Latest / Elon Musk Podcast / Musk Worried. xAI Just Lost Another Co-Founder

Transcript

Musk Worried. xAI Just Lost Another Co-Founder · Elon Musk Podcast

• 0:00
  Tony Wu resigned from XAI, making him the 5th Co founder to
• 0:04
  leave Elon Musk's artificial intelligence company since it
• 0:08
  launched in 2023. Wu helped build Alpha geometry
• 0:11
  in Minerva and the memorizing Transformer during his time at
• 0:15
  Google Deepmine. Before joining Elon Musk's
• 0:18
  startup, he had a PhD in machine learning from the University of
• 0:22
  Toronto and focused his entire career on teaching machines to
• 0:26
  reason like mathematicians. Now his departure comes just
• 0:30
  eight days after SpaceX absorbed X AI and $1.25 trillion all
• 0:36
  stock merger, the largest private company deal in history.
• 0:41
  So what's happening inside X AI that keeps pushing founding
• 0:45
  talent towards the exits? Wu posted his resignation on X,
• 0:50
  thanking Musk for the ride of a lifetime and writing that he
• 0:54
  would deeply miss the people, the war rooms and all those
• 0:58
  battles we have fought together. He hinted at launching something
• 1:02
  new, saying a small team armed with a IS can move mountains and
• 1:06
  redefine what's possible. We're going to break down every
• 1:09
  Co founder who is left. Axe AI, What each departure
• 1:13
  tells us about the company's internal state, how the SpaceX
• 1:16
  merger changes the calculus for remaining founders, and why XAI
• 1:20
  is brain drain could create an opening for competitors right
• 1:24
  when Musk needs stability most. And we'll get right into that
• 1:28
  after this very short break. Now Tony Wu is gone from XAI,
• 1:34
  and he took with him years of institutional knowledge about
• 1:37
  mathematical reasoning systems the no new hire can easily
• 1:41
  replace. Wu's research portfolio alone
• 1:45
  included core contributions to star, Alpha star and auto
• 1:49
  formalization work that sat at the center of XA is effort to
• 1:53
  build models that think through problems step by step rather
• 1:56
  than pattern match their way to answers.
• 1:59
  Now, the departure timeline at XAI tells a very telling story
• 2:04
  of accelerating attrition. Kyle Kazik left first in 2024,
• 2:10
  walking straight to Open AI, Musk's most better rival.
• 2:14
  Of course, we know about that. Christian Zagetti followed later
• 2:21
  that year, joining Morph Labs. Igor departed in August 2025 to
• 2:26
  launch a new venture, its venture capital firm focused on
• 2:30
  safe AI development. Greg Yang stepped back in
• 2:33
  January 2026 after a Lyme disease diagnosis.
• 2:37
  Moving into an informal advisory role and Wu's exit in February
• 2:41
  brings the total to five. Departure from XA is original.
• 2:44
  12 Co founders in under three years. 7 remain and Musk himself
• 2:49
  is one of them now. Each departure strips away a
• 2:52
  different layer of technical expertise.
• 2:55
  Kozak carry deep product knowledge to a direct
• 2:57
  competitor. These other ones co-authored
• 3:01
  papers with Wu on theorem proving.
• 3:05
  They shifted entirely out of building AI into funding it.
• 3:10
  Yang forced him out of daily operations and Wu's speciality
• 3:14
  building machines. They can solve Olympian level
• 3:16
  geometry problems and drive mathematical proofs.
• 3:20
  Filled a niche, then XAI cannot simply backfill with a job
• 3:23
  posting on LinkedIn. The company launched with 12 Co
• 3:26
  founders because it needed that density and that chemistry of
• 3:30
  talent from Google, DeepMind, Microsoft and Open AI to compete
• 3:34
  with everybody else. Losing 5 of them in rapid
• 3:37
  succession erodes the original vision those founders carried
• 3:40
  into that building. Now, I've been digging through
• 3:43
  the analytics of the show and about 39% of you are following
• 3:46
  the show, 61% of you have listened to an episode or have,
• 3:52
  you know, listened to two or three or 4, but just haven't hit
• 3:55
  the follower subscribe button. Been doing this for about 7
• 3:58
  years, eight years, going to do it for 10 more with your help.
• 4:02
  So please hit the subscribe button.
• 4:03
  Thank you so much. Now SpaceX, my favorite company.
• 4:08
  The merger adds a new variable to every remaining Co founders
• 4:12
  calculation. SpaceX acquired xai on February
• 4:15
  2nd in a deal structured as a share exchange, converting 1 XAI
• 4:20
  share into .1433 SpaceX shares. A SpaceX was valued at a
• 4:26
  trillion dollars in XAI. Add 250 billion in the
• 4:30
  transaction documents. The combined entity is preparing
• 4:33
  for an IPO that could raise 50 billion in value of the company
• 4:37
  at 1.5 trillion. SpaceX issued 250 billion in new
• 4:42
  shares to XAI shareholders, which diluted existing SpaceX
• 4:46
  investors significantly, The New York Times reported.
• 4:50
  The SpaceX long time backers were forced to shrink their
• 4:53
  ownership in the company drastically as a percentage to
• 4:57
  pay for the acquisition. For Co founders who stayed
• 5:00
  through XA is crash and burn phase, reportedly around a
• 5:04
  billion dollars per month. The merger created a direct path
• 5:08
  to liquidity through SpaceX stock ahead of one of the
• 5:11
  largest IP OS in history. Now Wu's timing suggests he
• 5:14
  waited for that conversion event before moving on.
• 5:17
  His shares are now SpaceX shares.
• 5:19
  I think he doesn't leave. He can exit with his equity
• 5:22
  intact while pursuing whatever comes next because he's going to
• 5:25
  have that money now. The merger itself was driven by
• 5:28
  Musk's fixation on orbital data centers.
• 5:31
  The idea that solar powered computing infrastructure in
• 5:34
  space can solve a is insatiable demand for electricity.
• 5:38
  Elon Musk wrote about these data centers require immense amounts
• 5:43
  of power and cooling, and the global electricity demand for AI
• 5:47
  simply cannot be met with terrestrial solutions.
• 5:50
  Now I'm not sure about this. I know there's regulations, but
• 5:56
  Elon Musk owns a power company. He owns Tesla solar panels.
• 6:02
  So do you think Elon Musk could build a bunch of solar panels in
• 6:04
  the desert? Probably.
• 6:07
  But you know, he would rather shoot these things into space
• 6:10
  and have AI data centers, maybe a million of them, floating
• 6:14
  around in low Earth orbit. And SpaceX did file with the FCC
• 6:19
  to launch up to 1,000,000 satellites as part of this
• 6:22
  orbital data center constellation. 1,000,000
• 6:26
  satellites in orbit. Musk estimated that space based
• 6:29
  AI compute could become the lowest cost option within two to
• 6:33
  three years. Jeff Bezos and Google's Project
• 6:36
  Sun Catcher are pursuing similar concepts, which gives the idea
• 6:40
  of some industry validation even if the engineering remains
• 6:44
  completely unproven. Now the pitch sounds crazy
• 6:49
  science fiction, but Elon Musk is putting real corporate
• 6:52
  structure behind it. XAI now operates as a fully
• 6:55
  owned subsidiary of SpaceX through a triangular merger
• 6:59
  designed to keep it legally insulated from the parent
• 7:02
  company. Now Reuters reported that this
• 7:04
  structure prevents XA is past liabilities from automatically
• 7:08
  becoming Spacex's new problems. That SpaceX and XAI legal
• 7:15
  separation is very valid because XAI carries some serious
• 7:20
  regulatory baggage into this deal.