PodcastAll-In PodcastJason Calacanis

Elon’s Anthropic Deal, The Next AI Monopoly?, “FDA for AI” Panic, Trading the AI Boom

The conversation centered on two tensions: how to regulate frontier AI without enabling regulatory capture, and whether compute capacity—limited by data‑center supply and power—will choke or amplify AI monetization. Rapid ARR expansion at Anthropic and OpenAI is reshaping market positioning and political scrutiny, while hyperscalers' cash flow and leased compute models create structural advantages. Speakers flagged near‑term infrastructure risks (protests, power constraints), recommended urgent cybersecurity hardening, and endorsed a roughly 500‑day bullish window as enterprise AI spend begins to show ROI.

AI SafetyArtificial IntelligenceMarket Breadth

Preview · 18 more points in the app

Read the Full Summary

Key Takeaways

Policymakers debating an 'FDA‑style' pre‑approval regime for AI face a real tradeoff: while guests urged faster, resourced government review and cyber hardening, they warned such an approval regime could enable regulatory capture and let Washington 'pick winners and losers.'

Hyperscale compute leasing (e.g., xAI leasing Colossus) is enabling frontier model deployment by deferring capex, but data‑center power limits and local opposition—roughly 9 GW of near‑term buildout with about half protested—threaten overall capacity expansion and therefore revenue ceilings for firms like OpenAI.

The three incumbent cloud giants have the financial firepower to dominate AI infrastructure—AWS (~$150B), Azure (~$108B) and Google Cloud (~$80B) run‑rates and strong free cash flow—yet new plays like Elon’s proposed NeoCloud (leveraging SpaceX/Starlink, Tesla assets, Powerwalls) aim to create an AWS‑style distributed rival.

Anthropic and OpenAI’s very rapid revenue acceleration (Anthropic reportedly tripling ARR Jan–Mar and further April gains; OpenAI showing fast monthly revenue) is concentrating market positioning and is already drawing heightened political and regulatory attention.

Hosts estimate coding represents roughly a $1 trillion annual TAM, but capturing it requires massive multi‑gigawatt compute scale; that constraint tempers upside even as early AI monetization and enterprise spend (coding tokens) suggest a near‑term bullish window of roughly 500 days while ROI materializes.

Speakers recommended equipping established cybersecurity vendors with frontier models and instituting identity‑based (KYC) controls during preview periods because frontier models are rapidly gaining cyber capabilities—creating an urgent need to harden systems and scan codebases now.

GEO-POLITICS

FDA‑Style Oversight — Reports that the White House is considering an FDA‑style pre‑approval regime for AI models sparked concern; guests argued for faster, well‑resourced reviews (finite review windows and cyber hardening) but warned such a regime could let Washington 'pick winners and losers' and enable regulatory capture.

“New York Times reported Trump is considering an executive order to create an, quote, AI working group…”—

TECHNOLOGY

Compute Leasing — Hyperscalers and new entrants are leasing large‑scale AI capacity (e.g., Colossus) so firms like xAI can run frontier models while deferring heavy upfront capex and competing in hyperscale.

“anthropic just added over 220 000 nvidia gpus over 300 megawatts of energy the deal is already having an impact as we've discussed here cloud users have been experiencing rate…”—

Anthropic Momentum — Speakers cited Anthropic's reported ARR tripling Jan–Mar plus further April gains and OpenAI's fast monthly revenue as the reason funds have increased exposure to compute/memory names and concentrated portfolio bets.

“First, we find out that from January 1st to March 31st, they grew from roughly 10 billion of ARR to 30 billion, so it tripled…”—

Coding TAM — Hosts estimated coding represents roughly a $1 trillion annual TAM and argued capturing large shares requires massive compute scale (multi‑gigawatt targets), prompting labs to pivot to coding products like Codex/GPT‑5.5 to chase monetization.

“there they expect to have five gigs by the end of this year 10 gigs by the end of next year It's kind of hard to get to those numbers…”—

ECONOMY

Data‑Center Limits — OpenAI's revenue is described as supply‑constrained by data‑center availability and, critically, power rather than end‑user demand; regional electricity pricing differences tied to local buildouts further limit near‑term revenue growth.

“The first is, as I mentioned a couple of weeks ago, and OpenAI's revenue performance has nothing to do with demand. Zero…”—

Capture Risk — Speakers warned firms may use safety narratives and regulation to insulate themselves from competition, again invoking Rockefeller analogies and citing model access restrictions as potentially anti‑competitive measures.

“I just want you to think for a second about the case of John D…”—

Cyber Defense — Guests recommended equipping established cybersecurity vendors with frontier models and enforcing identity‑based controls (KYC) during preview periods; labs already track API use and coordinate with government on suspicious activity and anti‑distillation.

“We have a giant cybersecurity industry in the United States whose sole job it is to protect systems and protect against breaches…”—

MARKETS

Hyperscaler Run‑rates — Hyperscaler cloud revenue is a central market driver: AWS, Azure, and Google Cloud were cited at roughly $150B, $108B, and $80B run rates respectively, with strong growth underpinning ongoing AI infrastructure demand.

“AWS is now on $150 billion run rate. Azure, 108 billion. GCP, Google Cloud, 80 billion.”—

FORECASTS

Retirement Protection — Speakers expected the Retirement Protection and Savings Act to pass decisively in California and argued it would effectively block a state wealth tax by constitutionally protecting retirement savings.

“The ballot initiative, the Retirement Protection and Savings Act, right, it's going to pass, going to pass with big numbers…”—

Donor Retreat — After an attack video targeted his house and safety concerns mounted, Ken Griffin said he'd shift efforts to Florida, illustrating how personal targeting can prompt major donors to reallocate political involvement.

“Mondami did an attack video on Ken Griffin's house…”—

Compute Delays — Speakers reported about 9 GW slated for near‑term buildout with roughly half reportedly protested—predicting delays or cancellations—and discussed activist coordination and policy shifts that let AI firms self‑provide power as reshaping infrastructure dynamics.

“You have about nine gigawatts that are supposed to come online this year. Almost 50% of it now is being protested…”—

Related Episodes

Important Disclaimers

Not Investment Advice: For education only — not financial, tax or legal advice.

AI-Generated: Made by AI from source transcripts and may contain errors.

No Guarantees: Past performance does not predict future results; investing carries risk.

Our Commentary · Request Removal