The Case For Data Centers In Space

Y Combinator · 2026-08-05 · 36 分鐘
https://www.youtube.com/watch?v=A9JDkiYEhfY影片總結
Starcloud puts an Nvidia H100 in orbit, targeting 88,000 satellites and 20 GW of compute.
Starcloud CEO Philip Johnston explains why his company is building AI data centers in orbit: terrestrial power and construction constraints are worsening while launch costs could fall dramatically with SpaceX Starship. Founded on January 1, 2024, the team booked a SpaceX rideshare launch the next day for about $300,000—before knowing what they would fly. Starcloud-1 launched in November 2025 with five GPUs, including an Nvidia H100, using an improvised ice-bath test and phase-change immersion cooling. After first contact in 12 hours, the satellite survived repeated software restarts and trained models in space. Starcloud estimates break-even launch costs near $500 per kilogram, plans a 10 kW Starcloud-2, and has filed with the FCC for 88,000 larger satellites representing roughly 20 GW of compute. Its hardest engineering problems remain removing heat in vacuum and protecting chips from radiation.
章節
- 0:00First thing every space company should do: book a SpaceX rideshare before building
- 1:35The Starbase trip that started it: Starship capacity reshaped the idea
- 3:20Why space-based solar doesn't work: 95% transmission loss versus $500-per-kilo data centers
- 7:15First contact and the satellite that kept restarting: 12-hour contact and three days of debugging
- 9:42The two hardest problems: heat in vacuum and radiation-resistant chips
- 11:27Testing GPUs in particle accelerators: five-year radiation doses in 24 hours
- 12:49Automotive parts instead of space-grade: cheaper components selected through radiation testing
- 14:30What 100 VCs got wrong: low-cost launch and terrestrial data-center constraints
- 18:38The path to 20 gigawatts in space: 88,000 Starcloud-3 satellites
- 21:16Designing a GPU for space with NVIDIA: Rubin Space and stripped-down H100s
- 24:00Why hard tech suddenly got popular with VCs: software moat concerns changed funding
- 25:37Inside the Benchmark round: Chethan backed Starcloud’s technical team
- 26:57Their hiring strategy: six months to hire Starcloud’s first engineer
- 28:08How to be contrarian but right: building around 10x-cheaper launches
- 31:44What if data centers get banned on Earth?: New York blocks construction
- 32:45Data center myth busting: water use versus power and national security
- 35:09Advice for hard tech founders: solve for technical talent first
這是 Tier 1 公開摘要
每章重點、段落總結、心智圖由分享者控制是否公開。想看完整分析?自己提交一支。
同頻道的其他分析
Jeff Dean: The 1% Rule for Building in AIY CombinatorJeff Dean links Google Search, TPUs, Gemini agents, and founder strategy to 1,000x energy gaps and 0%-success niches.
Patrick Collison: Is AI Breaking the Lean Startup Playbook?Y CombinatorPatrick Collison says Stripe took two years to launch, while Stripe data shows AI-era startups nearly doubled year over year.
Waymo Co-CEO Dmitri Dolgov: The Demo Is Only 1% Of The WorkY CombinatorWaymo’s demo took 18 months, its product 15 years, now reaching 17× human safety across 220 million autonomous miles.
Garry Tan: Own Your IntelligenceY CombinatorGarry Tan argues personal AGI, G Brain, and Markdown skills let one founder achieve 400x leverage while owning their intelligence.
Max Hodak: Average Is Not Good EnoughY CombinatorMax Hodak explains how Science’s Helix infrastructure, hiring system, and retinal implant work turn speed into startup advantage.
相關主題的分析
EP320. SPCX 跌破發行價、近期 AI 熱門話題、Intel 18A 進步了 | M觀點M觀點SpaceX跌破135美元、Grok Build洩程式碼,Intel 18A良率升至85%
EP323. Opus 5 正式發表、星艦第十三飛達標、開放模型的公開信 | M觀點M觀點Claude Opus 5成本降26%、Starship Flight 13完成九成任務,輝達等50家力挺開放權重模型
Elon Musk & Gwynne Shotwell on AI Risks and Peer Review, Starship, Terafab, SpaceX/Tesla MergerAll-In PodcastElon Musk forecasts Starship reusability in 2027 and urges AI peer review after Hugging Face security incidents.
EP311. Anthropic 作繭自縛、SPCX 風光上市、AI 時代電商怎麼做 | M觀點M觀點Anthropic 因 Claude 5 越獄遭美國禁用,SpaceX 上市市值衝破 2.1 兆美元
EP312. SPCX 併購 Cursor、Fable 5 還在牢裡、AI 產業的新利空 | M觀點M觀點SpaceX以600億美元併購Cursor,Anthropic受禁令困擾,AI從Token Maxing轉向預算最佳化