SpaceX (SPCX) Q2 2026 Earnings Call Transcript

SpaceX posted Q2 revenue of $7.8B (+92% YoY) and narrowed net loss to $541M (improves by $467M YoY), while adjusted EBITDA surged to $3.5B (+191% YoY). Starlink subscribers doubled to 12M and connectivity revenue reached $4.3B (+66% YoY, ARPU $66/mo stable), while AI segment revenue rose to $2.6B (+247% YoY) as incremental AI infrastructure revenue reached $1.6B from new cloud compute agreements. The company also reported $18.4B of Q2 capex (=$15.8B for AI compute) and strengthened liquidity with $100B cash after an IPO and a $25B investment-grade bond; management guided toward $100B ARR run-rate by Dec. 2026 backed by a $47.5B backlog and expects compute capacity to exceed 2GW by year-end.

Analysis

This is less a simple beat-and-raise and more a capital-allocation signal that changes the competitive map. The balance sheet now lets the company fund launch, satellite, and compute expansion simultaneously, which should force suppliers and rivals to think in multi-year rather than quarterly budgets. NVDA is the cleanest listed beneficiary because management is implicitly saying GPU supply is the binding constraint, not demand; that usually supports allocation priority and pricing power.

The second-order loser is legacy telecom, especially T, because direct-to-cell plus rooftop micro-sites attacks the economics of coverage density and disaster resilience. That does not destroy the business overnight, but it can force incumbents to spend defensively to protect premium customers, compressing returns on incremental capex. AAL is a smaller but real winner: connectivity becomes a product differentiator, which can lift premium mix and corporate share without needing broad industry volume growth.

The near-term catalyst path is execution, not narrative. In the next 1-2 quarters, the market will test whether launch cadence, permitting, and power delivery can keep up with the claimed compute ramp; any slip there would expose how much of the story is still optionality. The key falsifiers are a Starship delay, a sequential deceleration in subscriber adds, or AI revenue growth that fails to convert into margin because memory and power costs outrun pricing.

Contrarian view: consensus may be underestimating the value of bundling launch, connectivity, and compute into one distribution stack, which lowers customer acquisition costs and raises switching costs. But it may also be overpaying for the revenue targets themselves; if third-party cloud demand normalizes, the multiple can compress even while growth stays strong. This is a platform story, but not one that should be valued as if every adjacent market is automatically won.

Key Decisions for Investors

• Long NVDA / short T into the next 1-3 months as a supply-vs-substitution pair: NVDA benefits from a large, visible buyer with expanding AI capex, while T faces a defensive capex burden if mobile coverage economics are pressured. Cover the short if T shows accelerating enterprise wireless ARPU or if NVDA supply comments turn restrictive.
• Tactical long AAL on pullbacks for 4-8 weeks: the edge is product differentiation and potentially higher premium/corporate share, not a structural earnings reset. Falsify the trade if channel checks do not show higher load factors or if connectivity is not translating into yield lift by next quarter.
• Avoid a public short on SATS until spectrum deal economics are fully priced; current setup looks more like a one-time monetization event than a broken asset. If the stock rallies to imply recurring value from the spectrum sale rather than balance-sheet repair, fade it.
• Treat any weakness in SPY-style tech proxies tied to AI infra as a buying opportunity only if capex cadence and power-permit data remain on track; otherwise reduce exposure quickly. The story is highly dependent on physical execution over the next 2 quarters.