Anthropic's $2T IPO, Zuck's AI Manifesto, Nvidia's $500B AI Bet, Grok's Comeback

Summary

No explicit numeric entry or target prices were stated by any speaker for these tickers this episode — calls were qualitative (business-model/strategic theses), so Entry/Target/Δ are marked ”—”.

Theses (episode spine)

• Gavin is skeptical that the reported $2 trillion Anthropic IPO valuation is the real number, calling it likely underwriter sandbagging by whoever lost lead-left; he thinks it could trade to $3-4 trillion once the S-1 and roadshow results are visible.
• Sax argues Anthropic’s revenue growth can keep compounding given enormous AI demand, but physical constraints (compute, energy) and rising competition (OpenAI’s growth reportedly accelerating to ~20% month-over-month after its own pivot to coding) will determine how close they get to a $1 trillion run rate.
• Jason (Cal) is the outlier bear on Anthropic’s near-term growth: he expects it to slow considerably next year because cheap open-source models like GLM-4.6 are roughly 90% cheaper than Claude Opus and are increasingly being adopted by startups and corporate America; he doubts Anthropic reaches $1 trillion, guessing more like $300-400 billion.
• In an on-air bet, both Gavin and Sax take the under on Wall Street consensus for Anthropic’s 2027 revenue, each landing around $400-500 billion exit ARR next year versus roughly $100-120 billion this year.
• Sax and Gavin both praise Zuckerberg’s AI manifesto for reframing the debate as centralized versus decentralized power rather than simply open versus closed source, and both say they side firmly with decentralization and individual empowerment.
• Gavin argues the effective-altruist/Anthropic-aligned worldview believes AI is “too dangerous to distribute,” while Zuckerberg, Elon and Jensen believe it is “too dangerous to centralize” — he sides with the decentralizers, saying history has no real counterexample favoring centralization.
• Sax and Gavin dismiss the media narrative around data-center water and electricity costs as largely false, arguing data centers use less water than a golf course, tend to lower electricity costs when they build their own generation, and that rising blue-state electricity prices are actually driven by decarbonization policy.
• Gavin calls Nvidia’s $500 billion compute financing plan with Goldman, BlackRock and other banks/PE firms “brilliant” because it turns GPUs into a financeable, asset-backed asset class, removes the financing constraint that was capping the AI buildout’s TAM, and is not the circular financing Bill Gurley has warned about.
• Gavin says xAI’s Grok 4.6 and Grokbot releases show Elon has closed the gap with what was a two-lab (Anthropic/OpenAI) frontier duopoly in about six months, crediting the Cursor acquisition team and SpaceX veterans brought into xAI’s management; a larger Grok 4.7 model is expected within weeks.
• Jason and Sax disagree sharply on Amazon’s DSP driver-subcontracting model: Jason, a self-described large and growing long-term Amazon shareholder, argues Amazon should proactively convert delivery drivers to full employees to get ahead of a rising socialist political narrative, while Sax argues freedom of contract should prevail and that forced conversion would raise package costs by roughly $5.20 each and $664 per household per year in New York.

$NVDA (Nvidia)

Convergence / divergence: Both hosts view Nvidia’s move to help finance customer GPU purchases favorably — Gavin frames it as a structural TAM-unlock, Sax as a mechanical financing analogy — with no disagreement on the strategic merits.

Speaker calls:

• Gavin (bullish, unspecified timeframe): Calls Nvidia’s $500B compute financing initiative with Goldman, BlackRock, KKR and other banks/PE firms brilliant because it turns GPU compute into a financeable, asset-backed asset class, removes the financing constraint that was capping the AI buildout’s addressable market, and could make Nvidia effectively a capital-light cloud via revenue-share royalties above a guaranteed offtake floor.
• Sax (neutral): Compares the financing structure to airplane asset-backed financing and notes Nvidia has to standardize GPU reference designs so Wall Street can package and securitize them the way mortgages are securitized.

Cross-check:

• Price: $224.89 (P/E 34.3x, mkt cap $5.46T). Next earnings: 2026-08-26.
• Recent headlines worth knowing: Announced ~$500B AI-compute financing initiative with Goldman, BlackRock, KKR and others to turn GPU compute into a financeable asset class; earnings due Aug 26, 2026, just after this episode aired.
• Inconsistencies: None flagged — Gavin’s bullish framing of the financing deal is consistent with the stock trading near all-time highs at a still-elevated (though below its 12-month average) P/E.

$WDAY (Workday)

Convergence / divergence: Only Gavin offered a view; no divergent take from Sax or Jason on this story.

Speaker calls:

• Gavin (bullish, unspecified timeframe): Says the reported Silver Lake bid for Workday (stock up 17% on the news) signals private equity returning to software after a two-year death-spiral driven by AI/open-source pressure, calling it a sign that software names like Workday have been oversold and that a private-equity backstop for the sector is back.

Cross-check:

• Price: $198.68 (P/E 53.5x TTM, mkt cap ~$43-50B). Next earnings: 2026-08-20.
• Recent headlines worth knowing: Reuters/Bloomberg reported Silver Lake in talks to take Workday private at roughly $43B; shares jumped ~17-18% on the report, matching the 17% move cited on the podcast; deal is not confirmed as final.
• Inconsistencies: None flagged — Gavin’s characterization of a 17% pop matches reported figures; talks are still preliminary per reporting, so a completed buyout is not guaranteed.

$AMZN (Amazon)

Convergence / divergence: Jason and Sax diverge on policy (whether Amazon should voluntarily convert DSP drivers to employees), not on the stock itself — Jason’s investment stance is bullish/long-term while Sax offered no explicit position, only a cost-of-regulation argument.

Speaker calls:

• Jason (Cal) (bullish, long-term): Discloses he is a long-term Amazon shareholder with a very big position that he keeps increasing every year, and argues Amazon should proactively convert DSP delivery drivers to full employees (at an estimated cost of about 25 cents per delivery) to get ahead of a rising socialist political narrative around subcontracted labor.
• Sax (neutral): Argues Amazon’s DSP contractor model should be left alone under freedom of contract, warning that forcing full-employee conversion would raise package costs by about $5.20 each, add roughly $664 per household per year in New York, and put around 5,000 jobs at risk.

Cross-check:

• Price: $278.09 (P/E 22.1x TTM, 31.2x forward; mkt cap ~$3.0T). Next earnings: ~late Oct 2026 (Q3, exact date TBD).
• Recent headlines worth knowing: Stock hit a new all-time high and topped $3T market cap after a stronger-than-expected Q2 earnings beat reported July 31, 2026; New Jersey AG sued Amazon over its DSP driver subcontracting model, and NYC Mayor Mamdani pushed municipal rules to force direct employment of drivers.
• Inconsistencies: None flagged — Jason’s disclosed bullish long-term position lines up with the stock trading at all-time highs; his call to voluntarily raise driver costs is a policy opinion, not reflected in current valuation.

Topics discussed

Anthropic’s rumored $2 trillion IPO and revenue ramp

Summary: The Financial Times reported Anthropic is targeting a $2 trillion IPO as soon as October, which would top SpaceX’s $1.75 trillion record. The hosts debate whether that headline number reflects real value or is underwriter sandbagging, and whether the company’s roughly 10x year-over-year revenue growth (heading toward $100-120B run rate this year) can be sustained given rising competition from OpenAI and cheap open-source models.

Speaker views:

• Gavin: Thinks $2 trillion is likely a lowballed leak from bankers who lost the lead-left mandate, and that the real price could settle at $3-4 trillion; internally Anthropic is very confident, with people close to the company saying Dario has mused Anthropic could be one of very few private companies left standing.
• Sax: Believes growth can continue given massive AI demand and a large addressable market in knowledge work, but flags physical constraints — compute and energy — as the real limiting factor, alongside share loss to OpenAI (whose growth has reportedly accelerated to ~20% month-over-month after pivoting into coding) and open source.
• Jason (Cal): Expects Anthropic’s growth to slow considerably next year, arguing there is no way it reaches $1 trillion in revenue; points to GLM-4.6 pricing roughly 90% cheaper than Claude Opus and increasing corporate embrace of open-source models as the reason, guessing Anthropic lands closer to $300-400 billion.

Potential impact: The hosts argue Anthropic’s upcoming quarterly S-1 disclosures will become the AI industry’s key “pace car” signal — if its growth wobbles, they expect a chain reaction through SpaceX’s chip purchases, Nvidia’s revenue, and the broader AI capex buildout given the compute supply chain’s interdependence.

Zuckerberg’s AI manifesto: centralized vs. decentralized AI

Summary: Mark Zuckerberg published a roughly 6,500-word essay arguing for open-source, decentralized AI, framing superintelligence as an invention rather than automation and arguing AI safety should come from balance of power rather than centralized control. Both hosts read it and largely agreed with its framing, seeing it as a direct rebuttal to the effective-altruist worldview associated with Anthropic and past Biden-administration AI policy circles.

Speaker views:

• Sax: Says he found himself nodding along, agreeing Zuckerberg exposed a contradiction in effective-altruist doomers building a future they claim to fear; describes past Washington meetings where officials wanted to centralize AI control in a small government-blessed cartel, which he calls a pipe dream given the rest of the world will get AI regardless.
• Gavin: Frames the split as anthropic/EA types believing AI is too dangerous to distribute versus Zuckerberg, Elon and Jensen believing it’s too dangerous to centralize; says he sides with decentralization because history has no real example where centralized control produced better outcomes, and values a “rich variety” of competing AIs.

Potential impact: Both hosts argue that if the U.S. had gone the centralized/heavily-regulated route Anthropic’s own policy allies pushed for in Washington, it would have compressed the six-month lead time Anthropic needs to justify premium pricing, letting Chinese open-source competitors catch up faster and undermining both Anthropic’s IPO case and U.S. competitiveness.

Data-center water and electricity panic

Summary: The hosts push back hard on media coverage claiming AI data centers strain local water and electricity supplies, saying the narrative stems from an inaccurate figure (off by roughly 100,000x) and ignores that data centers can lower electricity costs when they build their own generation and sell excess back to the grid.

Speaker views:

• Jason (Cal): Says the water-use narrative is simply wrong (data centers reportedly use less water than a golf course, and water recirculates), and points to a Wall Street Journal story about a town called Ellendale that was revitalized by a new data center as an underreported counter-narrative.
• Sax: Argues rising electricity prices in blue states are actually driven by decarbonization and green-energy regulation rather than data centers, and that the only legitimate concern is noise pollution, which is manageable by siting facilities away from residential areas.

Potential impact: Both hosts frame this as a live political fight, noting Texas Governor Abbott’s newly announced energy-audit requirement for new data centers and warning that unchecked “moral panic” coverage could slow the U.S. buildout relative to China.

xAI’s Grok comeback: Grok 4.6 and Grokbot

Summary: xAI released Grok 4.6 and Grokbot this week, with benchmark charts (cited from Databricks, a $190B-valued company) showing Grok 4.6 near or at the frontier on quality while being notably cheaper than rivals. The hosts discuss how quickly xAI closed the gap with what had been a two-lab frontier duopoly.

Speaker views:

• Gavin: Says he had been calling it a frontier duopoly between Anthropic and OpenAI as recently as last week and now thinks that view was wrong; credits xAI’s acquisition of the Cursor team and a management reshuffle bringing in SpaceX veterans for closing the gap in about six months, with a larger Grok 4.7 model expected within weeks.
• Jason (Cal): Notes the tension in Elon simultaneously hosting Anthropic’s compute (with a contractual right to reclaim it on 90-day notice) while increasingly competing with them via Grok, framing xAI’s large compute clusters as both a call option to become a true frontier lab and a put option to sell spare compute to rivals at attractive prices.

Amazon DSP driver lawsuits and the gig-economy regulation fight

Summary: New Jersey’s AG sued Amazon over its use of subcontracted Delivery Service Partners (DSPs) to shield itself from driver-related liability and labor obligations, and NYC Mayor Zohran Mamdani has pushed municipal rules that would force Amazon to hire drivers directly. The hosts debate whether this is legitimate corporate cost-shifting or a necessary flexible labor arrangement.

Speaker views:

• Jason (Cal): As a long-term Amazon shareholder, argues Amazon should proactively convert DSP drivers to full employees (citing an estimated 25-cents-per-delivery cost) to get ahead of a socialist political narrative, drawing a parallel to Howard Schultz’s decision to raise Starbucks barista wages and benefits.
• Sax: Argues the DSP arrangement is a mutually beneficial, market-evolved structure that should be left to freedom of contract, citing a study estimating forced conversion would raise per-package costs by about $5.20, cost New York households roughly $664 per year, and put about 5,000 jobs at risk.

Potential impact: The hosts frame this as a preview of a 2026 midterm and 2028 campaign issue, with Democratic socialists likely to use gig-economy subcontracting as a core talking point against large employers.