Google, Amazon, and OpenAI Explore Space Data Centers as AI Power Use Rises

Major technology firms are exploring orbital data hubs to meet rapidly rising AI power demand, with Google planning two small Tensor-chip units by early 2027 (in partnership with Planet Labs) and Starcloud already sending a high-end Nvidia unit to space. Global data-center energy demand is projected to rise from ~415 TWh in 2024 to ~945 TWh by 2030, but economics remain a barrier—current launch costs near $1,500/kg must fall toward ~$200/kg by 2035 and technical issues (heat, upgrade space, orbit risk) persist—so expect early tests in coming years but only long-term, conditional deployment at scale.

Analysis

Market structure: Hyperscalers and AI chip vendors (NVDA, GOOGL, AMZN, CRM) are primary beneficiaries as they capture high-margin AI workloads and can internalize any niche space-linked infra; Planet Labs (PL) is a small-variance beneficiary tied to early tests. Incumbent data‑center REITs (DLR, EQIX) and utilities face margin pressure from higher cooling/energy capex as global data‑center demand doubles from ~415 TWh (2024) to ~945 TWh (2030), while launch‑cost economics ($1,500/kg today vs. ~ $200/kg target by 2035) cap full disruption for >1 decade.

Risk assessment: Tail risks include catastrophic launch failures, regulatory bans on offshore/space data for sovereignty (probability medium), and insurability spikes after radiation/repair losses; immediate risk (days–weeks) is trading/system outages (CME example), short term (1–36 months) is tech demos/failures, long term (>5–10 years) depends on launch‑cost curves hitting <$1,000/kg by 2028 and ~$200/kg by 2035. Hidden dependencies: heavy reliance on SpaceX commercialization, radiation‑hardened chip progress, and global energy prices which can swing TCO ±20–40%.

Trade implications: Near term (0–12 months) favor NVDA (core AI demand) and cloud champions GOOGL/AMZN — consider concentrated 2–3% long NVDA, 1–2% GOOGL/AMZN exposure; small option punts in PL (0.5%) for 12–36 month optionality tied to Starship milestones. Hedge/rotate: underweight data‑center REITs (DLR, EQIX) and select utilities by 2–4% as secular capex rerates; use NVDA 6–12 month 15–25% OTM call spreads funded by selling near‑term calls on AMZN/GOOGL to monetize elevated IV.

Contrarian angle: The market underestimates operational friction — latency, repairability, radiation losses and data‑sovereignty barriers make space hubs a niche, not a near‑term replacement; enthusiasm may be priced into growth multiples of chip/cloud names but actual adoption is contingent on hitting <$1,000/kg by 2028. Historical parallel: early cloud infrastructure investment produced winner‑take‑most consolidation; if that repeats, concentrate on durable moats (NVDA, GOOGL, AMZN) rather than speculative space infra names until cost curves materialize.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in NVDA within the next 30 days to capture near‑term AI demand; pair with a 6–12 month 15–25% OTM call spread (size = 0.5–1% portfolio) to limit premium spend and capture upside into next two earnings cycles.
• Add 1–2% long positions in GOOGL and AMZN (0.5–1% each) over 1–3 months to play cloud-driven workload absorption; fund by reducing 2–4% exposure to data‑center REITs (DLR, EQIX) and selected utilities whose TCO will be stressed as AI energy use rises.
• Initiate a 0.5–1% speculative long in PL as a binary, 12–36 month catalyst bet tied to Starship/launch cost progress; set hard stop if market‑wide launch costs are >$1,000/kg in 2028 or if Starship program misses major milestones in 12 months.
• Hedge operational/exchange tail risk: buy a 3–6 month ATM put on CME sized to 0.5–1% of portfolio or reduce CME exposure by 1–2% after evaluating post‑outage remediation; reason: single‑event operational risk can cause outsized P&L hits in minutes.
• Sector tilt: immediately overweight Cloud/AI infra (GOOGL, AMZN, CRM, NVDA) by 3–5% and underweight Utilities/Data‑Center REITs by 3–5% for a 12–24 month horizon; revisit if launch costs decline to <$1,000/kg by end‑2028 or if NVDA revenue guidance misses by >5%.