of Apollo GPU leases, yet reaching 8 GW still requires roughly $42bn of new FY27E debt.
Exhibit 50: Base SI compute reaches 8.0 GW in FY27E, ahead of the 6.0 GW Street pace
Bull (CEO plan)Consensus implied0 GW10 GW20 GW30 GW40 GWFY25AFY26EFY27EFY28EFY29EFY30EActualEstimates0.8 GW2.7 GW8.0 GW12.5 GW16.0 GW19.0 GW

Source: Form 424B4; Q3 FY26 10-Q; Penmark Research model

Exhibit 51: Our 8.0 GW FY27E compute forecast puts group sales 8.7% above consensus
Metric ($bn unless stated)FY26EFY27EFY28E
SI compute, Penmark (GW)2.78.012.5
Implied Street / CEO view (GW)2.76.0 / 12.711.0 / n.d.
Group revenue, Penmark49.41196.10343.80
Group revenue, consensus48.80180.37330.90
Difference vs consensus+1.2%+8.7%+3.9%
SI capex, Penmark ($40/W)(74.00)(212.00)(180.00)
Group FCF, Penmark(61.45)(110.37)(0.69)
Spot DCF delta (+2 GW / +4.7 GW)n/a+$6.54 / +$15.38per share

Source: Q3 call; MarketScreener/Yahoo Finance; Penmark model.

What it is worth

Our 8 GW base path values SI at roughly $1.1tn after charging $392bn of two-year SI capex. Reaching 10 GW adds about $6.5 per share, and management’s bull build adds over $15. Conversely, a 1 GW delay cuts ~$3.3 per share and tips our call to Sell.

Dated signposts

  • Dec 2026: Texas grid review of SpaceX power sites (confirm: no cap; break: 1 GW delay).
  • Late Feb 2027 (4Q26): Check year-end compute near 2.7 GW and December ARR at $100bn+.
  • Jul 2027 (2Q27): Confirm ~0.4 GW monthly pace versus >0.65 GW bull pace.
What would prove us wrong:Energising over 300,000 GB300s monthly in early 2027 with secured turbine power would track toward 12.7 GW at year-end. If debt costs stay near 7%, spot value rises by about $15 per share to nearly $104. Our 12-month target would then reach ~$134.8 (−1.3%), remaining a Hold.
PART 3 · THE THESIS · 11 SI RENT AND MARGIN DURABILITY
1 · THE CLAIM · SI RENT AND MARGIN FADEFading SI rent to $26/W leaves FY28E EBITDA 7.5% below consensus

What the market believes

At the current share price, investors imply roughly $34.6/W of FY30E SI rent while consensus extrapolates nearly 69% group EBITDA margins. Bulls treat recent scarcity pricing above $30 to $50/W and sub-one-year payback as permanent.

The evidence (three independent sources)

  • Short-fuse contracts & buyer concentration. Cloud capacity contracts allow 90-day cancellation after three months. Customer B alone generated nearly 35% of third-quarter group revenue.
  • Peer supply wave & forward pricing. CoreWeave and Nebius are expanding capacity aggressively and signing multi-year deals near $20 to $25/W as GB300 supply normalises.
  • Fixed power/cooling cost math. Site power, cooling and staffing stay near $11/W. A $4.5/W rent fade from FY27E to FY32E therefore trims SI margin by 9pp to 55%.
Exhibit 52: Base SI rent settles at $26.0/W by FY32E, just below the $27.03/W Sell trigger
$15/W$20/W$25/W$30/W$35/WFY27EFY28EFY29EFY30EFY31EFY32EBull $29/WBase $26/WBear $16/WFY30E Sell trigger $27.03/W

Source: Penmark Research model

Exhibit 53: Lower SI rent and margins leave FY28E EBITDA $17.1bn below consensus
MetricFY27EFY28EFY30E
SI rev / watt, Penmark ($/W/yr)30.5029.3227.18
Consensus / Price-implied ($/W/yr)30.5030.5030.50 / 34.59
SI Adj. EBITDA margin, Penmark64.0%62.0%58.0%
SI Adj. EBITDA, Penmark ($bn)104.43186.33275.88
Group Adj. EBITDA, Penmark ($bn)121.85210.06312.34
Group Adj. EBITDA, consensus ($bn)112.67227.14n.a.
Penmark vs consensus EBITDA+8.2%−7.5%n/a
Hold/Sell boundary ($27.03/W FY30E)−$0.15/W$87.53/shTips to Sell

Source: Form 424B4 & 10-Q; Penmark model. n.a. = not available.

What it is worth

SI pricing and margin durability offer over $23 per share of upside if bulls are right, yet also form our Sell trigger. Holding SI rent at $27/W adds ~$4.6 per share. Holding 62% margin adds nearly $19, whereas a small FY30E rent slip tips our call to Sell.

Dated signposts

  • 1H 2027: Initial GB300 renewals; check whether multi-year contracts replace 90-day exits.
  • Oct 2027 (3Q27 10-Q): Customer B share (>35% raises repricing risk; <25% diversifies).
  • Feb 2028 (FY27 10-K): Full-year SI margin against our 64% forecast.
Why this is the knife-edge variable:SI represents roughly 92% of operating segment EV and faces a five-year $24/W GPU refresh cycle. Every $1/W move in medium-term rent shifts equity value by roughly $1 per share. In the bear case, a supply glut, a Customer B exit and a slower build cut spot value below $10 per share.
PART 3 · THE THESIS · 12 STARLINK FLOOR AND BULL SKEW
1 · THE CLAIM · STARLINK FLOOR AND UPSIDE SKEWStarlink cash flow and bull-case skew lift weighted value $22.46 above base

What the market believes

Pure AI-cloud bears focus on our base DCF alone and short the stock, while 32 of 34 rated analysts carry Buy calls with a ~$256 mean price target (PT). Bulls treat $500bn+ ARR and global mobile leadership as the base outcome.

The evidence (three independent sources)

  • Reported 10-Q Starlink profitability. Connectivity sales rose 69% year on year to about $5.2bn in Q3. Enterprise and government sales grew 91% as nine-month EBITDA margin neared 64%.
  • Starship V3 step-change in orbital cost. Flight 14 deployed 26 V3 satellites rated near 1 terabit per second (Tbps) each. That represents roughly 20 times prior satellite capacity.
  • EchoStar mobile spectrum moat. Buying Advanced Wireless Services-4 (AWS-4) and H-block spectrum for about $9.3bn cash plus shares unlocks mobile coverage in early 2028.
Exhibit 54: Connectivity revenue grows 5.1x from FY25A to reach $58.4bn by FY30E
Connectivity revenueAdj. EBITDA margin (right)$0bn$20bn$40bn$60bn50%55%60%65%FY24AFY25AFY26EFY27EFY28EFY29EFY30EActualEstimates

Source: Form 424B4; Q3 FY26 10-Q; Penmark Research model

Exhibit 55: Space and Connectivity adds $190.1bn of DCF value to anchor our Hold call
Metric ($bn unless stated)FY26EFY27EFY30E
Consumer subscribers, end (m)15.522.542.0
Connectivity revenue18.7027.8158.43
Connectivity Adj. EBITDA (63→60%)11.8417.5235.06
Space & Conn. unlevered FCFn/a(9.13)9.72
Valuation summaryPenmark model
Space & Conn. DCF EV (vs $202.48bn peer mult.)$190.10bn ($13.23/sh)
Base value (65% wt): spot / 12m$88.49 / $98.63
Bull (25%) / Bear (10%) spot$209.78 / $9.84 (+$22.46)
Prob.-weighted spot / 12m target$110.95 / $123.66 (−9.5%)

Source: Form 10-Q; Penmark Research model.

What it is worth

Space and Connectivity provides a ~$190bn DCF floor (~$13 per share) and turns FCF-positive in FY28E. Segment unlevered FCF reaches nearly $35bn by FY40E. Pairing that floor with a 25% bull weight adds over $22 per share above our base case, lifting our target to Hold.

Dated signposts

  • Dec 2026: Starship Flight 15 dual booster and ship catch.
  • 30 Nov 2027: Regulatory close of EchoStar spectrum (~$9.3bn cash).
  • Early 2028: Commercial Starlink Mobile launch on new spectrum.
Counter-evidence & unvalued options:Shorting SpaceX on our base DCF alone carries sharp right-tail risk, with about 141m shares sold short. Delivering our bull compute and rent path lifts forward value past $233 (+71%). Monetising TeraFab, orbital compute, Mars or a Tesla tie-up adds further unvalued upside.
PART 4 · THE NUMBERS · 13 TRACK RECORD AND QUALITY OF REPORTED NUMBERS

Non-cash charges held nine-month GAAP EPS at −$0.15 despite $27.1bn of sales

Operating scale is real, yet Adjusted EBITDA overstates distributable profit before depreciation and amortisation (D&A) and equity pay. Revenue grew roughly a third per year through FY25A on Starlink volume before mergers and acquisitions (M&A) and GPU rents lifted nine-month sales to $27.1bn. Free cash flow excluding customer prepayments remained negative at roughly $40bn.

Exhibit 56: Depreciation and equity pay absorb 79% of our $26.4bn FY26E Adjusted EBITDA
US$m (unless stated)FY23AFY24AFY25A9M26AFY26E
Reported revenue (Space / Conn. / SI)10,38714,01518,67427,08349,408
Revenue growth y/y (organic + xAI/Cursor M&A)n/a+34.9%+33.2%n.d.+164.6%
Adjusted EBITDA3,8215,3506,58413,47426,402
Adjusted EBITDA margin36.8%38.2%35.3%49.8%53.4%
Depreciation & amortisation(2,635)(3,824)(6,701)(8,934)(13,907)
Ongoing + acquisition SBC (Cursor in 9M26)(679)(784)(1,947)(3,833)(6,949)
GAAP operating income (EBIT)(3,505)466(2,589)7055,545
Cash from operations / Free cash flow4,520 / 1055,776 / (5,387)6,785 / (13,952)13,097 / (34,972)24,509 / (61,445)

Source: Form 424B4 (FY23–FY25, restated for xAI combination); Q3 FY26 Form 10-Q; Penmark model. FY23A includes $4,012m restructuring/impairment.

Exhibit 57: SI and Connectivity generated $23.7bn, or 88% of nine-month group revenue
SIConnectivitySpace$0bn$10bn$20bn$30bnFY23AFY24AFY25A9M26A$10bn$14bn$19bn$27bn

Source: Form 424B4 prospectus (FY23A-FY25A); Q3 FY26 Form 10-Q (9M26A)

Exhibit 58: D&A and SBC create a $20.9bn wedge between FY26E Adj. EBITDA and GAAP EBIT
Adj. EBITDAGAAP EBITFY23AFY24AFY25A9M26AFY26E−$3.5bn$3.8bn$0.5bn$5.3bn−$2.6bn$6.6bn$0.7bn$13.5bn$5.5bn$26.4bn

Source: Form 424B4; Q3 FY26 10-Q; Penmark Research model

Exhibit 59: Deducting $17.52bn of prepayments and ongoing equity pay prevents cash double-counting
Accounting areaReported metric impactQuality assessment & Penmark valuation treatment
Share-based compensation$2,616m ongoing + $1,217m Cursor awards in 9M26A.Added back in Adj. EBITDA ($6.95bn FY26E). We deduct ongoing after-tax SBC as a real cash cost (−$93.79bn Corporate DCF present value (PV) with NWC).
Customer prepayments$17,520m deferred revenue at 30 Sep (+$5,404m vs Dec 25).Boosts GAAP operating cash flow (+$5.40bn in 9M26A). We exclude prepayments from DCF FCF and deduct $17.52bn (−$1.22/sh) in our EV bridge.
GPU & satellite livesGross PP&E $110.01bn ($43.21bn servers, $23.75bn CIP).5-year server depreciation lags cash capex during rapid expansion; DCF captures full cash capex plus $24/W five-year GPU refresh from FY29E.
xAI & Cursor restatement$48.42bn goodwill & $6.35bn intangibles on balance sheet.Prospectus restated FY23–25 for xAI common-control merger; Cursor added replacement SBC ($7.5bn FY27E, $6.5bn FY28E, $4.0bn FY29E before zeroing).

Source: Form 424B4; Q3 FY26 Form 10-Q; Penmark Research model.

PART 4 · THE NUMBERS · 14 FORECASTS (1 OF 2)

Group revenue grows 11x from FY26E to reach $541.1bn by FY30E

We model 14 explicit years through FY40E so both the initial build wave and five-year GPU refreshes enter cash flow. Every near-term estimate sits inside management’s guidance bands. Our medium-term SI rent fade puts FY28E EBITDA 7.5% below consensus.

Exhibit 60: Group revenue scales from $196.1bn in FY27E to $541.1bn by FY30E
Avg compute5.35 -> 17.50 GWRev per watt$30.5 -> $27.2/WSI segment$163.2bn -> $475.7bnConsumer subs22.5m -> 42.0m endARPU & E&G$64 -> $56.5; E&G $32bnConnectivity$27.8bn -> $58.4bnRevenue CAGR+11.3% FY27E-30EEBITDA margin-2% -> +20%Space$5.1bn -> $7.1bnGroup revenue FY27E -> FY30E$196.1bn -> $541.1bn

Source: Penmark Research model (FY27E to FY30E path). E&G = enterprise and government; CAGR = compound annual growth rate.

Exhibit 61: SI drives 83% of FY27E group revenue as compute scales to 19.0 GW
US$bn (unless stated)FY24AFY25AFY26EFY27EFY28EFY29EFY30E
Space revenue (Adj. EBITDA margin)3.80 (30%)4.09 (16%)4.65 (0%)5.11 (−2%)5.73 (8%)6.41 (15%)7.05 (20%)
Consumer end subs (m) × ARPU ($/mo)4.4 × 918.9 × 8115.5 × 67.522.5 × 64.029.5 × 61.036.0 × 58.542.0 × 56.5
Connectivity revenue (63%→60% margin)7.6011.3918.7027.8137.5447.9858.43
SI end GW / avg GW × $/W/yr0.3 GW0.8 GW2.7 GW8.0 / 5.35×30.512.5 / 10.25×29.316.0 / 14.25×28.119.0 / 17.5×27.2
SI revenue (64%→58% margin)2.623.2026.06163.18300.53400.85475.65
Group revenue (growth y/y)14.0218.67 (+33%)49.41 (+165%)196.10 (+297%)343.80 (+75%)455.24 (+32%)541.13 (+19%)
Group Adj. EBITDA (margin)5.35 (38.2%)6.58 (35.3%)26.40 (53.4%)121.85 (62.1%)210.06 (61.1%)270.74 (59.5%)312.34 (57.7%)

Source: Form 424B4; Q3 FY26 Form 10-Q; Penmark Research model.

Exhibit 62: Rent fade leaves FY28E Adjusted EBITDA 7.5% below Street consensus
Line & periodPenmarkStreet / GuideGapAnalytical reason for gap
Dec 26 ARR / 1Q27 rev$103.1bn / $33.9bn≥$100bn / $33–37bnInside2.7 GW exit compute at $27.5/W 4Q stub + $5.9bn 4Q Connectivity lands inside guide.
FY27E revenue$196.10bn$180.37bn Street+8.7%Street capex lags ~220,000 GB300s/mo pace (8.0 GW end-FY27E vs ~6.0 GW implied Street).
FY27E Adj. EBITDA / EPS$121.85bn / $4.24$112.67bn / $4.03+8.2% / +5.3%Higher SI GW drives +8.2% EBITDA; $42.87bn D&A and $13.38bn total SBC narrow the EPS gap to +5.3%.
FY28E revenue / EBITDA$343.80bn / $210.06bn$330.90bn / $227.14bn+3.9% / −7.5%12.5 GW lifts revenue +3.9%, but SI fade ($29.32/W, 62% margin vs Street 68.6% group) cuts EBITDA −7.5%.

Source: Q3 call guidance; MarketScreener/Yahoo Finance consensus; Penmark model.

Long-run operating path to FY40E:Connectivity scales to 68m consumer subscribers and about $129bn of revenue by FY40E at a 60% EBITDA margin. SI reaches 35.5 GW and roughly $904bn of sales at a 55% margin. Group revenue then tops $1.0tn.
PART 4 · THE NUMBERS · 14 FORECASTS (2 OF 2)

Three-year capex of $514.2bn lifts gross debt to $116.5bn before cash inflects

Cash conversion turns positive in FY29E, then absorbs a $127bn GPU refresh wave in FY32E. Keeping a $30bn minimum cash floor requires $40bn of GPU leases plus about $51bn of new debt over two years. Surplus FY29E cash retires all new debt and starts buybacks in FY30E.

Exhibit 63: Free cash flow troughs at -$110.4bn in FY27E before turning +$72.5bn in FY29E
−$150bn−$100bn−$50bn$0bn$50bn$100bn$150bnFY24AFY26EFY28EFY30EFY32EActualEstimates$100bnBreak-even

Source: Form 424B4; Q3 FY26 10-Q; Penmark Research model

Exhibit 64: Gross debt and GPU leases peak at $116.5bn in FY28E before repayment
$0bn$25bn$50bn$75bn$100bn$125bnFY24AFY26EFY28EFY30EActualEstimatesGross debt & leases $50bnCash & securities $30bnMin cash $30bn

Source: Form 424B4 FY24A-25A balance sheets (cash ex marketable securities); Penmark FY26E+ model.

Exhibit 65: Surplus cash funds $132.2bn of share buybacks starting in FY30E
US$bn (unless stated)FY26EFY27EFY28EFY29EFY30EFY31EFY32E
Cash from operations (incl. prepayments)24.51127.30199.58241.16280.08304.88328.98
Capex (new build + $24/W GPU refresh)(85.57)(228.34)(200.26)(168.63)(153.69)(167.51)(229.32)
EchoStar spectrum cash payments(0.39)(9.33)0.000.000.000.000.00
Free cash flow (reported definition)(61.45)(110.37)(0.69)72.52126.39137.3799.67
Memo: FCF ex customer prepayments(66.62)(116.61)(11.53)70.48119.52131.6994.95
Net GPU lease + new debt draw / (repay)0.0076.990.69(58.68)(8.00)(8.00)(3.00)
Share buybacks (grown at cost of equity, Ke)0.000.000.000.00(132.24)(129.37)(96.67)
End cash / Gross debt & leases63.4 / 38.830.0 / 115.830.0 / 116.543.8 / 57.830.0 / 49.830.0 / 41.830.0 / 38.8

Source: Penmark Research model. FY26E spectrum cash line shows Q4 stub ($385m) after $856m paid in 9M26A.

Exhibit 66: Adding 4.7 GW of compute lifts spot DCF value by $15.38 per share
Driver shock (other inputs at base)FY28E EBITDASpot DCF ($/sh)Delta vs $88.49Call impact
SI compute +4.7 GW from FY27E (12.7 GW bull build)+$88.3bn103.87+$15.38Remains Hold ($134.80 target)
SI compute −1.0 GW from FY27E (grid/permit delay)−$18.8bn85.22−$3.27Tips to Sell
SI yield ±$3.0/W across FY28E–FY30E±$19.1bn91.57 / 85.40±$3.08Downside = Sell
Starlink consumer net adds +1.0m / −1.0m+$0.9bn / −$0.5bn88.69 / 88.38+$0.21 / −$0.11Remains Hold
Discount rate (WACC) −50bp / +50bpUnchanged96.20 / 81.69+$7.71 / −$6.80+50bp = Sell

Source: Penmark Research model.

Working capital & prepayments:Customer prepayments rise by about $5bn in FY26E, $6bn in FY27E and $11bn in FY28E alongside SI scale. Because the $17.5bn September balance funds unearned revenue, our valuation bridge deducts it in net debt rather than double-counting cash.
PART 4 · THE NUMBERS · 15 VALUATION (1 OF 2)

Both our spot DCF and peer-multiple check sit well below the share price

We value SpaceX with a 14-year sum-of-the-parts DCF, cross-checked against segment FY27E EV/EBITDA. Our 11.3% WACC is the blended annual return required by debt and equity holders. Our 3.0% terminal growth rate is the yearly cash flow growth assumed after FY40E, and peer multiples sit nearly 9% below our spot DCF.

Exhibit 67: Valuation ranges cluster below the $137 share price, with our target at $124
SOTP DCF, spot bear-bull12m rolled scenariosSOTP DCF, WACC +/-1ptPeer EV/EBITDA FY27EStreet 12m targets$0$200$400$600$800$10$210$11$234$76$105$54$95$70$800Price $137Target $124

Source: Penmark Research model (cut-off 13 Nov 2026)

Exhibit 68: Our EV/EBITDA cross-check lands 8.7% below our spot base DCF
Valuation componentPrimary SOTP DCF ($m)$/shFY27E EV/EBITDA ($m)$/shBasis & peer benchmark
Space & Connectivity EV190,10413.23202,47514.09DCF vs 11.62x TMUS/IRDM median on $17,420m EBITDA
SI (compute + Grok/X/Cursor) EV1,123,99278.20932,87964.90DCF vs 8.93x CRWV/NBIS median on $104,432m EBITDA
Corporate (after-tax SBC & NWC)(93,789)(6.53)n/an/aOngoing SBC expensed in peer EBITDA; PV deducted in DCF
Enterprise value (terminal value, TV = 49.2%)1,220,30784.901,135,35478.99Cross-check 7.0% below DCF EV before spectrum cash
Cash & securities ($89,855m) + digital ($1,573m)91,4286.3691,4286.3630 Sep 2026 balance sheet
Debt & leases (−$38,844m) & prepayments (−$17,520m)(56,364)(3.92)(66,077)(4.60)Cross-check also deducts −$9,713m unpaid EchoStar cash
Equity value (14,373.8m diluted shares)1,255,37187.341,160,70580.75Includes 261.8m EchoStar consideration shares
Value at price date (13 Nov) / 12m base$88.49 spot$98.63 12m$80.75 spot−8.7% gapDCF rolled 44d at 11.46% Ke; multiples at $136.63: 15.9x EBITDA, 32.2x P/E

Source: Form 10-Q; Penmark Research model. Multiples at $123.66 target: 14.4x FY27E EV/EBITDA and 29.1x FY27E P/E.

Discount rate and unvalued options: We build cost of equity from a 5.1% Treasury yield, a 4.2% equity risk premium (ERP) and a 1.50 beta. Pre-tax debt costs 7.0% (about 5.4% after tax). Starlink’s subscription cash flows behave like a telecom carrier’s, so the peer median overstates group risk, though we stress-test higher betas next. Our peer multiple check sits 8.7% lower because FY27E averages ~5.4 GW before later expansion, and we assign zero value to TeraFab, orbital compute, Mars or a Tesla combination.
PART 4 · THE NUMBERS · 15 VALUATION (2 OF 2)

Five of six valuation stress tests push our Hold call to Sell

Exhibit 69: Spot DCF spans $65.00 to $133.63 across discount rate and terminal growth inputs
$122.0$124.3$127.0$130.0$133.6$102.2$103.5$105.0$106.7$108.6$86.9$87.7$88.5$89.4$90.4$74.8$75.2$75.7$76.1$76.7$65.0$65.2$65.4$65.7$65.9Terminal growth2.0%2.5%3.0%3.5%4.0%WACC9.3% (-2pt)10.3% (-1pt)11.3% (Base)12.3% (+1pt)13.3% (+2pt)

Source: Penmark Research model (spot value at 13 Nov 2026)

Exhibit 70: Spot DCF clears the current share price only at $137.30 under dual bull shifts
$46.2$47.9$49.5$51.1$52.8$64.1$66.5$69.0$71.4$73.9$81.9$85.2$88.5$91.8$95.0$99.8$103.9$108.0$112.1$116.2$117.7$122.6$127.5$132.4$137.3FY27E compute shift-2 GW-1 GWBase (8 GW)+1 GW+2 GWSI $/W shift-$6/W-$3/WBase ($26/W)+$3/W+$6/W

Source: Penmark Research model (FY27E-FY40E shifts)

Exhibit 71: Bull skew and the forward roll lift our base DCF to our $123.66 target
$0$25$50$75$100$125S&C EVSI EVCorporateCash & dig.Debt & prep.Roll 13 NovBaseBull skewWeighted12m rollTarget$13.23+$78.20−$6.53+$6.36−$3.92+$1.15$88.49+$22.46$110.95+$12.71$123.66

Source: Penmark Research model

Exhibit 72: Five of six valuation stress tests push our 12-month rating from Hold to Sell
Price-implied / WACC testSpot $/sh12m PTCall
Implied WACC: 8.99% (−236bp vs 11.34%)136.63n/aPrice
Implied FY30E SI rent: $34.59/W (+$7.41)136.63n/aPrice
Discount rate −100bp (Ke 10.44%)105.01145.40Hold (+6.4%)
Discount rate +100bp (Ke 12.48%)75.66106.70Sell (−21.9%)
Beta / terminal / base testSpot $/sh12m PTCall
Peer-rule beta 2.04 (Ke 13.73%)63.4390.44Sell (−33.8%)
High-end peer beta 2.55 (Ke 15.89%)48.1870.00Sell (−48.8%)
Terminal RONIC cut to 12.8%86.52120.92Sell (−11.5%)
Base case alone (no bull skew)88.4998.63Sell (−27.8%)

Source: Penmark Research model. A reverse DCF solves for the cash-flow or discount-rate input needed to match today’s share price.

Exhibit 73: Probability-weighted scenarios yield our $123.66 target, just 0.5pp inside Hold
Scenario / reconciliation lineWeightEnd-27 GWFY28 $/WFY30 SI mgnSpot $/sh12m $/shReturn vs $136.63
Bear (supply glut & customer exit to $16/W)10%5.0 GW$23.0046.0%9.8410.97−92.0% (Sell)
Base (8.0 GW build; fade to $26/W & 55%)65%8.0 GW$29.3258.0%88.4998.63−27.8% (Sell)
Bull (CEO 12.7 GW plan; $28–34/W; faster V3)25%12.7 GW$33.0065.0%209.78233.82+71.1% (Buy)
Probability-weighted value & 12m target100%n/an/an/a110.95123.66−9.5% (Hold)
Rating band triggers (at $136.63 / at $123.66 PT)SI −$0.15/W/yr ($27.03/W FY30E = $87.53 spot) tips to SellSell <$122.97 PTBuy >$150.29 PTPrice: >$137.40 Sell / <$112.41 Buy

Source: Penmark Research model. Subjective scenario weights carry low conviction.

PART 4 · THE NUMBERS · 16 CATALYSTS

Seven dated catalysts through early 2028 test our 0.5pp Hold buffer

Our target sits less than $1 above the Sell threshold, so every catalyst over the next 15 months can move our rating. Below we map each dated milestone, compare our forecast against consensus or guidance, and name the stock driver at stake.

Exhibit 74: Seven major milestones through early 2028 test our 8.0 GW compute path
Nov-26Jan-27Mar-27May-27Jul-27Sep-27Nov-27Jan-28Mar-28SI Compute4Q26 SI build to 2.7 GWFY27 SI ramp to 8.0 GWEquity & M&ADay-180 lock-up expiry1-yr Musk & insider unlockConnectivityEchoStar spectrum closeStarship V3 & Mobile launch15 Nov 2026$37.5bn capex$40bn Apollo leaseUp to 798m shs6.4bn Musk shs$9.3bn cash + 262m shs

Source: Form 424B4; Q3 FY26 10-Q; 26 Oct 2026 Form 8-K; Q3 call

Exhibit 75: Dated milestones test our Buy and Sell rating thresholds through early 2028
DateEvent & stock driverPenmark vs Street / guideWhat moves our target or rating
8 Dec 2026180-day IPO lock-up release (Liquidity)Up to 328.4m–797.6m Class A eligible.Technical supply test (95.4m 3M avg daily vol); pullback below $112.41 crosses +10% to Buy.
Dec 2026Starship Flight 15 & Texas grid review (Compute & orbital)Dual booster/ship catch; no permit moratorium.−1 GW permit delay cuts DCF −$3.27/sh (PT <$122.97 = Sell); Flight 14 added +15.5%.
~25 Feb 20274Q26 results & 1Q27 guide (Compute & rent)We: 2.7 GW, $37.5bn capex, $103.1bn ARR, $33.9bn 1Q27 vs $48.8bn FY26 Street.GW below 2.5 or a 1Q27 guide below $33bn points to a slower build or softer rents, tipping us to Sell; above 3.0 GW supports the bull case.
May–Jun 20271Q27 lock-up & TeraFab test chip (Compute & rent)Up to 351.9m sh; TeraFab unvalued in base.First 90-day GB300 renewals: multi-year take-or-pay at ≥$30/W (rents held at $27/W add +$4.63/sh; SI margin held at 62% adds +$18.84/sh).
12 Jun 20271-yr IPO lock-up expiry (Governance / float)Up to 351.9m sh + 6.4bn Musk sh eligible.Watch any 10b5-1 filings or collateral pledges; ~7.8bn shares (~60%) restricted >1 yr at IPO.
Jul / Nov 20272Q27 GW print & EchoStar close (Compute & spectrum)We: ~5.35 GW avg FY27E; $9.33bn spectrum cash by 30 Nov.>0.65 GW/mo pace (+4.7 GW = +$15.38/sh) lifts our target to $134.80 (−1.3%; Hold); FCC delay hurts Space & Conn.
Early 2028FY27 10-K & Starlink Mobile launch (Rent & spectrum)We: 64.0% FY27E SI margin; commercial mobile launch.Confirms whether SI margin holds 62% (+$18.84/sh) or fades toward 55% vs Street 68.6% group.

Source: Form 424B4; Form 10-Q; Q3 call transcript; Penmark Research model.

Next numbers to watch (late-February release):Watch year-end nameplate compute against our 2.7 GW estimate and first-quarter revenue guidance near $34bn. Also track fourth-quarter Starlink net adds toward 15.5m subscribers and FY27 debt or lease spreads near 6% to 7%.
PART 4 · THE NUMBERS · 17 RISKS

A 50bp jump in discount rates cuts spot value by $6.80 per share

With our rating just 0.5pp above Sell, risks cut sharply in both directions. Small rent, grid or rate setbacks push the call to Sell, whereas bull-case execution creates +71% upside risk against a cautious view.

Exhibit 76: Six core risks span −$78.65 bear downside to +$121.29 bull upside per share
Risk categoryLikelihoodSpot $/sh impactMechanism & what breaks our Hold thesis
SI rent fade / Customer B exitMedium–High−$3.08 to −$78.65Customer B is 34.5% of 3Q sales; cloud deals carry 90-day exit rights. −$0.15/W/yr tips to Sell; bear glut ($16/W) hits $9.84/sh.
WACC / funding spread shockMedium−$6.80 to −$25.06$428.6bn FY27E–28E capex needs $40bn GPU leases + $50.7bn debt. +50bp WACC = −$6.80/sh; peer beta 2.04 = $63.43 spot (Sell).
Grid, water & turbine delaysMedium−$3.27 per −1 GWMulti-GW AI halls strain Texas/Tennessee grids, water & turbines. A −1.0 GW FY27E slip cuts spot to $85.22 (Sell).
Starship / FAA / FCC / AI rulesMediumUp to −$13.23Starship anomaly, FAA launch caps, FCC delay on EchoStar AWS-4 ($9.33bn) or European/US X & Grok fines impair $190.1bn Space & Conn. EV.
Key-person & controlled boardStructuralGovernance disc.Elon Musk holds 84.4% voting power (Class B 10x votes; elects 51% of board), leads Tesla/other firms, and SpaceX guarantees $20.21bn of Valor leases.
Upside risk: 12.7 GW bull build25% weight+$15.38 to +$121.29Securing +10 GW in FY27E at $33/W lifts base +$15.38/sh and full bull to $209.78 spot ($233.82 12m = +71%, Buy).

Source: Form 424B4 Risk Factors; Q3 Form 10-Q; Penmark Research model.

Exhibit 77: Faster compute adds $15.38 per share across our valuation sensitivity range
+4.7 GW 2027 compute10Y Treasury yield -50bpSI rent +$3/W FY28-30EFY27E SI rent +$2/WStarlink +1m/yr net addsStarlink -1m FY27E addsAnchor 90d cut -$3/WSupply wave -$3/WGrid/permit -1.0 GWFunding spread +50bp+$15.38+$7.71+$3.08+$0.39+$0.21−$0.11−$0.59−$3.08−$3.27−$6.80

Source: Penmark Research model (spot $/sh impact)

What ends the thesis & environmental, social and governance (ESG) exposure:Any two of FY30E rent below $27/W, end-FY27E compute below 7 GW or debt spreads over 100bp wider pushes our call to Sell. Permitting moves value directly via Electric Reliability Council of Texas (ERCOT) and Tennessee Valley Authority (TVA) power and water approvals plus launch-pad reviews.
PART 5 · APPENDIX · 18 OWNERSHIP AND TRADING

Top 10 institutions hold 2.47bn shares ahead of staged lock-up releases

Exhibit 78: Leading institutional filers held 2.47bn shares, or 32.1% of Class A stock
Rank & institutional holderShares (m)% Class A ref.% dilutedRank & institutional holderShares (m)% Class A ref.% diluted
1. Alphabet Inc.551.197.16%3.83%6. Public Investment Fund154.152.00%1.07%
2. Valor Management LLC503.416.54%3.50%7. BAMCO Inc. (Baron)145.781.89%1.01%
3. FMR LLC (Fidelity)302.563.93%2.10%8. D1 Capital Partners126.041.64%0.88%
4. VY Capital271.813.53%1.89%9. NVIDIA Corp.122.761.60%0.85%
5. Gigafund171.832.23%1.20%10. SC US (TTGP) / Sequoia122.521.59%0.85%
Top 5 subtotal1,800.8023.40%12.53%Top 10 / All 1,645 13F filers2,472.05 / 3,672.4132.12% / 47.72%17.20% / 25.55%

Source: SEC Form 13F (as of 30 Jun 2026; 7,696.3m Class A reference; 14,373.8m model diluted shares). Single post-IPO 13F snapshot; prior-quarter changes unavailable.

Exhibit 79: Short interest fell 32% from its July peak to 141.1m shares
0m50m100m150m200m250m15Jun30Jun15 Jul31 Jul14Aug31Aug15Sep30Sep15Oct30Oct23m111m165m208m184m174m162m181m167m141m

Source: Financial Industry Regulatory Authority (FINRA) short interest (1.22 days to cover at 30 Oct 2026)

Exhibit 80: The $256.41 Street mean target sits at 2.1x our target
Street low $70Price $137Street high $800Penmark $124Street mean $256

Source: Investing.com and price-target.com (~8 Nov 2026; 32 Buy, 2 Sell)

Exhibit 81: Average daily turnover of $14.05bn faces roughly 7.8bn unlocking shares
Trading / float itemDisclosed facts & market impact
Liquidity & trading range3M average daily volume: 95.43m shares ($14.05bn/day). Intraday range since 12 Jun listing: $104.83–$225.64 (closing range $108.27–$211.39; price $136.63, 13 Nov last trade).
Short interest (30 Oct)141.05m shares short (−15.7% vs 15 Oct; down 32.1% from 207.78m 31 Jul peak), equal to 1.22 days to cover.
Insider activityPresident/COO Gwynne Shotwell sold 342,170 shares on 22 Sep 2026 ($151.15–$154.72) alongside option exercises under a pre-arranged Rule 10b5-1 plan; no open-market sales by CEO Elon Musk.
Staged lock-up schedule8 Dec 2026 (180d): up to 328.4m–797.6m Class A; post-4Q26 results: up to 351.9m; 18 Mar 2027: up to 176.0m; post-1Q27 results: up to 351.9m; 12 Jun 2027 (1yr): up to 351.9m + 6.4bn Musk shares (~7.8bn shares / ~60% restricted >1 yr at IPO).
Index & Street ratingsDual-class structure (10x Class B votes) limits S&P 500 eligibility under current governance rules; Nasdaq-listed. Street: 32 Buy / 2 Sell (Hold count unpublished; KeyBanc Hold 21 Oct), mean PT $256.41.

Source: Form 424B4; Form 4; FINRA short interest; yfinance; Penmark Research model.

Float & Street positioning takeaway:Roughly $14bn of average daily turnover and 1.2 days to cover let the stock absorb routine institutional flow. Yet about 7.8bn restricted shares (~60% of shares out) unlock through June 2027. Meanwhile, 32 of 34 rated brokers carry Buy calls.
PART 5 · APPENDIX · 19 FINANCIAL MODEL

Group free cash flow reaches $99.7bn in FY32E even after server replacements

Exhibit 82: Operating income scales from $0.5bn in FY24A to $226.5bn by FY32E
Line item (US$m)FY24AFY25AFY26EFY27EFY28EFY29EFY30EFY32E
Revenue14,01518,67449,408196,101343,798455,244541,131671,101
Adjusted EBITDA5,3506,58426,402121,852210,062270,740312,344370,587
D&A + total SBC + other op. items(4,884)(9,173)(20,857)(56,252)(88,829)(106,769)(116,652)(144,073)
GAAP operating income (EBIT)466(2,589)5,54565,600121,234163,971195,692226,514
Net income / Diluted EPS ($/sh)18 / n.d.(4,937) / n.d.4,717 / 0.2160,451 / 4.2497,915 / 6.81127,890 / 8.90153,126 / 10.89177,828 / 13.67
Cash from operations / Capex5,776 / (11,163)6,785 / (20,737)24,509 / (85,569)127,297 / (228,344)199,577 / (200,262)241,156 / (168,634)280,084 / (153,691)328,982 / (229,315)
Free cash flow (ex prepayments memo)(5,387)(13,952)(61,445) / (66,622)(110,375) / (116,614)(685) / (11,533)72,522 / 70,483126,393 / 119,52299,666 / 94,946
Cash & securities / Gross debt & leasesn.d.n.d.63,382 / 38,84430,000 / 115,83730,000 / 116,52343,844 / 57,84430,000 / 49,84430,000 / 38,844
Net PP&E / Shareholders’ equityn.d.n.d.122,185 / 185,014308,360 / 269,948435,589 / 382,958511,259 / 523,953559,820 / 555,665711,678 / 701,526

Source: Form 424B4; Q3 Form 10-Q; Penmark Research model. Balance sheet rolls from 30 Sep 2026 ($89,855m cash, $38,844m debt, $176,768m equity).

Exhibit 83: Quarterly revenue climbs from $4.1bn in 1Q25A to $22.3bn in 4Q26E
Quarterly metric (US$m)1Q25A2Q25A3Q25A4Q25A1Q26A2Q26A3Q26A4Q26E
Space revenue8657461,6668096199621,7671,300
Connectivity revenue2,4752,5873,0973,2283,2574,2915,2485,900
SI revenue7277388139238182,5617,56015,125
Group revenue4,0674,0715,5764,9604,6947,81414,57522,325
Group Adjusted EBITDA1,7301,2142,1691,4711,1273,5388,80912,928
GAAP net income(528)(1,008)(1,054)(2,347)(4,276)(541)4,4045,130
Capex4,1402,8256,8816,89110,10718,36919,59337,500
Starlink subs (m) / SI GW5.0 / 0.3n.d.7.2 / 0.68.9 / 0.810.3 / 1.012.0 / 1.413.6 / 1.715.5 / 2.7

Source: Form 424B4; Q2 & Q3 Form 10-Q; Penmark model. Full quarterly statements beyond 4Q26E not modelled beyond 1Q27E $33,871m revenue check.

Exhibit 84: Forward P/E compresses from 32.2x in FY27E to 12.5x by FY30E at spot
Ratio / per-share metricFY24AFY25AFY26EFY27EFY28EFY29EFY30E
Average diluted shares (m; buybacks from FY30E)n.d.n.d.14,112.014,242.914,373.814,373.814,060.3
ROE / ROA / ROCE (%)n.d.n.d.2.6 / 2.0 / 2.526.6 / 18.5 / 21.530.0 / 20.7 / 27.428.2 / 22.1 / 30.328.4 / 24.1 / 33.0
Net debt / equity (%) & Net debt / EBITDA (x)n.d.n.d.−13.3% / −0.93x31.8% / 0.70x22.6% / 0.41x2.7% / 0.05x3.6% / 0.06x
P/E / EV/EBITDA at $136.63 (Div. yield 0.0%)n.m.n.m.639.9x / 73.4x32.2x / 15.9x20.1x / 9.2x15.4x / 7.2x12.5x / 6.2x

Source: Penmark Research model. FY26E ROE/ROA/ROCE use 30 Sep–31 Dec average assets/equity; EV multiples use $1,938.5bn spot EV.

Explicit bridge to terminal value: From FY32E through FY40E, group revenue compounds to about $1.0tn and unlevered FCF reaches roughly $269bn. SI reaches 35.5 GW and Connectivity hits 68m consumer subscribers by FY40E. Terminal cash flows then grow 3% at a 15% RONIC.
PART 5 · APPENDIX · 20 REFERENCE

Non-cash charges create a $20.9bn FY26E wedge between Adjusted EBITDA and GAAP EBIT

Exhibit 85: GAAP net income reaches $60.5bn in FY27E after $56.3bn of D&A and equity pay
US$mFY23AFY24AFY25A9M26AFY26EFY27EFY28E
Adjusted EBITDA3,8215,3506,58413,47426,402121,852210,062
Less: Depreciation & amortisation(2,635)(3,824)(6,701)(8,934)(13,907)(42,869)(73,734)
Less: Ongoing + Cursor acquisition SBC(679)(784)(1,947)(3,833)(6,949)(13,383)(15,095)
Less: Restructuring, impairment & M&A(4,012)(276)(525)(2)000
GAAP operating income (EBIT)(3,505)466(2,589)7055,54565,600121,234
Net interest, other & income tax(1,123)(448)(2,348)(1,118)(828)(5,149)(23,319)
Reported (GAAP) net income(4,628)18(4,937)(413)4,71760,45197,915

Source: Form 424B4; Q3 Form 10-Q; Penmark model.

Glossary & ratio definitions

  • Nameplate compute (GW). Energised information technology (IT) power capacity across SI GPU clusters (~1.7 GW at September 2026).
  • SI yield ($/W/yr). Annual SI revenue divided by average nameplate watts, fading from about $30.5/W to $26/W by FY32E.
  • Starlink V3 & AWS-4. Next-generation ~1 Tbps Starship-deployed LEO satellites and EchoStar mobile spectrum closing by November 2027.
  • Customer prepayments. Upfront deferred revenue (~$17.5bn at September 2026), excluded from valued FCF and deducted in EV.
  • Ratios. ROE = net income / avg equity; ROA = net income / avg net assets; ROCE = EBIT / avg capital employed; Net debt = debt & GPU leases less cash/securities.

Data sources & reporting currency

  • Model & currency: US$ reporting (US issuer, no American Depositary Receipt, ADR); Penmark Research model with market data to 13 Nov 2026.
  • Filings & KPIs: SEC Forms 424B4, 10-Q and 8-K plus the October Q3 earnings call; KPIs checked against the 10-Q.

Methodology & model limitations

  • Multi-segment archetype. Combines Space (backlog), Connectivity (subscriptions) and SI (capacity × yield) for a controlled dual-class issuer listed in June 2026.
  • Departures. With five months of trading history, we drop 12-month stock series, use a 1.50 beta, and deduct ongoing SBC and customer prepayments.
  • Known limits. Base 8 GW FY27E compute sits below management’s +10–20 GW ambition. Segment PP&E uses historical capex share (~30% Space & Conn.), and cash tax ramps 8% to 21% without net operating loss (NOL) vintage tracking.
  • Unvalued options. Zero DCF value assigned to a potential Tesla tie-up, TeraFab, orbital compute or lunar and Mars Starship.
Exhibit 86: Peer medians of 8.93x for SI and 11.62x for Connectivity anchor our cross-check
ComparableRole in Penmark modelEV ($bn)FY27E EV/EBITDAEBITDA mgnBeta
CoreWeave (CRWV)SI multiple & beta comparable (AI GPU cloud)76.04.54x63.2%2.06
Nebius Group (NBIS)SI multiple & beta comparable (SI median 8.93x)70.413.32x54.6%2.59
T-Mobile US (TMUS)Space & Conn. multiple & beta comparable (wireless carrier)284.07.08x40.6%0.30
Iridium (IRDM offer)Space & Conn. multiple comparable (median 11.62x)8.016.16x (LTM)56.8%1.05*
RKLB / ASTS / VSATBeta-only space & satellite cohort (peer-rule beta 2.04)n.m.n.m.n.m.2.50–2.92

Source: MarketScreener; Multiples.vc; 24/7 Wall St; Penmark model. LTM = last twelve months. *IRDM excluded from beta set (under offer); GSAT excluded (Amazon deal).

Reporting conventions: US$; fiscal year ends 31 December; market data to 13 November 2026.
PART 5 · APPENDIX · 21 DISCLOSURES

Analyst certification and important disclosures

Ratings

Ratings are based on the upside from the share price to our 12-month price target, excluding dividends. The price target is weighted across bull, base and bear scenarios.

BuyUpside to price target above +10%
HoldUpside to price target from −10% to +10%, inclusive
SellUpside to price target below −10%

We review ratings at each set of results and whenever the facts change materially. A share price can move outside its band before the rating changes, and targets may not be reached.

Analyst certification

The Penmark Team certifies that (1) the views in this report accurately reflect its views about the subject company and its securities, and (2) no part of its compensation was, is or will be directly or indirectly related to the specific ratings, price targets or views in this report.

Contact

Penmark Research

Questions about our research:
research@penmarkresearch.com

All our research and coverage:
penmarkresearch.com

Basis of preparation

This report uses information publicly available up to its publication date. Share prices are closing prices on the trading day before the publication date, unless stated otherwise. Forecasts, valuations and price targets are Penmark Research’s own, and forecast periods are marked E. Sources are shown beneath each exhibit.

Valuation and price targets

Price targets are based on the valuation methods set out in each report and are weighted across bull, base and bear scenarios. The main risks to each rating and price target are described in the report. Penmark Research updates its coverage after each set of results but is not obliged to update a report at any other time.

Independence and conflicts of interest

Penmark Research is independent. It does not provide investment banking, brokerage, market-making or asset management services, does not trade securities, and does not accept payment from any company for research coverage. Companies do not see or review Penmark Research reports before publication.

No member of the Penmark Team holds a long or short position, directly or through derivatives, in the securities of any company Penmark Research covers.

Regulatory status

Penmark Research publishes general, impersonal investment research on a regular basis. It is not registered as an investment adviser with the US Securities and Exchange Commission or any state securities authority, is not a registered broker-dealer and is not a member of FINRA. Penmark Research does not give advice tailored to any person’s circumstances.

Not investment advice or an offer

This report is for information only. It is not personalised investment advice and is not an offer or solicitation to buy or sell any security or other financial instrument. It does not take into account any reader’s objectives, financial situation or needs. Readers should make their own investment decisions and seek independent professional advice where appropriate.

Forward-looking statements and risk

Forecasts, valuations and price targets are estimates based on assumptions that may prove wrong, and they may not be achieved. Opinions are current as of the publication date and may change without notice. Investing in equities involves risk, including the loss of capital. Share prices can be volatile, and past performance is not a reliable indicator of future results.

Data and sources

Company financials, operating data and filings come from company filings, earnings releases and earnings calls. Market data and any consensus estimates come from third-party sources named beneath the relevant exhibit; consensus figures from aggregators are unofficial. Penmark Research does not use other brokers’ research. Third-party data belong to their providers and may not be extracted from this report.

Accuracy and liability

Information is taken from sources believed to be reliable, but Penmark Research makes no representation or warranty as to its accuracy, completeness or timeliness. To the fullest extent permitted by law, Penmark Research accepts no liability for any loss arising from use of this report. Descriptions of litigation and regulatory matters reflect allegations and public records, not findings, unless stated. If a published report is corrected, the corrected version carries an “Updated” date at penmarkresearch.com.

Distribution and access

Penmark Research reports are available to registered users of penmarkresearch.com. Downloads and programmatic access are governed by the Penmark Research subscription terms. This report is not directed at, or intended for, any person in Singapore or India, and access from those countries is blocked. It is not intended for distribution to, or use by, any person in any jurisdiction where that would be contrary to law or regulation or would subject Penmark Research to registration or licensing requirements there.

Copyright and permitted use

© Penmark Research. All rights reserved. Penmark Research is a trade name of Brightriver Labs. Readers may share a link to this report but may not copy, reproduce, redistribute, sell or use its content, including for text and data mining, without written consent. Contact: research@penmarkresearch.com.

Keep reading

Sign in to Penmark