More Trillion Dollar IPOs, Anthropic $3T, Zuck's Price War, China Ends Open Source?, Trump Accounts

Summary

Theses (episode spine)

• Chamath and Brad both expect Anthropic and OpenAI to IPO within the next 6-9 months, using SpaceX’s “textbook” IPO (raised $75B at $1.75T, now trading near $2T on ~$35B forward revenue, up ~25% from the offer price) as the template for pricing, lockup structure, and index inclusion.
• Brad says Altimeter would be “a buyer at scale and at size” in both Anthropic and OpenAI IPOs even at valuations up to $3 trillion, citing Anthropic’s rumored trajectory toward $100B+ revenue this year.
• Chamath warns that AI token costs at his portfolio company 8090 are doubling every 45 days while measured productivity gains are only about 5%, and argues this looming ROI reckoning is exactly why the labs should IPO now, before it “seeps into the water table.”
• Brad counters that enterprise AI adoption is still extremely early relative to a TAM of every company on the planet, and that Anthropic/OpenAI could plausibly 3-5x revenue again next year if they exit 2026 above $100B combined.
• Despite open-source and cheaper models improving, the frontier labs are not losing share of wallet: Sax cites data showing open source’s share of enterprise AI spend fell from 19% to 11% year over year, and Brad calls the market a “duopoly” dominated by Anthropic ($60B+ ARR) and OpenAI ($40B+ ARR).
• Sax argues most enterprises would like to diversify off expensive closed frontier models to cut costs, but lack the technical capability to build the routing/harness middleware needed to do so, leaving convenience-driven demand for Claude and GPT models intact.
• Reuters reported China may restrict overseas access to its own top AI models (Alibaba, ByteDance, Zipu/GLM) over IP-theft and national-security concerns; Sax believes this would hurt China more than the US and is likely “chess playing” rather than a genuine threat.
• Chamath flags energy, not chips or software, as the real constraint on US AI scaling, citing an internal analysis showing the US will be roughly three Californias’ worth of power short of projected 2050 demand.
• Brad Gerstner detailed the launch of Trump (“Invest America”) accounts, seeded with $1,000 at birth and invested in the S&P 500: 1.5M+ accounts were created and $1B+ deposited in the first 24 hours after the July 4 launch, with pledges including $6B+ from Michael & Susan Dell, $350M in SpaceX shares from Gwynne Shotwell, and $250M from Micron.
• Sax highlights the Trump accounts as an unusually powerful tax-advantaged savings vehicle — employers can contribute up to $2,500 tax-free per child, and the accounts can later roll into a Roth IRA, effectively giving every child an IRA from birth.

SPACEX (SpaceX)

Convergence / divergence: All three see the IPO as a success and a template for future frontier-AI listings; there’s no real divergence here — Brad and Chamath are more forward-looking bullish (“meat on the bone,” a smart pre-emptive move), while Jason’s framing is more descriptive of the round-trip in price action.

Speaker calls:

• Brad (bullish, long-term): “Calls SpaceX’s IPO textbook: it priced at $1.75T, is now trading near $2T on roughly $35B of forward revenue (up about 25% from the IPO price), and he still sees ‘a lot of meat on the bone.’”
• Chamath (bullish): “Says it was smart for Elon to take SpaceX public first, ahead of an industry-wide AI token-cost and ROI reckoning he sees coming.”
• Jason (neutral): “Notes SpaceX ran up to $200/share post-IPO before settling back to about $150 (right at the IPO price), making it the seventh-largest company in the world.”

Cross-check:

• Price: $145.39 as of 2026-07-10 (SPCX), market cap ~$1.93T. No P/E disclosed. No earnings date confirmed yet post-IPO.
• Recent headlines worth knowing: IPO’d June 12, 2026 on Nasdaq at $135/share; jumped 19% on debut to ~$2.1T cap; hit an all-time high of $225.64 on June 16; has since pulled back amid a broader ~$600B AI-linked sell-off. Average 12-month analyst price target ~$242 (range $62-$800).
• Inconsistencies: Brad’s “trading near $2T” roughly matches the podcast-day figure, but the current cap (~$1.93T) is below the post-IPO high — the stock has round-tripped much of its post-IPO pop.

ANTHROPIC (Anthropic, private, pre-IPO)

Convergence / divergence: Both see an Anthropic IPO as imminent and desirable, but for different reasons — Brad frames it around growth and demand from crossover buyers like Altimeter, while Chamath frames urgency around getting ahead of a coming AI-ROI skepticism cycle.

Speaker calls:

• Brad (bullish, next 6-9 months, target up to $3T): “Says Anthropic is rumored to be trending over $100B in revenue this year, and that Altimeter would be an enthusiastic buyer at scale even at a $3 trillion IPO valuation.”
• Chamath (neutral): “Argues the labs should get their IPOs done now, before rising token costs and flat productivity gains force investors to demand proof of ROI.”

Cross-check:

• Price: N/A (private). Last private mark ~$965B (Series H, closed ~June 2026). Targeting an October 2026 Nasdaq listing per a confidential S-1 filed June 1, 2026.
• Recent headlines worth knowing: raised $65B at the $965B valuation; revenue run-rate reported near $47B by late May 2026 (up from ~$9-10B at end of 2025); reportedly on pace to exceed $50B run-rate and possibly reach its first profitable quarter.
• Inconsistencies: Brad’s on-air claim of “trending over $100B in revenue this year” and a $3T IPO target is well above the ~$47B run-rate and ~$965B private mark found in press coverage — a large gap worth flagging as an aggressive/optimistic framing rather than a reported figure.

OPENAI (OpenAI, private, pre-IPO)

Convergence / divergence: Both agree OpenAI trails Anthropic in IPO readiness and could go out second, but they diverge on the reason — Brad cites corporate-restructuring complexity, while Chamath flags the underlying business quality (cash burn, consumer mix) as the concern.

Speaker calls:

• Brad (bullish, long-term, target over $1T): “Says OpenAI’s revenue has ‘ticked back up’ to a rumored ~$70B run rate and it could still IPO above $1 trillion, though corporate-restructuring complexity likely means it goes out after Anthropic.”
• Chamath (bearish): “Notes OpenAI’s cash burn was still quite high last they heard, given its more consumer-reliant, diffuse business compared to Anthropic’s enterprise focus.”

Cross-check:

• Price: N/A (private). Last private mark ~$852B (from a ~$122B round earlier in 2026). Confidential S-1 filed June 8, 2026, originally targeting a September 2026 IPO.
• Recent headlines worth knowing: revenue run-rate reported around $25B annualized in early 2026 (~34x sales at the $852B mark); heavily loss-making, with internal projections of ~$14B in losses for 2026 and profitability not expected until ~2030; more recent reporting suggests OpenAI may push the IPO to 2027; price-war dynamics with rivals (e.g., Meta) cited as a risk.
• Inconsistencies: Brad’s on-air claim of a ~$70B revenue run-rate is roughly 3x the ~$25B figure found in press coverage — worth flagging as notably higher than public reporting suggests. Also, reports of a possible 2027 IPO delay would push out the “trillion-dollar IPO” timeline discussed on the show.

Topics discussed

Trillion-dollar AI IPO wave (SpaceX, Anthropic, OpenAI)

Summary: The hosts reviewed SpaceX’s “textbook” IPO — raised $75B at a $1.75T valuation, now trading near $2T on about $35B forward revenue — as the blueprint for Anthropic and OpenAI’s expected IPOs. They discussed controversy over fast-tracking SpaceX into major indexes given its size and the historical volatility of newly public companies (peak-to-trough drawdowns of ~50% in the first six months).

Speaker views:

• Brad Gerstner: SpaceX’s index inclusion was handled well despite legitimate concerns about jamming a newly public, untested stock into passive funds; he expects Anthropic and OpenAI to follow the same playbook and both to be public within roughly a year.
• Chamath Palihapitiya: The key question for any of these IPOs is what price the market will clear at, not whether the businesses are good; he thinks it’s smart to IPO before industry-wide AI ROI concerns intensify.

Potential impact: A blockbuster Anthropic or OpenAI IPO at a $1-3T valuation would be a major test of public-market appetite for AI companies and could set the template (lockups, index timing) for future frontier-lab listings.

AI token cost economics and enterprise ROI

Summary: Chamath described a portfolio company (8090) whose token costs are doubling every 45 days while productivity gains are only about 5%, framing this as an industry-wide ROI reckoning. Brad countered with examples like Uber and DoorDash showing early but rapidly compounding enterprise adoption across a massive addressable market.

Speaker views:

• Chamath Palihapitiya: Token spend is compounding much faster than measurable productivity gains, and at some point sophisticated investors will demand hard evidence of ROI from companies spending millions on tokens.
• Brad Gerstner: Enterprise adoption is still early relative to a TAM covering every company on the planet, and revenue growth of 30%+ a year on already-massive bases is unprecedented; he expects Anthropic/OpenAI could 3-5x again next year if they exit 2026 above $100B.

Potential impact: If an earnings miss or broader AI ROI disappointment occurs, Sax and Chamath suggested companies would first cut costs (including AI token spend) rather than lay off staff, which could pressure token demand and frontier-lab revenue growth.

Frontier labs vs. open-source models — duopoly forming?

Summary: The hosts debated whether cheaper open-source and Chinese models (e.g., GLM 5.2) are eroding demand for frontier models like Claude and GPT. Sax cited data showing open source’s share of enterprise AI spend fell from 19% to 11% year over year, even as usage of both categories grows.

Speaker views:

• David Sacks: Enterprises would like to shift spend to cheaper open models to control costs, but most lack the technical ability to build the routing middleware required, so closed frontier models keep gaining share of wallet by being the most convenient choice.
• Brad Gerstner: The market is becoming a duopoly by revenue — Anthropic at roughly $60B+ ARR and OpenAI at roughly $40B+ ARR dwarf everyone else — and there is no evidence on the field today that commodity open models are closing the intelligence/revenue gap.
• Chamath Palihapitiya: Raises the non-consensus possibility that intelligence does not converge at all — if models become recursively self-improving, the frontier’s lead over open source could widen rather than close over the next 2-3 years.

Potential impact: If the duopoly thesis holds, it strengthens the case for premium IPO valuations for Anthropic and OpenAI; if convergence happens instead, it favors verticalized/open-source players and pressures frontier-lab pricing power.

China considering restrictions on its own open-source AI models

Summary: Citing Reuters reporting, the hosts discussed that Chinese regulators met with Alibaba, ByteDance, and Zipu (GLM) about limiting overseas access to top Chinese models, driven by IP-theft and national-security concerns, including fear the US could weaponize a model against Chinese interests.

Speaker views:

• David Sacks: If China wants to harm the US, restricting open models it benefits from makes some sense, but he thinks the story is probably overstated and is more likely Chinese “chess playing” than a real threat, since it would hurt China more than the US.

Potential impact: Sacks argued the US should continue supporting open-source development (citing Nvidia and Reflection) rather than restrict its own labs, and floated (half-jokingly) that a Chinese AI-safety “doomer” movement would help the US win the AI race by slowing Chinese labs.

Energy as the binding constraint on AI scaling

Summary: Chamath cited an internal team analysis showing the US will be roughly three Californias’ worth of power short of projected 2050 demand even before accounting for AI. Jason added that Taiwan’s reliance on LNG (only 2-3 weeks of reserves) creates a parallel energy risk tied to chip supply if China blockades Taiwan.

Speaker views:

• Chamath Palihapitiya: The US has an enormous energy shortfall coming based on projected load growth to 2050, independent of AI-specific demand.
• Jason Calacanis: Taiwan’s limited LNG reserves mean a Chinese blockade could immediately cut off energy needed to keep chip production running, compounding the US energy bottleneck.

Potential impact: Both hosts framed energy (nuclear, solar, batteries) rather than chips or software as the real gating factor for further AI infrastructure buildout in the US.

Meta’s AI pricing strategy shift

Summary: The hosts discussed Mark Zuckerberg’s announcement of MuseSpark 1.1, a coding model he claims matches frontier quality at a fraction of the cost, framing it as a shift from Meta’s earlier open-source strategy toward a direct price war with the frontier labs.

Speaker views:

• Chamath Palihapitiya: Believes Meta “flubbed” its original open-source strategy with Llama but is now pivoting to compete on cost, calling the price-war framing an interesting new vector of competition.

Trump (“Invest America”) accounts launch

Summary: Brad Gerstner detailed the July 4 launch of Trump accounts — $1,000 seeded at birth into an S&P 500-linked account, with $5,000/year contribution limits from family and up to $2,500 tax-free from employers. The app hit #1 in downloads with 1.5M+ accounts created and $1B+ deposited in the first 24 hours, with philanthropic pledges from Michael & Susan Dell (over $6B, $250 to 25 million children), Gwynne Shotwell/SpaceX ($350M in shares), and Micron ($250M, up to $1,000/employee).

Speaker views:

• Brad Gerstner: Frames the program as potentially the largest direct philanthropic platform in US history, targeting $100B raised in the first 12 months, and says the administration’s goal is to auto-create accounts for all 50-70 million children under 18 within 90 days.
• David Sacks: Highlights the accounts’ tax advantages as historically unusual — comparable to maxing a 401(k) match or Roth IRA — since contributions and growth are tax-free, employer contributions up to $2,500 are tax-free, and the account can later roll into a Roth IRA, potentially compounding to $10M+ by retirement if maxed from birth.

Potential impact: Sacks and Gerstner argued this could shift the share of Americans who own equities from roughly 50% toward 70-75%, and Gerstner suggested it could eventually serve as a supplemental, ownership-based alternative alongside (not replacing) Social Security.