Space Exploration Technologies Corp. (SPCX)

News Summary

• SpaceX's SI (AI compute) segment hit $7.6B in Q3 revenue, up $5B sequentially and ~10x year-over-year, driven primarily by renting data-center capacity to customers including Anthropic, Google, and Reflection AI; management disclosed payback periods under one year at pricing slightly above the $30–$50/watt range, with Musk suggesting compression to six months in 2027 as utilization rises—a direct read on AI infrastructure pricing power and demand intensity relevant to any compute hardware vendor.
• SpaceX is deploying 200,000 GB300 GPUs per month and expects 2.5–3 GW of nameplate compute by end of 2026 scaling to at least 13 GW by end of 2027; critically, Musk stated SpaceX tried to offload inference workloads to hyperscalers and found none could supply GB300s, GB200s, TPUs, or H100s at scale, underscoring acute AI hardware scarcity that benefits alternative chip suppliers like Cerebras.
• Terafab, the SpaceX-Tesla-Intel chip manufacturing JV, is targeting a first SRAM test chip by May–June 2027 and volume production by Q3 2028 with an ambition of producing one terawatt of AI chips annually; separately, SpaceX's agentic Grok Bot reached 800,000 paid weekly users in October, more than doubling month-over-month, with enterprise customers over $1M run-rate growing 50% QoQ—signaling rapid agentic AI adoption that is pulling forward inference compute demand.
• Q3 revenue of $14.6B beat consensus by ~$1.9B (+161% YoY); adjusted EPS of $0.45 beat the $0.13 estimate; adjusted EBITDA of $8.8B (+306% YoY) beat consensus by ~$2.2B; gross margin expanded to 62.4% from ~44–51% in prior quarters; SpaceX guided Q1 2027 revenue of $33–37B and Q4 capex of $35–40B, reaffirming a $100B annualized revenue run rate by year-end 2026 and $500B–$1T ARR by end of 2027.
• SPCX shares rose ~8% in extended trading; analyst reaction was broadly bullish with JPMorgan raising its price target to $500, Barclays to $300, Morgan Stanley maintaining $300, and Goldman citing compute demand 'substantially ahead of supply'; the dominant press framing was that SpaceX has transformed into a de facto cloud and AI infrastructure provider, with Reuters noting its business 'looks closer to a cloud provider's than an AI lab's.'

Financial Highlights

Earnings Call Highlights

SpaceX's Q3 2026 call was dominated by an almost evangelical tone around AI compute infrastructure, with Elon Musk positioning SpaceX's Super Intelligence (SI) segment as the defining growth engine of the company and projecting $500B–$1T in annual recurring revenue by end of 2027. Management was unusually specific and confident about compute deployment metrics, framing SpaceX's vertical integration and engineering culture as a structural moat against hyperscalers who cannot source power, cooling, or even chips at comparable speed. Analyst sentiment was broadly enthusiastic, with questions probing the balance between internal and external compute monetization rather than challenging the underlying thesis. The dominant narrative was that AI infrastructure scarcity—not chip design—is the binding constraint, and SpaceX believes it has cracked that constraint in ways competitors cannot replicate quickly. The SI segment's $7.6B quarterly revenue, up $5B sequentially, gave the bullish framing hard financial grounding.

• SpaceX is deploying 200,000 GB300 GPUs per month and expects to reach 2.5–3 gigawatts of nameplate compute by end of Q4 2026, with high confidence in 13 gigawatts total by end of 2027 and a non-zero probability of exceeding 20 gigawatts—figures that would make SpaceX one of the largest AI compute operators on earth.
• SI segment revenue hit $7.6B in Q3, up $5B quarter-over-quarter, with cloud services agreements driving the ramp; Bret Johnsen disclosed that new compute deployments are generating payback periods of under one year at pricing 'slightly above the high end' of the $30–$50 per watt range, and Musk suggested payback could compress to six months in 2027.
• Musk confirmed SpaceX is targeting a functional test chip (SRAM) from its Austin research fab by May–June 2026 and volume production at Terafab (Gibbons Creek) by Q3–Q4 2028, with the stated ambition of producing a terawatt of AI chips per year—directly threatening incumbent chip suppliers and foundries.
• Grok bot, SpaceX's agentic AI assistant launched in August, already has 800,000 paid weekly users and is more than doubling month-over-month; Grok Enterprise customers with over $1M in run-rate revenue grew 50% quarter-over-quarter pro forma, signaling rapid enterprise agentic adoption that would compete directly with other AI inference and agent platform providers.
• Musk argued that AI agents will consume orders of magnitude more bandwidth per unit than humans, projecting 1,000x demand growth for connectivity in 3–5 years, and framed compute scarcity as so acute that SpaceX tried to offload inference workloads to hyperscalers and found none could supply GB300s, GB200s, TPUs, or even H100s at scale—a pointed signal about the tightness of the AI hardware market that Cerebras competes in.

Sources

• Musk says SpaceX-Tesla merger is 'a good idea' as companies form deal ...
• SpaceX stock rises after Q3 earnings beat. What analysts say - CNBC