CNBC / David Faber (2025, secondary)

• Interviewer/venue: David Faber, on CNBC’s Closing Bell: Overtime, recorded at the Tesla Gigafactory in Austin, Texas.
• Format: Live televised sit-down — David Faber’s second interview with Musk the same day (May 20, 2025). The earlier segment is tracked separately at CNBC / David Faber (2025); this page covers only the post-intermission “twofer” and does not repeat what the first one already established.
• Date: May 20, 2025.
• Trust tier: verified (Tier 1). The raw is CNBC’s official published transcript; the source_url is the cnbc.com transcript page (HTML, so #:~:text= fragments work).
• Quote citation: every linked block quote below (a > "…" [🔗](…) line carrying a 🔗) is byte-accurate to the raw and spoken by Musk (David Faber is the interviewer — none of his words are quoted). Each is anchored to the CNBC transcript page with a #:~:text= fragment whose decoded snippet is a verbatim substring of the quote. Fragments are deliberately apostrophe-free (live CNBC pages render apostrophes as curly characters, so an ASCII apostrophe in a fragment would fail to highlight), with in-snippet hyphens percent-encoded (%2D). A few > callouts carry no 🔗 — the documentary note below and the provenance note further down — and are wiki-authored editorial notes, not quotes, per the convention used across the source pages.

Summary

The same-day sequel to the first Faber interview, and the more forward-looking of the two: where the first caught Musk on government, free speech and the robotaxi rollout, this one is mostly about the future he is betting on — humanoid robots, the AI compute frontier, and how AI development ends. For the wiki the highest-signal material is not the deal talk (Uber, an xAI–Tesla merge, his pay package) but three reasoning clusters.

On humanoid robots, he makes his strongest demand claim anywhere in the wiki — “humanoid robots will be the biggest product ever,” demand “insatiable” — and, more revealingly, lays out a model of how they learn: imitation via motion-capture, then the “threshold” breakthrough of learning from video, then self-play modeled explicitly on “how does a child learn.” The wiki opens a dedicated Humanoid robots page for this belief cluster.

On the AI frontier as a physics problem, he restates a long-standing prediction — the limiter on AI was chips, then becomes electrical equipment, then power generation itself — and folds in his admiration of China’s build-out. The compute-and-power reasoning is tracked on xAI and Grok.

On how AI ends, he softens (but does not retract) the “20% chance of annihilation” line into a precautionary stance, and frames the choice in movie terms: a Gene Roddenberry (Star Trek) outcome versus a James Cameron (Terminator) one — “we want the Roddenberry outcome.” That, plus “we have courtside seats to the big bang of intelligence explosion,” extends AI existential risk and Humanity’s bright future. A separate, sharp aside — that breakthrough innovation requires questioning authority, which he reads as America’s cultural edge over China — extends Curiosity and truth-seeking.

Documentary note: this page reports Musk’s stated predictions and reasoning and attributes them to him; it does not endorse or rebut them. The demand, timeline, compute and China claims are his forecasts/characterizations, recorded as such.

Key quotes (verbatim, CNBC-anchored — Elon Musk only)

On humanoid robots — demand and learning

The demand thesis, staked under pushback that the robots are “decades away”:

“I think, I think humanoid robots will be the biggest product ever. The demand will be insatiable.” 🔗

The universality intuition behind it:

“who wouldn’t want their own personal C3PO or R2D2. Everyone’s going to want one.” 🔗

The “threshold” capability he says unlocks everything — learning a task from watching video:

“if optimus can watch videos, YouTube videos, or how to videos, or whatever. And based on that video, just like a human can learn how to do that thing, then you really have task extensibility that is dramatic, because then it can learn anything very quickly.” 🔗

The learning model proper — a robot learning the way a child does:

“you want the robot to self-play. So you say, how does a child learn? Well, a child has toys and a child plays with the toys.” 🔗

“Once you have a lot of robots, you can do this self-play, which is that you just put the robot in a room with toys and have the robot, literally have the robot play with toys.” 🔗

On the AI frontier — chips, then power

His long-standing limiter prediction, restated — and pushed one link further down the chain:

“a few years ago, I made a very obvious prediction which is that the limitation on AI will be chips. And it’s still chips, kind of chips today, then it will be electrical equipment” 🔗

The scale claim for his own training cluster:

“we have the most powerful training cluster in the world right now, which is over 200,000 GPUs, training coherently.” 🔗

His admiration of China’s capabilities — talent and sheer build-out:

“I do want to emphasize that the sheer number of smart, talented people in China who work very hard is amazing.” 🔗

On breakthrough innovation — question authority

His sharpest mind-line in the interview: the cultural precondition for breakthrough innovation is the willingness to question authority — which he reads as the U.S. advantage:

“to have breakthrough innovation, you have to question authority. That fundamentally your breakthrough, you’re questioning the conventional wisdom when you do a breakthrough innovation.” 🔗

On how AI ends — Roddenberry vs Cameron

The intelligence-explosion framing — front-row, not in control:

“I feel like we’re, you know, we’re in the big bang of the intelligence explosion.” 🔗

The softened-but-not-retracted risk stance, after Faber notes he says “20% chance of annihilation” less often now:

“I think we should always consider that there’s some chance of a bad outcome, to try to protect against the bad outcome.” 🔗

The choice cast as two movies — and which one he is rooting for:

“are we in a Star Trek movie or like are we in a Gene Roddenberry movie or a James Cameron movie? Which movie are we in here? And you could either have a Roddenberry or a Cameron outcome. And let’s, I think in this case, we want the Roddenberry outcome.” 🔗

On control — just enough not to be ousted

Asked why he wants ~25% of Tesla, he frames it as just enough to not be removed, not true control:

“just enough control to not get ousted by activist investors at some point in the future” 🔗

Faber floats becoming “an Ellison-like figure” at Tesla; Musk redirects to describing Larry Ellison’s relationship to Oracle — “Owner, I think would be – he’s the owner of Oracle” — and then:

“He’s not the CEO, he’s just the owner.” 🔗

This last quote is Musk describing Ellison, not a label he applies to himself; it is recorded here only as context for the control exchange, not as a self-conception.

Provenance note: most of this interview is operational/financial (no Tesla–Uber deal, an xAI–Tesla merge “not out of the question” but with “no plans to do so,” his 2018 pay package). The wiki treats that as business context and does not mine it for mind-content; the control material above (the ~25% reasoning) is included only as his stated rationale for the stake he seeks, not as corporate-governance reporting.

Connections (pages touched)

• Humanoid robots — created: the demand thesis (“biggest product ever,” “insatiable”) and the staged child-like learning model (Mocap imitation → learning from video → self-play with a reward function).
• AI existential risk — extended: the 2025 softening of the “20% annihilation” line into a precautionary stance, and the Roddenberry-vs-Cameron framing of how AI ends; all prior content preserved.
• Humanity’s bright future — extended: the “big bang of the intelligence explosion / courtside seats / won’t be boring” optimism, the bright-future lens applied to the AI transition.
• Curiosity and truth-seeking — extended: “breakthrough innovation requires questioning authority,” his clearest civic-scale statement of the question-authority instinct (the China/US contrast).
• xAI and Grok — extended: the compute-frontier reasoning — “most powerful training cluster… over 200,000 GPUs,” the chips→electrical-equipment→power-generation limiter chain, and the gigawatt-class build-out.
• Elon Musk — extended with a “What the CNBC / David Faber (2025, secondary) interview reveals” section threading the robot, AI-frontier and AI-ending reasoning together, plus the ~25%-control rationale; all prior content preserved.
• Tesla — extended: the Airbnb-plus-Uber mental model for the autonomous fleet and “Tesla has all the ingredients… overnight,” his answer to “why not buy Uber.”
• Autonomous driving — linked: the shared-fleet logic restated here, already developed from the first interview.
• Sustainable abundance — linked: the robot demand claim sits adjacent to the abundance framing (Faber, not Musk, voices “sustainable abundance” here).