All-In PodcastAnthropic's $2T IPO, Zuck's AI Manifesto, Nvidia's $500B AI Bet, Grok's Comeback
CHAPTERS
- 0:00 – 2:28
Gavin Baker returns: SpaceX glow, politics chit-chat, and setting the agenda
The episode opens with Jason welcoming Gavin Baker and David Sacks, trading banter about investing, SpaceX momentum, and recent political interviews. They quickly pivot to the show’s packed docket focused on AI business fundamentals, regulation narratives, and market structure.
- •Gavin reflects on SpaceX as a ‘magical moment’ and moving on to the next opportunity
- •Sacks and Jason discuss Rahm Emanuel and the Democratic Party’s internal factions
- •Tone-setting: skepticism of ideological extremes and focus on real-world execution
- •Transition into breaking AI/markets news
- 2:28 – 9:24
Anthropic’s rumored $2T IPO: timing, run-rate shock, and banker ‘lead-left’ games
Jason lays out the Financial Times report: an October IPO window, unprecedented $100B+ run-rate, and a $2T valuation headline. Gavin argues the $2T figure may be strategic posturing from bankers, and that pricing discipline matters to avoid post-IPO volatility.
- •FT report: October IPO chatter; Polymarket odds and market signaling
- •Valuation math: ~16–20x sales vs other high-multiple public comps
- •Gavin: leaks often come from banks jockeying for lead roles; possible sandbagging
- •Anthropic’s product ‘harness’ retains value even if model leadership narrows
- 9:24 – 11:56
Can the ramp continue? TAM vs physical constraints (compute, energy, data centers)
Sacks and Jason stress the core question: exponential growth implies massive future ARR, but could hit hard limits in compute and energy availability. The discussion frames AI demand as enormous while emphasizing that supply-side realities may be the gating factor.
- •Sacks: 10x growth implies eye-popping next-year ARR projections; is TAM big enough?
- •Demand likely expands via agents and new use cases; token consumption grows
- •Main constraint may be physical: compute supply, power generation, and energy logistics
- •Competitive dynamics: OpenAI and others pivoting (especially to coding) to reclaim share
- 11:56 – 24:05
Building the AI ‘atoms’: turbines, natural gas, and the data-center backlash narrative
Gavin describes the gritty operational reality of scaling power and infrastructure—‘atoms, not bits’—and argues the U.S. can ramp manufacturing and deployment quickly. The trio pushes back on media claims about water and electricity impacts, emphasizing local tax benefits and manageable externalities.
- •Infrastructure build-out requires heavy industry execution: turbines, remote deployment, coordination
- •Examples: repurposing jet engines as turbines; capacity expansion at major industrial suppliers
- •They argue common claims are wrong: water usage exaggerated; grid effects misunderstood
- •Politics/regulation risk: energy audits, permitting, and public perception campaigns
- 24:05 – 27:32
Open-source pressure vs premium frontier tokens: pricing, commoditization, and why diversity matters
Jason argues open-source models could force a meaningful pricing reset, especially in enterprise, echoing historical open-source adoption. Gavin counters with a pro-open-source stance for strategic diversity, while also claiming open-source can make frontier models more valuable as orchestrators.
- •Jason: open-source is dramatically cheaper and increasingly viable for startups and enterprises
- •Corporate precedent: open-source adoption to avoid vendor lock-in and ‘hostage’ pricing
- •Gavin: key benefit is diversity of AIs—avoid dominance by 1–3 centralized models
- •Frontier vs open-source equilibrium: frontier retains premium value for best-in-class outcomes
- 27:32 – 45:07
Zuckerberg’s AI manifesto: decentralization, individual empowerment, and a critique of ‘AI nuclear’ framing
They break down Meta’s long essay positioning: superintelligence as invention, broad distribution via open models, and safety through balance of power. Sacks frames it as a direct rebuttal to effective-altruist centralization, arguing decentralization and competition are historically safer.
- •Zuck: if AI future is dystopian, why rush to build it? (shot at ‘doom’ narratives)
- •Safety framing: concentration of power is risky; decentralization empowers individuals
- •Sacks’ Washington perspective: prior pushes for cartelization/commission-style regulation
- •Argument: AI ≠ nuclear weapons; consumers and businesses universally need AI access
- 45:07 – 56:32
‘Go direct’ leadership: CEOs using X as strategy, protection, and narrative control
Jason highlights a broader trend: major CEOs writing directly to the public via long-form posts on X. Gavin argues having an authentic, personal communication channel is now a form of reputational armor—especially in politicized scrutiny environments.
- •Trend: leaders (Zuck, Jensen, Nikesh) increasingly publishing long-form directly
- •Jason: Zuck’s post appears human-written; signals a more mature communications stance
- •Gavin: relying solely on corporate PR is risky; personal voice is ‘protection’ and ‘power’
- •Meta/Zuck rebrand chatter blends culture, authenticity, and strategic positioning
- 56:32 – 58:03
All-In Summit announcements: speakers, festival build-out, and experiential ‘frontier tech’ programming
Jason shifts into event mode, announcing new Summit speakers and describing a more immersive festival-style setup. The segment emphasizes demos, wellness and gifting suites, and upcoming party/DJ announcements.
- •Summit dates/location: Sept 13–15 in Los Angeles at the Shrine
- •New speakers: Steve Hilton, Spencer Pratt, Abigail Shrier (+ surprise guest), Blake Scholl
- •Expanded format: AI/robotics demos, merch gifting suite, wellness/IV/red-light offerings
- •Call to action: applications at theallinsummit.com
- 58:03 – 1:03:18
Nvidia’s $500B financing vision: GPUs as an asset class, residual value guarantees, and ‘AI central bank’ vibes
The group unpacks Nvidia’s plan with Goldman/BlackRock and others to finance AI compute at massive scale. Gavin explains why top asset managers validating GPU cashflows matters, and how residual value guarantees plus revenue-share structures could lower financing costs and expand the market.
- •Structure: shift from upfront CapEx to financed systems with compute rental cashflows
- •Gavin: big-name capital validates GPUs as financeable, long-lived assets
- •Residual value guarantees and revenue share/royalty-like economics for Nvidia
- •Sacks: removing financing constraints unlocks the market’s true TAM; analogy to plane leasing
- 1:03:18 – 1:14:22
How the AI market could ‘pile up’: overbuild risk, pace-car signals, and why Anthropic’s S-1 matters
Sacks steelmans the downside: compute overbuild could create ‘dark GPUs’ like post-dotcom dark fiber. They argue Anthropic’s public reporting could become the industry’s demand barometer, and that profitability disclosures may break bearish narratives about subsidized tokens.
- •Primary systemic risk: compute glut and collapsing spot prices, not demand weakness per se
- •Jason: ‘governors’ on overbuild include physical build constraints and customer buy-button dynamics
- •Anthropic IPO/earnings could become the key demand ‘pace car’ for the whole supply chain
- •Gavin: many macro/value bears assume token subsidies; argues profitability is real up the stack
- 1:14:22 – 1:27:31
Amazon DSP subcontracted drivers: accountability vs flexibility, and why this fights into 2028 politics
A lawsuit and political push in NJ/NY renew scrutiny of Amazon’s Delivery Service Partner model. Sacks argues for freedom of contract and warns costs would rise sharply; Jason argues Amazon is externalizing risk and should proactively improve worker treatment to defang socialist narratives.
- •NJ AG lawsuit + NYC political pressure to force Amazon to hire drivers directly
- •Sacks: DSP model enables flexible scaling, lowers costs, supports small businesses
- •Jason: model can shield Amazon from liability/benefit obligations; invites populist backlash
- •Debate ties to broader ideological battle: corporate trust, worker protections, and cost of living
- 1:27:31 – 1:35:19
Grok’s comeback: Grok 4.6 pricing, benchmark positioning, GrokBot UX, and SpaceX’s ‘call/put option’ strategy
They assess xAI’s Grok 4.6 as moving onto (or near) the frontier at disruptive pricing, with early positive vibes from developers and benchmarks. The conversation frames SpaceX/xAI’s strategy as both competing at the model layer while monetizing compute as a fallback, creating a powerful optionality loop.
- •Grok 4.6: strong cost-capability ‘Pareto frontier’ positioning; third-party benchmarks cited
- •Operational reasons for speed: bringing in SpaceX ‘aces’ and leveraging strong engineering execution
- •GrokBot: easier on-ramp for non-coders vs more complex agent frameworks
- •Strategic tension: hosting Anthropic compute while competing; optionality via selling spot compute
- 1:35:19 – 1:39:29
Breaking news close: Silver Lake reportedly in talks to buy Workday—what it signals for software valuations
Late in the episode, Sacks flags a report that Silver Lake may acquire Workday, framing it as a meaningful turn for battered software stocks. They argue PE interest returning could reflect a market belief that the worst is priced in—and that open-source AI may ultimately backstop parts of the software ecosystem.
- •Report: Workday jumps on acquisition chatter; interpreted as PE re-engaging in software
- •Sacks: software drawdown reflects fear of model-layer dominance; open source helps re-balance
- •Jason: ‘AI-first’ operational playbooks can make legacy software more profitable post-acquisition
- •Implication: public market pessimism can create buyout opportunities and valuation floors
