Space Exploration Technologies (SPCX) closed Friday at $136.97, barely above the $135 price of June's record IPO. The stock has traded as high as $225.64 and as low as $104.83 in roughly ten weeks as a public company. Two forces explain that range. One is a share-supply schedule that has repeatedly outrun demand.
The other is a capital expenditure line that now runs well above revenue. Terafab, the Grimes County semiconductor complex SpaceX confirmed on August 6, sits at the intersection of both. It is the most ambitious piece of the story and the least contractually binding. Investors should price it accordingly.
The SPCX Spending Problem is Already on the Income Statement
SpaceX reported second-quarter revenue of $7.81 billion — up 92% year over year and about $1 billion ahead of consensus — while the net loss narrowed to $541 million and adjusted EBITDA landed at $3.5 billion.
Those are strong numbers. The capital expenditure figure is what moved the stock.
| Q2 2026 Metric | Result | Context |
| Revenue | $7.81B | Up 92% year over year |
| Capital expenditure | $18.37B | Roughly 2.4x quarterly revenue |
| AI capex | $15.83B | 86% of total capex |
| Net loss | $541M | Narrowed by $467M |
| Cash and securities | $100B | Post-IPO and post-bond |
| Long-term debt and finance leases | $36.8B | Up from $22.0B at year-end 2025 |
| Backlog | $47.5B | Company disclosure |
Management indicated the next two quarters should look similar on capex. Analysts now model full-year spending above $45 billion. Against a roughly $23 billion trailing revenue base, that is an extraordinary ratio. Only one segment carries the load. Connectivity delivered $4.3 billion in revenue on 66% growth and is the profit engine.
Starlink now counts 12 million subscribers, yet average revenue per user has slipped to $66 from $85 a year ago. Subscriber growth is real, but pricing is moving the wrong way — and that matters when this is the segment funding everything else. The AI segment grew revenue by 247% to $2.6 billion. It is also where the $15.83 billion went. Investors are being asked to underwrite a compute buildout on the promise of returns that have not yet appeared in operating income.
Terafab is a 2028 Story Sold in 2026
The confirmed facts are narrow. SpaceX and Tesla (TSLA) plan to build a vertically integrated fab complex in Grimes County, Texas, with a first-phase commitment of roughly $16.8 billion. The finished campus is meant to span more than 100 million square feet and employ at least 3,000 people, and a May regulatory filing floated total potential investment as high as $119 billion.

Texas contributed a $30 million Enterprise Fund grant, and Grimes County approved a full property tax abatement paired with fixed annual payments. Now, the caveats, which have received far less coverage. SpaceX's own IPO prospectus describes only a "general framework for the future development of Terafab."
Financial terms, intellectual property rights, timelines, and capital commitments were all left to future negotiation. The risk factors go further and state that neither Tesla nor Intel is obligated to stay in the project. That language came from the issuer, not from a skeptic. The process technology is Intel's (INTC) 14A node. Intel has committed to high-volume 14A production in 2028.
Elon Musk has said 14A is state of the art and not yet complete. Neither company has disclosed the commercial structure. Musk has never used the word licensing. Analysts at Yole Group have argued the arrangement may function more as an Intel foundry expansion with SpaceX, Tesla and xAI as anchor customers than as a standalone SpaceX-owned venture. That distinction determines who books the revenue and who carries the depreciation.
Scale expectations also deserve scrutiny. Yole's memory analysts estimate that satisfying Terafab's stated demand would require on the order of 10 million wafer starts per month for memory alone. Published capacity targets for the site are a fraction of that. The gap suggests either the demand figure is aspirational or the facility plan is incomplete. For now, Tesla's near-term silicon still comes from Samsung's Taylor fab and from Taiwan Semiconductor (TSM), and Terafab does nothing to change that through 2026 or 2027.
Supply of Shares, Not Supply of Chips, is the Near-Term Driver
The more immediate variable is float. Roughly 911 million shares became eligible for sale on August 6, a block larger than the existing tradable float. A further tranche of about 319 million shares was unlocked on August 20 and pushed the stock back below its IPO price. Additional releases are scheduled for December. That the stock absorbed both events without collapsing is telling. Nvidia (NVDA), for its part, disclosed a stake of roughly $21 billion at quarter-end, and institutional buying has so far mopped up the insider supply.
| Near-term variable | Status | Why it matters |
| Share unlocks | August tranches absorbed; December pending | Determines whether the float overhang is priced in |
| Starlink ARPU | $66, down from $85 | The only profitable segment is monetizing less per user |
| Quarterly capex | Guided flat near term | Sets the cash burn floor through 2027 |
| Terafab structure | Undisclosed | Decide whether SpaceX or Intel carries the fab economics |
The SPCX Investment Case: Bull vs. Bear
The bull argument is coherent. SpaceX holds a durable launch cost advantage, a subscriber business compounding at 66%, and $100 billion of liquidity to fund an aggressive compute strategy. Management has verbally targeted a $100 billion annualized revenue run rate by December and pulled forward its long-range revenue ambitions. Wall Street consensus remains constructive, with an average target well above the current price. The bear argument does not require disputing any of that.
It requires only noting what the numbers already show — capex running at 2.4 times revenue, ARPU sliding in the one segment that actually earns money, verbal ambitions standing in where formal guidance belongs, and a flagship factory that is non-binding on paper and years from producing anything. None of it contradicts the bull case. It simply sits underneath it.
For investors, the practical implication is to separate the two clocks. Terafab is a long-dated call option on vertical integration, and it should be valued as one. The next four quarters of SPCX will be decided by Connectivity margins, the December unlock, and whether AI segment revenue starts closing the gap with AI segment spending.
What SPCX Holders Should Watch In Q3?
Watch the Q3 report for two specific items. First, whether capex holds at the guided level or drifts higher. Second, whether Starlink ARPU stabilizes. Those two lines will tell you more about the stock than any rendering of a fab in Grimes County.
By Udi Jacoby, 18 Years of Financial Market Experience & Trade The Pool Senior Analyst