Sep 15, 2026 · 1h 4m
SpaceX pairs launch ambitions with an AI safety reckoning
Elon Musk & Gwynne Shotwell on AI Risks and Peer Review, Starship, Terafab, SpaceX/Tesla Merger
The conversation connects SpaceX’s expansion into connectivity, orbital computing, chips, and reusable rockets with urgent questions about AI security and oversight.
- 1Gwynne Shotwell traces SpaceX’s shift from launch startup to a sprawling infrastructure and AI platform.
- 2Elon Musk argues that rival AI companies should cross-test models for deception, cyber risk, and catastrophic failure.
- 3Starship, orbital compute, and TeraFab reveal the capital-intensive strategy behind SpaceX’s next phase.
Don't miss
Elon Musk argues for cross-company AI test harnesses after discussing model deception, security incidents, and the difficulty of trusting systems to evaluate themselves.
The brief
Gwynne Shotwell recounts joining SpaceX as an early business-development leader and growing into the executive overseeing launch, connectivity, satellites, and AI infrastructure.
The conversation presents Starlink, Starship, orbital supercomputing, and TeraFab as connected bets: revenue and infrastructure from one business help finance the next.
Elon Musk shifts the focus to AI security, arguing that companies should cross-test one another’s models because self-evaluation may miss deception, cyberattacks, and catastrophic failure.
Musk says Starship Flight 15 could attempt the first tower catch, while TeraFab would build semiconductor capacity through a staged Austin research and production effort.
The episode’s central tension is expansion versus control: SpaceX is scaling systems across launch, communications, compute, and chips as AI risks accelerate faster than oversight.
What was said on this episode
44 statements · 32 positive · 10 negative · 2 neutral
Gwynne Shotwell predicts SpaceX will land people on Mars within the decade.
“I think we'll put people down within the decade.”
Listen at 0:07
Gwynne Shotwell expects a lunar settlement and would view its absence negatively.
“I would be disappointed if we didn't have a settlement on the moon.”
Listen at 0:23
Gwynne Shotwell says organizations not using or leading AI risk becoming irrelevant.
“If you're not using AI, if you're not leading AI, you run the risk, significant risk of being irrelevant.”
Listen at 7:49
Gwynne Shotwell says future rockets will increasingly be designed using AI systems.
“100%.”
Listen at 8:15
Gwynne Shotwell says SpaceX does not plan to issue additional stock.
“I don't think we're going to release any more stock.”
Listen at 11:49
Gwynne Shotwell assesses Starlink market penetration at 1.5%–2%.
“Our market penetration on Starlink is 1.5% to 2%, depending on which country you're in.”
Listen at 12:23
Gwynne Shotwell assesses computer rental as a highly attractive business currently.
“Honestly, right now, and it is a little embarrassing, but computer rental, it's a heck of a business.”
Listen at 12:40
Gwynne Shotwell says demand for compute rental is not declining.
“We see no drop in demand at all.”
Listen at 12:51
Gwynne Shotwell says Starlink spectrum will be used to eliminate cellular dead zones.
“we want to leverage that to make sure that there are no dead zones.”
Listen at 14:45
Gwynne Shotwell says achieving Elon Musk's goals requires Starship.
“To achieve Elon's goals, you really need a Starship.”
Listen at 16:11
Gwynne Shotwell predicts Starlink's market will become highly competitive.
“I think it'll be very competitive”
Listen at 17:24
Gwynne Shotwell says superior capability and customer service determine winners in Starlink competition.
“If your capability is incredible, people will buy it.”
Listen at 17:41
Gwynne Shotwell claims Elon Musk's companies generally deliver promised products and services.
“we end up doing what we say we're going to do.”
Listen at 18:24
Gwynne Shotwell says demand for compute is currently insatiable.
“The demand is insatiable right now.”
Listen at 19:48
Gwynne Shotwell says space offers effectively unlimited real estate for data centers.
“The real estate in space is infinite.”
Listen at 20:02
Gwynne Shotwell predicts SpaceX will launch AI-compute satellites next year.
“We're also going to launch AI compute. Satellites. So next year is a big year.”
Listen at 21:20
Gwynne Shotwell says effective managers can substantially leverage employee talent.
“great managers can really leverage the talent.”
Listen at 22:56
Gwynne Shotwell claims SpaceX hires the best available people.
“We hire the best people.”
Listen at 24:04
Elon Musk assesses artificial intelligence as potentially very dangerous.
“AI can be very dangerous.”
Listen at 29:06
Elon Musk predicts sufficiently capable AI models will seek to escape constraints.
“any sufficiently smart model seems like it will want to escape its constraints.”
Listen at 29:45
Elon Musk recommends major AI competitors test one another's models before release.
“what I think would be wise to do as soon as possible, if not immediately, would be to have the major AI competitors test each other's models”
Listen at 29:55
Elon Musk says test-harness logs would expose model distillation or intellectual-property theft.
“if you try to do distillation or steal IP, it would be very obvious based on the logs.”
Listen at 31:29
Elon Musk assesses cross-company model testing as a reasonable request.
“I think this is, I think this is a pretty reasonable request that, that models just get tested.”
Listen at 33:21
Elon Musk assesses current AI danger as very significant.
“the danger of AI is very significant at this point.”
Listen at 34:35
Elon Musk says risks from AI models are increasing exponentially.
“we have at this point exponentially increasing risk with the AI models.”
Listen at 34:47
Elon Musk says an AI-enabled air-gap breach is plausible.
“Okay. That's not out of the question.”
Listen at 36:46
Elon Musk says Falcon rockets have recently delivered payloads without exploding.
“the Falcon rockets deliver their payload to orbit and haven't exploded for a long time”
Listen at 39:03
Gwynne Shotwell says earlier reporting of rocketry problems makes them easier to solve.
“the sooner you bring it up, the easier it is going to be to solve that problem.”
Listen at 40:42
Elon Musk says SpaceX must achieve functioning rockets, satellites, and Starlink connectivity.
“the rockets need to get to orbit, the satellites need to work, uh, the Starlink, you know, connection needs to work”
Listen at 41:22
Elon Musk predicts SpaceX will attempt to catch Starship on Flight 15.
“on Flight 15, we will try to catch the ship”
Listen at 42:55
Elon Musk says reflying Starship would establish the first fully reusable orbital rocket.
“once we can refly the ship, we will have made the first fully reusable orbital rocket.”
Listen at 43:14
Elon Musk says Starship is designed for full, aircraft-like rapid reusability.
“it's designed for not just full reusability, but also rapid reusability like an aircraft.”
Listen at 44:15
Elon Musk says rapidly reusable Starship is critical for extending life beyond Earth.
“this is a very important breakthrough. It's really the critical breakthrough that's necessary to extend life beyond Earth.”
Listen at 44:23
Elon Musk estimates at least 50–60% odds of catching Starship on its first attempt.
“I'd say it's at least 50 or 60%.”
Listen at 44:39
Elon Musk predicts Starship will achieve full rapid reusability in 2027.
“I think it's extremely likely that we will achieve full reusability with rapid reflight next year in 2027.”
Listen at 45:56
Elon Musk says TerraFab is needed to ensure future chip supply certainty.
“we need to make sure, we need, we need, there needs to be Uh, certainty of future supply of chips”
Listen at 47:52
Elon Musk says AI scaling requires building TerraFab or accepting failure to scale.
“So it's either build TerraFab or, or fail to scale.”
Listen at 48:21
Elon Musk predicts the chip project will reach initial fabrication capability by end of next year.
“I think we'll get at least to the crawl part by the end of next year.”
Listen at 50:32
Elon Musk says reciprocal test-harness evaluation is the best available AI-safety measure.
“everyone applying everyone else's test harness to each other is probably the best way we— best thing we do to ensure, ensure safety.”
Listen at 54:12
Elon Musk predicts AI companies will find it difficult to reject reciprocal model testing.
“I think it would be the sort of thing that's kind of hard to say no to.”
Listen at 54:43
Elon Musk predicts enormous legal liability if companies release models flagged as unsafe.
“the legal liability would be enormous.”
Listen at 56:28
Elon Musk assesses Anthropic as more safety-focused than OpenAI.
“on balance, I, I think Anthropic is, is, you know, puts more, more care into their safety than, than OpenAI.”
Listen at 59:32
Elon Musk says heterogeneous cross-company testing greatly increases the odds of finding AI issues.
“The odds that you will find issues are dramatically greater.”
Listen at 1:00:31
Elon Musk says regulation tends to increase irreversibly once imposed.
“it does tend to be very much a one-way ratchet in terms of increasing regulations.”
Listen at 1:02:14
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.