SPCX Slides 43% From Peak as SpaceX AI Capex Risk Divides Investors

SpaceX AI capex risk is the central tension in the SPCX investment case, even as second-quarter revenue nearly doubled year-on-year to $7.8 billion. The stock trades 43% below its June peak, yet Interactive Investor ranked SPCX as the second most-bought equity on its platform in July, and AJ Bell reported it topping its buy charts over the past month.

Those buying the dip are not wrong to see an opportunity. The Q2 numbers are genuinely strong across most lines. The question is whether the capex trajectory, the defining component of SpaceX AI capex risk, will consume returns faster than the revenue can grow into them.

What SpaceX’s Q2 2026 Results Actually Show

According to the SpaceX Q2 2026 earnings release, revenue of $7.8 billion represented 92% growth from $4.1 billion in the prior-year quarter. Adjusted EBITDA of $3.5 billion was up 191% from $1.2 billion. The net loss narrowed to $541 million, an improvement of $467 million from a loss of $1.0 billion a year earlier. SpaceX ended the quarter with $100 billion in cash, cash equivalents and marketable securities, alongside a backlog of $47.5 billion.

The Connectivity segment, which houses Starlink, delivered 66% revenue growth and 79% income from operations growth year-on-year, driven by a doubling of Starlink subscribers. SpaceX was also awarded over $6 billion in multi-year US government contracts for its Starshield defence programme. AI segment revenue reached $2.6 billion, up 213% sequentially and 247% year-on-year, primarily driven by new cloud service agreements and growth in Grok and X subscriptions.

For context, full-year 2025 revenue was $18.7 billion against a net loss of more than $4.9 billion. The pace of improvement is real; the profitability gap is real too.

SpaceX AI Capex Risk and the Google Cloud Anchor

The AI segment’s growth has substantial contractual backing. In June 2026, SpaceX entered into a Cloud Service Agreement with Google LLC covering access to approximately 110,000 NVIDIA GPUs and related infrastructure. Under the agreement, Google has committed to pay SpaceX $920 million per month from October 2026 through June 2029, with capacity ramping through September at a reduced rate. Cloud service agreements had generated $14.1 billion of contracted sales in aggregate by the end of Q2.

The delivery risk is concrete. If SpaceX fails to provide the committed GPU capacity by 30 September 2026, Google may, after a one-month grace period, terminate the agreement or accept a reduced allocation at a proportionally lower fee. After 31 December 2026, either party may terminate on 90 days’ notice. That termination clause is a live variable for the rest of this year.

On expenditure, Q2 capital spending came in at $18.37 billion, marginally below analyst consensus of $18.58 billion. JPMorgan analyst Doug Anmuth has projected capex approaching $200 billion in both 2027 and 2028. Musk has stated the annual revenue run rate could reach $100 billion by year-end, though his forward projections carry a well-documented optimism premium. That capex-to-revenue ratio, if Anmuth’s numbers prove accurate, is what makes SpaceX AI capex risk the central valuation debate.

Cursor, Starmind, and the Physics of Orbital Computing

The most consequential recent corporate action was the acquisition of Anysphere, the company behind the Cursor AI coding tool. SpaceX completed the merger on 14 August 2026, with Cursor surviving as a wholly owned subsidiary. The implied equity value of Cursor was $60.0 billion, settled primarily through the issuance of 389,289,254 shares of SpaceX Class A common stock to Cursor shareholders, plus additional shares converted from vested restricted stock units.

Starmind, SpaceX’s planned constellation of orbital data centres, is the longer-dated bet. The company has applied to the FCC for authority to operate up to one million satellite data centres. Nvidia confirmed on 24 August 2026 that the first-generation Starmind satellite will be based on an optimised version of its Vera Rubin NVL72 architecture. On timing, Musk has targeted Q4 2027 for Starmind AI1 launches; SpaceX’s own SEC filings describe deployment beginning ‘as early as 2028,’ while the company’s CFO suggested demonstrations could be possible in 2027 with meaningful orbital compute targeted for 2029.

The economics present a harder constraint. A July 2026 policy brief by Bain found that satellite data centre launches need to cost between $50 and $100 per kilogram to be economical. SpaceX’s Falcon Heavy can currently only launch payloads at approximately $1,500 per kilogram, a gap of an order of magnitude that only a fully operational Starship can close. The Starmind AI1 satellite is also too large to launch on Falcon 9.

One calendar marker worth tracking: the 180-day lock-up period tied to SpaceX’s S-1/A prospectus, filed 3 June 2026, expires around 30 November 2026. Insider supply arriving near that date could add downward pressure on a stock already trading at a stretched multiple.

The immediate binary is the Google GPU delivery deadline of 30 September 2026. Clean delivery reinforces the AI revenue thesis and shores up the contracted backlog. A shortfall hands Google a termination option and gives the market a concrete reason to reprice the AI segment from the top down.