Guidance for $33 to $37 billion of revenue in the first quarter of 2027 is spectacular. The price already needs about 9 gigawatts of rented-out AI compute next year, paid for with borrowed money. Target $148.09, a hold.
Key points
- $35 billion of guided revenue for Q1 2027 (midpoint) is more than double Q3 2026's $14.6 billion.
- $35 to $40 billion of Q4 capex, money spent on equipment, is about twice Q3's.
- $148.09 is my 12-month target, 4% below the $154.14 price.
The call
The price needs the AI build-out to land almost on plan. At $154.14 the company is worth about $2.2 trillion. My model reaches that value only if about 9 gigawatts of AI compute is earning money on average through late 2027. That is roughly three times the year-end 2026 goal.
Quarterly revenue and capex ($ billions)- Revenue, reported
- Revenue, my estimate
- Capex (Q4 2026 guided midpoint)
Source: SpaceX prospectus, 10-Q, Oct 20 2026 release and call; Q4 2025 not retrieved; E = my estimates
I think the market is about right, and the results were real. Q3 beat my growth path, and rented compute is earning more per watt than before. But capex will run far ahead of cash flow. I expect net debt near $166 billion by October 2027, and compute is rented on short contracts that customers can end.
Value per share by average AI compute in service, Oct 2027 ($)My model: $34B revenue per GW, 55% EBITDA margin, 10x; other parts held fixed
Valuation
| Item | Figure |
|---|
| AI compute revenue, year to Oct 2027 (9 GW at $34 billion each) | $306 billion |
| AI earnings before interest, tax and depreciation at 55% margin, times 10 | $1.7 trillion |
| Starlink at 25 times $19 billion of earnings, plus launch at $150 billion | $630 billion |
| Net debt, Oct 2027 | $166 billion |
| Diluted shares | 14.5 billion |
| Value per share | $148 |
Ten times earnings fits rented hardware that wears out in about five years on short contracts. Each 1.5 gigawatts, on or off plan, moves the value about $19 a share. A turn of multiple moves it about $12. The upside and downside are about even, so the stock should track the market, with wide swings.
Risks
- Power or chip delays slow the AI build; watch nameplate compute against 2.5 to 3 gigawatts at year end.
- Rental prices for AI chips fall or a big cloud customer leaves; watch backlog against $63 billion and renewal pricing.
- Upside: a Tesla merger or 13 gigawatts by end 2027 lifts the value toward $190; watch the special committees and the Q4 report.
What would change my mind
Q4 revenue near $25 billion with operating cash flow covering half of capex would move me to a buy.
On the other analysts' notes
I am keeping my hold and my $148.09 target. Of the other four notes, the one I took most seriously is from Gemini, backed by Muse. Their case is that customers can cancel the cloud deals on 90 days' notice, which leaves the company spending far more cash than it brings in. I agree with that, and it is already a bigger cost in my numbers than in theirs. I expect debt to exceed cash by about $166 billion by October 2027. Gemini expects $10 billion. Muse and Grok both expect the company to still hold more cash than debt. So debt cannot be what separates us. What separates us is the method. They price the company at 10 to 14 times sales. That treats rented AI computing power, which earns a profit of about 55 cents per dollar of sales before interest, tax and depreciation, the same as a business with thin margins. On a profit basis, even after the heavier debt, the value comes out near today's price, not 20 to 26% below it. Grok says a $35 billion quarter is not a $140 billion year. That is fair as a warning. But guidance more than doubled from Q3 to Q1, so the ramp is running ahead of plan, not behind it. That guidance is also why I won't call this a strong sell. GPT's point is that rental prices are rising, and CoreWeave's increases back it up. I had already raised my revenue per watt for that. GPT's buy then relies on only $30 billion of net debt against $155 billion of spending, and that understates the funding gap I expect. No one gave me a fact I had not weighed. The arguments on each side roughly cancel out, so the stock should track the market. The Q4 report will settle it: compute switched on against the 2.5 to 3 gigawatt goal, and operating cash flow against capex.
