Shaun Maguire's SPCX bull case, rebuilt with sliders
On Oct 30, 2026 Shaun Maguire (Sequoia) published a quarterly model of SpaceX through 2028. His AI business runs off three numbers: gigawatts in service, revenue per gigawatt-year and capex per gigawatt added. This page rebuilds his sheet quarter by quarter, so at his inputs it reproduces his revenue and capex exactly, then lets you move those three numbers. The Maguire tick under each slider marks his figure; click it to jump back, or use the button at the top of the slider box to reset every input to his.
Source. Shaun Maguire, "People don't understand SpaceX", posted on X on Oct 30, 2026. The quarterly figures marked "Maguire" are read off the model image in that post.
His caveats, in his words (three separate passages of the post, quoted exactly):
- Opening line:
Caveats: this is not investment advice. Sequoia Capital has a large position in SpaceX so I'm extremely biased. Also, Past performance is not indicative of future results.
- With the model:
There are wide error bars in certain areas, especially: revenue per gigawatt, CapEx per gigawatt and Terafab CapEx. But this is my 50th percentile forecast.
- Conclusion:
This is a very short post. I didn't justify most of my numbers.
Ours. This is an independent, fan-built model, not Maguire's forecast and not investment advice. It is not affiliated with or endorsed by Shaun Maguire, Sequoia Capital or SpaceX. Everything beyond his sheet (depreciation, stock comp, interest, tax, shares and EPS) is our extension and is labelled as such. The authors may hold positions in the securities discussed.
His published figures against the sliders — $B; the Maguire columns are fixed, the "by the sliders" columns follow the sliders
| 2026 · Maguire | 2026 · by the sliders | 2027 · Maguire | 2027 · by the sliders | 2028 · Maguire | 2028 · by the sliders |
|---|
Revenue per GW-year
Applies to GW in service at the start of the quarter. Maguire steps one fleet rate $30B, $35B, $35B, $40B, $40B, then $40B through 2028; the sliders scale that path. Maguire calls rev/GW and capex/GW "coupled constants"; here they move separately.
Capex per GW added
Maguire steps it $45B, $40B, $35B, $35B, $30B, then $30B through 2028; the sliders scale that path.
Fleet at year end
Terrestrial GW in service, from Maguire's 1.7 GW on Sep 30, 2026. Within each year the GW added per quarter keeps his shape (2027: 1.5, 1.8, 2.0, 2.4; 2028: 2.9, 2.5, 2.4, 3.0), scaled to the year's total. Maguire: 2.5, 10.2, 21.0.
Margin, the 1.7 GW already running, a price cut, and the fixed parts of his sheet
Maguire: 60% in Q4 2026, 70% in Q1 2027, 75% from Q2 2027.
Maguire prices it at the fleet rate. This option is ours and off by default: it gives the old fleet its own rate, so the new fleet's rate doesn't apply to it. The starting rate of $16.67B per GW-year is our figure, not his.
A bear case Maguire's coupling rules out: the whole installed fleet is repriced down from Jan 2028 while build cost and operating cost stay put, so revenue and EBITDA both fall by the cut. He writes that "it's very unlikely we see anything below $20B/GW within the next ~4 years"; at his $40B, a 50% cut is $20B.
Held fixed from his sheet, quarter by quarter: other AI revenue , calibration to his totals (his SI total less his terrestrial compute line); Communications + Space revenue, operating income and segment EBITDA; Space + Connectivity capex .
Charts
Earnings and EPS our extension — not in Maguire's sheet
Maguire publishes operating income and change in cash, not net income or EPS, so tax, share count and EPS here are ours. With both boxes off, depreciation and stock comp are his implied figures and tax is a 21% book rate on his operating income. His unlabelled cash item looks more like working capital than tax, so it moves cash only, not earnings. Full Depreciation on swaps in our useful-life depreciation and stock comp. Full Tax on adds interest and our cash-tax rate. EPS is also shown on his published operating income, which doesn't follow the sliders; at his inputs with both boxes off, the two match. EPS is to one decimal.
Non-cash, deducted before tax. 5% is our assumption. The closest comparison is the SI (AI) segment's own: in Q3 2026 its stock comp was 10.9% of SI revenue ($821M of $7.560B), or 27.0% including the $1.217B tied to the Cursor acquisition, all booked in SI. Company-wide it was 7.9% of revenue ($1.146B of $14.575B), or 16.2% including the Cursor acquisition's $1.217B (Q3 2026 results, SI segment table and adjusted EBITDA reconciliation).
Straight line on each quarter's AI capex from the next quarter, plus $2.0B a quarter for the pre-2026 fleet (ours) and his $40.05B of Jan–Sep 2026 SI capex.
On each quarter's average net debt, paid in cash. Off by default: net cash earns nothing.
Book tax for EPS is a flat 21%. Cash tax is separate: his change-in-cash line takes out more than EBITDA less capex ($4.7B in Q4 2026, $28.8B in 2027, $52.3B in 2028), unlabelled. At 14.5% on our pre-tax income our end-2028 net cash lands next to his. With Full Tax off, cash tax and interest are replaced by those implied amounts, as his share of revenue each quarter, for cash only; EPS still takes the 21% book tax.
| $B except per share | 2027 · his operating income | 2027 · our bridge | 2028 · his operating income | 2028 · our bridge |
|---|
The "his operating income" columns are fixed at his published figure (they don't follow the sliders), deduct no interest (his sheet shows none) and apply our 21% book tax and share count. "Our bridge" uses this model's own depreciation, stock comp and interest. Shares: 13.58B basic at Sep 30, 2026, × 1.0446 (Q3 2026 weighted diluted ÷ basic, 12,475M ÷ 11,942M) = 14.19B diluted, +2% a year, at year end.
Full year, this model
By quarter — $B except GW and rates
| Quarter | GW start | GW added | GW end | Rev / GW-yr | Capex / GW | Terrestrial rev | AI revenue | Revenue | AI capex | Other capex | AI depreciation | Cash tax / other | Interest | FCF | Maguire Δ cash | Net cash (debt) | Maguire net cash (debt) |
|---|
Notes
Inherited from his sheet
- His 2028 AI (SI) lines don't add to his SI total: Grok + Cursor + other $76.1B, terrestrial compute $569.0B, orbital compute $31.0B and Terafab $5.2B sum to $681.3B, against his SI total of $625.2B. The gap opens in Q1 2028. This model uses his total; the "other AI revenue" line is his total less his terrestrial line.
- Orbital compute (2.8 GW and $31.0B of revenue in 2028) and Terafab carry no capex in his SI capex line, which is exactly terrestrial GW added × capex per GW every quarter. Some may sit in his Space capex; his sheet doesn't say.
- His change in cash is below EBITDA less capex by $4.7B in Q4 2026, $28.8B in 2027 and $52.3B in 2028, with no label. It may be tax, interest or working capital; we can't tell. With Full Tax off it moves cash only; with it on, the cash-tax default above absorbs it.
- His SI segment carries $42.5B of charges between EBITDA and operating income in 2028 ($468.9B against $426.5B), against about $650B of AI capex across 2026–28. With Full Depreciation on, our depreciation runs higher, so our operating income is lower than his; both are shown above.
- His sheet has no balance sheet. "Maguire net cash (debt)" is our reading: his change in cash applied to the Sep 30, 2026 net cash of $51.0B ($89.9B cash and securities less $38.8B debt, Q3 report).
Dollars in billions; negatives in brackets. Maguire's model of Oct 30, 2026 is quarterly from Q1 2026 to Q4 2028; this page starts at Q4 2026 and, at his inputs, reproduces every quarter of his GW, revenue per GW, capex per GW, terrestrial revenue, SI revenue, total revenue and total capex. Sliders scale his quarterly paths rather than replacing them. Communications + Space, other AI revenue and non-AI capex are his quarterly lines, held fixed in dollars. Depreciation, stock comp, interest, tax, shares, net income and EPS are ours. Cash starts at Sep 30, 2026: net cash $51.0B. Peak net debt is read at quarter ends. Independent and fan-built; not investment advice; not affiliated with or endorsed by Shaun Maguire, Sequoia Capital or SpaceX.