Elon Musk – "In 36 months, the cheapest place to put AI will be space”
In this episode, John and I got to do a real deep-dive with Elon. We discuss the economics of orbital data centers, the difficulties of scaling power on Earth, what it would take to manufacture humanoids at high-volume in America, xAI’s business and alignment plans, DOGE, and much more. 𝐄𝐏𝐈𝐒𝐎𝐃𝐄 𝐋𝐈𝐍𝐊𝐒 * Transcript: https://www.dwarkesh.com/p/elon-musk * Apple Podcasts: https://podcasts.apple.com/us/podcast/dwarkesh-podcast/id1516093381?i=1000748400389 * Spotify: https://open.spotify.com/episode/4nah0x1qQF2hxgJnv8PlmN?si=88be1f9b1a8b47b4 𝐒𝐏𝐎𝐍𝐒𝐎𝐑𝐒 - Mercury just started offering personal banking! I’m already banking with Mercury for business purposes, so getting to bank with them for my personal life makes everything so much simpler. Apply now at https://mercury.com/personal-banking - Jane Street sent me a new puzzle last week: they trained a neural net, shuffled all 96 layers, and asked me to put them back in order. I tried but… I didn’t quite nail it. If you’re curious, or if you think you can do better, you should take a stab at https://janestreet.com/dwarkesh - Labelbox can get you robotics and RL data at scale. Labelbox starts by helping you define your ideal data distribution, and then their massive Alignerr network collects frontier-grade data that you can use to train your models. Learn more at https://labelbox.com/dwarkesh To sponsor a future episode, visit https://dwarkesh.com/advertise. 𝐅𝐔𝐑𝐓𝐇𝐄𝐑 𝐑𝐄𝐀𝐃𝐈𝐍𝐆 Notes on Space GPUs — turned my Elon prep into a blog post: https://www.dwarkesh.com/p/notes-on-space-gpus 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 0:00:00 - Orbital data centers 0:36:46 - Grok and alignment 0:59:56 - xAI’s business plan 1:17:21 - Optimus and humanoid manufacturing 1:30:22 - Does China win by default? 1:44:16 - Lessons from running SpaceX 2:20:08 - DOGE 2:38:28 - TeraFab
Read Video · 文字稿与深度分析
本集已有完整文字稿 + AI 深度分析
免费注册 · 无需信用卡 · 注册即获 150 积分,足够解锁本集
- 📄 完整文字稿含时间戳
- ✨ AI 摘要、关键词与思维导图
- 💡 核心要点与精彩引言
节目时间轴
Elon Musk argues that space will become the cheapest place to run AI within 36 months because terrestrial energy scaling is hitting hard limits.
- Musk predicts that in 36 months or less—possibly 30—the most economically compelling place to put AI will be space, because solar panels there produce about five times more power than on the ground and require no batteries.
- Outside of China, electrical output is essentially flat while chip output grows exponentially, so the only scalable answer is space-based solar power, since covering Nevada in solar panels would require impossible permitting.
- Space is fundamentally a regulatory play: it is harder to scale power on the ground than in space, and the atmosphere alone causes roughly a 30% energy loss for terrestrial solar.
The terrestrial power bottleneck is not just generation but the entire supply chain, especially gas turbine blades and vanes.
- Building a gigawatt of power for xAI's Colossus 2 required ganging together turbines, dealing with Tennessee permit issues, and running high-power lines across the border into Mississippi.
- The limiting factor for gas turbines is the cast vanes and blades, made by only three casting companies worldwide that are massively backlogged, with turbines sold out through 2030.
- Musk estimates that servicing 330,000 GB300s—including networking, CPU, storage, peak cooling, and servicing margin—requires roughly a gigawatt at the generation level, far more than naive chip-power calculations suggest.
SpaceX and Tesla plan to scale solar cell production to 100 gigawatts per year to enable orbital data centers.
- Both SpaceX and Tesla have a mandate to reach 100 gigawatts a year of solar cell production, doing the whole stack from raw materials to finished cells.
- Solar cells are already farcically cheap at around $0.25-0.30 per watt in China, and in space they become roughly ten times cheaper because no batteries are needed.
- Space-based solar cells are cheaper to make than terrestrial ones because they need no heavy glass or framing to survive weather events.
关键概念
- space data centers— Musk's core prediction that the cheapest place to run AI will be in orbit within 36 months.
- energy bottleneck— The central argument that flat electricity output outside China is the limiting factor for AI scaling.
- solar power in space— Space solar panels deliver roughly five times the power of ground-based panels with no batteries needed.
精选金句
In 36 months, but probably closer to 30 months, the most economically compelling place to put AI will be space.
🔥— A bold, specific prediction that overturns the assumption that data centers must stay on Earth.
Any given solar panel can do about five times more power in space than on the ground.
🤯— Reveals a surprising physical advantage of space that most people never consider.
可执行的洞察
🧠Strategic Thinking
Focus on the limiting factor, not everything at once.
This week, list your top three bottlenecks and delegate or drop everything else.
Space is a regulatory and energy arbitrage play.
Identify one regulatory or physical constraint in your industry and brainstorm a workaround.
👥Management and Hiring
Believe your interaction over the resume.
In your next interview, spend 20 minutes on a technical problem and ignore the CV.
Skip-level meetings prevent filtered information.
Schedule a skip-level meeting with a junior engineer this week and ask what's really blocking them.
转录文字与 AI 洞察均由模型自动生成,可能存在少量误差。识别效果与音频质量、语速和发音清晰度相关——如有内容看起来不对,以原始音频为准。
播客与视频,已可阅读
音频播客

Essentials: Control Your Brain Chemistry for Focus, Motivation & Well-Being
Huberman Lab
2026年8月6日34:37EN
BA Bites - The Top 10 Ways Business Analysts Influence Strategy (Without the Title)
The Better Business Analyst Podcast
2025年6月11日17:38EN
Cook Samin Nosrat (‘Salt Fat Acid Heat’) Returns with ‘Good Things’
Talk Easy with Sam Fragoso
2025年9月7日1:12:44EN
vol.37 “我只是做自己 却冒犯到了一些人”该怎么办
天真不天真
2025年9月30日1:07:03ZH-Hans
#6: Offseason vs. Preseason mit Niklas Jauch
We talking about practice
2021年3月21日1:00:10DE
163.- “Mujeres, dinero y el miedo a incomodar” con Maca Riva
LA MAGIA DEL CAOS con Aislinn Derbez
2026年5月26日1:27:54ES
视频

Uber Sells the Dream, Waymo Logs the Autonomous Miles
The Road to Autonomy
2026年2月28日53:54EN
Marketing Agents Masterclass (GROW your startup)
Greg Isenberg
2026年8月5日43:59EN
Sam Altman on Building OpenAI & Betting on the Impossible
David Senra
2026年8月23日1:18:17EN
从「上瘾模型」到「专注力训练」,如何在被算法理解的世界里重新找回主动?| 英文访谈 S9E33
声动活泼
2025年10月16日50:48ZH-Hans
Alles geschieht zur richtigen Zeit — vertraue und entspanne dich | Stoizismus
Stoische Stärke
2026年8月18日50:11DE