Where Will SpaceX Be in 3 Years?

The article highlights SpaceX’s shift toward AI infrastructure: a $60B Anysphere acquisition and neocloud monetization with Alphabet’s 3-year deal (~$30B by 2029) and Anthropic paying ~$15B annually for Colossus 1 capacity. Starlink remains the profit engine with ~12M subscribers, $11.4B sales in 2025, and $4.4B operating income, targeting 15M U.S. subscribers by 2030, plus a planned 1,200-satellite Starship launch mid-2027. For the rocket segment, Goldman estimates core revenue could rise to $8.3B by 2030 (from $4.1B in 2025), with Starship expected to cut marginal launch costs by 90% versus Falcon rockets; overall, the outlook is positive but flagged as execution-dependent.

Analysis

The market implication is less about the private company itself and more about who gets paid along the stack if a new, capital-intensive AI tenant emerges. The cleanest second-order winner in the public tape is NVDA, because any credible neocloud scale-up is a demand amplifier for accelerators, interconnect, and power delivery; GOOGL also benefits if outsourced capacity lets it smooth AI demand without fully owning the capex burden. The subtle loser is the hyperscaler model itself: if capital-light capacity sharing proves economical, it pressures every cloud operator to defend utilization and can cap valuation multiples tied to “efficient growth.”

The bigger risk is that headline contract value and long-dated TAM talk can outrun actual free cash flow. In the next 1-3 months, the market will likely trade on incremental signs of real delivery: rack deployment, customer renewal behavior, and whether margins survive the buildout; over 6-18 months, the decisive variable is whether launch cadence and power availability scale without forcing repeated equity raises or debt dilution. For the satellite franchise, subscriber growth alone is not enough; if terminal subsidies or launch costs rise faster than ARPU, the operating leverage story can stall despite strong top-line optics.

I would treat TMUS acquisition chatter as low-conviction optionality rather than a tradable catalyst; telecom M&A tends to need regulatory and financing windows that are measured in quarters, not weeks. GS has a modest positive skew if private-market financing and strategic deal activity accelerate, but that is a fee-pool tailwind, not a core earnings driver. The contrarian miss is that the scarcest resource may be power and permitting, not demand; that favors grid, transmission, and data-center infrastructure more than the story-stock names attached to the narrative.

Key Decisions for Investors

• Buy NVDA on 5-8% pullbacks over the next 1-3 months as the cleanest public proxy for AI capacity buildout; thesis fails if hyperscaler and neocloud capex guides roll over for two consecutive quarters.
• Stay long GOOGL only as a quality AI-infrastructure exposure, but size modestly; the upside is in optionality from outsourced compute, while the risk is margin dilution if capacity demand proves more cyclical than advertised.
• Do not chase TMUS on acquisition speculation; wait for a binding strategic bid or FCC/process milestone before taking risk. If the stock rallies purely on rumor, that is a fade candidate rather than a long.
• Use GS as a secondary beneficiary of private-market funding and M&A, but keep it as a watchlist name unless deal volumes actually accelerate; the catalyst window is 6-12 months, not immediate.
• Alert: if future disclosures show SpaceX-style capacity deals converting into sustained FCF and not just booked revenue, rotate toward power/grid and infrastructure proxies; if not, treat the narrative as overextended and trim AI infrastructure exposure.