Most important take away

This was the “most aggressive quarter in American capitalism” — the five largest companies on the planet are accelerating revenue at 20-30%+ while letting CapEx grow 50-60% and consume most of their free cash flow. The structural reality: hyperscalers are largely doing distribution and infrastructure work for two privately held LLM companies (Anthropic and OpenAI), so the durable upside increasingly accrues to the model layer (Anthropic raising $50B in 48 hours at ~$900B is the proof) and to enterprise transformation players like Palantir that can move corporate AI initiatives in $10M+ chunks.

Summary

Actionable insights and investment / career takeaways from the discussion:

Mag 7 earnings — buy/sell signals

• Buy Amazon, sell Microsoft. Amazon ($181B revenue, AWS $37B, fastest growth in 15 quarters) now has Anthropic models on AWS distribution and benefits from the application boom. Microsoft is the warning: strip out AI/Copilot/Azure-AI growth and the rest of the company is flat to slightly down. If the AI bet is wrong, the valuation premise is wrong.
• Alphabet was the clear winner this quarter (cloud backlog nearly doubled to $462B, ~60% cloud growth), but the panel argues it is “too boring to buy” at this valuation; was the right call a year ago.
• Meta crushed earnings ($56B revenue, EPS $10.44) but got punished because CapEx jumped from $125B to $145B with no clean attribution. Wall Street can model Google’s CapEx-to-revenue; it cannot model Meta’s. Meta is also the only Mag 7 not benefiting from the application/token-selling boom — least preferred of the four.
• Apple beat across the board, no AI story, just buybacks and margins. Memory price inflation (driver: AI demand) is starting to push hardware prices up — watch this as a stealth-inflation theme.

The structural thesis (most actionable framing)

• The five biggest market-cap public companies are effectively doing distribution and CapEx for two private LLM companies (Anthropic, OpenAI) that own the IP. Anthropic’s tokens grew ~15x in Q1; Gemini went from 10B to 16B tokens/min Q4 to Q1 — i.e. the most aggressive quarter in American capitalism is still under-performing the privates.
• Implication: the attractive sliver of value over time accrues to the model layer. Of the panel’s “one dollar to deploy” question — Anthropic over Sierra, not even a contest, despite Sierra’s downside protection.

Anthropic $50B raise at ~$900B

• Raised in ~48 hours by email, no IPO blocks, no governance rights. As long as you can do that in private markets, you don’t have to go public. Both Anthropic and OpenAI may slip IPO timing into 2027. Every $1 of Anthropic revenue requires ~$3-4 of CapEx, and forecasting one year out you’re committing ~$30B of CapEx per $1B of run-rate revenue — there is no such thing as too much cash on the balance sheet.

Palantir — most concrete actionable name

• Stock: $349B market cap, RPO up 134% to $4.45B, rule of 40 at 145%, gap revenue growth ~32%. Priced beyond perfection but if the AI buying compression Carp described is real, growth could justify it.
• Career/strategy insight: Palantir is the only software company that can credibly sell $10M-$100M AI transformation initiatives to Fortune 500 CEOs. Point-solution AI startups (Sierra, Harvey, etc.) max out around $200K-$2M deals — too small to be a CEO’s “top 2 initiative.” Big companies have to spend big to do big things.
• Knock-on opportunity: HubSpot/Shopify/Salesforce implementation agencies that develop genuine AI deployment expertise will see “infinite demand.” Most won’t make the transition.

SaaS apocalypse — partly over for the right names

• Atlassian (+29%), Twilio (+20%), Five9 (+23%) all beat. Five9 dismissed as not interesting at 9% reaccel.
• Two-prong test for AI-beneficiary SaaS: (1) monetize AI in your installed base, and (2) attract net new customers. Atlassian does (1) but maybe not (2). Twilio does both — net new customer count may be up ~40% YoY because every AI startup (11Labs, Sierra at $15B running on Twilio, Replit, Lovable) uses it. Twilio is the more interesting buy.
• Watchlist: HubSpot announced an “agents at parity with humans” platform. If this works it should drive HubSpot reacceleration in 12 months and signal hope for the rest of the GTM SaaS stack. If it fails there, write off the rest of the category.
• Pattern for trading legacy SaaS: when these companies are priced as if going to zero (3x revenue), there’s often a 2x bounce available; when they get to 6-7x they’re vulnerable to correction.

Sierra at $15.8B / $950M raise / 105x revenue

• Worry: tam expansion is being flattered by the “replace headcount” narrative. CX software market is ~$20-30B today, labor spend $400B; competing AI-first vendors will compete each other down. Hundred-times multiples lean very hard into the future.
• Counter-narrative: the LLMs-eat-everything thesis is wrong — Sierra and Palantir both prove there’s real value to be added on top of LLMs. Token spend is sub-10% of COGS for these companies; the value is in software + domain knowledge.

Token economics — the key research question

• The big unknown: what is the steady-state token spend as a % of salary for a fully AI-first organization? At 20-30% it justifies multi-hundred-billion / half-trillion dollar markets. At 5% it’s a much smaller story.
• Coding is the canary in the coal mine. Cursor at $44B run rate ($100M/day) implies very heavy token usage per developer.
• Surprising data point from Lemkin: SaaStr’s autonomous AI VP of Marketing + AI VP of Customer Success cost a combined $254/month in tokens to run, replacing multiple humans. Engineering token intensity in their portfolio sample is 2-15%, not 20%. Implication: non-coding agentic workflows may be far cheaper than expected, which is bullish for adoption but raises questions about Cursor-scale revenue durability.
• Hire more AI-pilled engineers, not fewer: if 20% token spend triples a $200K engineer’s productivity, you hire 15 instead of 10, not 5. ROI on developers goes up.

Career and management advice (heavily emphasized)

• The “$250K SDR”: there is now a market for very small numbers of $250K SDRs who can drive 20x output via AI. There is essentially no market left for a $60K SDR. Develop AI skills like a laser or get displaced.
• Coinbase laying off 15% saying “no more managers of managers — every employee must be an individual contributor.” Lemkin’s framing: “Anyone on LinkedIn that talks about their team, fire them.” A CMO today should be able to run their own campaigns by directing agents — the 1-2% of CMOs who can will dominate.
• Lead from the front with AI. Spend 10%+ of your time using the technology yourself or you will be in the grip of the experts and disconnected from reality.
• Brian Armstrong’s broader pattern: when he says something publicly, assume less cant and hypocrisy than the average CEO blaming AI for layoffs. His “no politics at work” call aged well; this one likely will too.
• Work-from-home Fridays were dismissed as a 3-day weekend; do not invest in companies that operate that way and do not invest your career in them if you want to compete.

Musk vs Altman trial — implications

• Musk admitted xAI partly distilled OpenAI models. Greg Brockman’s stake disclosed at ~$30B.
• The legal merits will likely turn on technical issues most coverage ignores: statute of limitations (Musk may have known too early to bring the case in time) and standing (a lot of Musk’s contributions went to a Donor Advised Fund, which is a separate legal entity, so the DAF may be the harmed party, not Musk personally). Don’t trade on courtroom drama.

Stocks / investments mentioned with explicit views

• Buy: Amazon, Twilio, Anthropic (private)
• Sell / avoid: Microsoft, Meta (least bad sell), Five9
• Hold / fairly priced: Alphabet (great but priced in)
• Speculative: Palantir (priced beyond perfection but justifiable if doubling continues), HubSpot (catalyst from agent platform), Cursor (depends on whether token-spend thesis holds)
• Cautious: Sierra at 105x revenue
• Honorable mentions: Vanta (re-accelerating to ~63% growth at $300M ARR), Founders Fund’s new $6B fund

Chapter Summaries

1. Mag 7 earnings — the most aggressive quarter in American capitalism. The five largest market-cap companies are growing 20-40% in sub-segments while letting CapEx grow 50-60%, consuming most of their free cash flow. They’re effectively servicing two private LLM companies that own the IP.

2. Google / Alphabet — the clear winner. Cloud backlog doubled to $462B; search hasn’t been disrupted; SaaStr’s own SEO is up 60% YoY. But token-growth metrics show Gemini under-performing Anthropic and OpenAI for coding, the canary use case.

3. Microsoft and Meta as the warnings. Strip out AI growth at Microsoft and the company is flat. Meta got punished because CapEx jumped to $145B with no attribution, no spreadsheet model Wall Street can build, and no application-boom benefit.

4. The buy/sell call. Buy Amazon (Anthropic on AWS, application boom tailwind), sell Microsoft. Apple aside discussing memory-price inflation as a stealth driver.

5. Palantir blowout. RPO up 134%, rule of 40 at 145%. Why it’s uniquely positioned: only software vendor that can credibly sell $10M+ AI transformation deals to Fortune 500 CEOs whose top initiative is AI. Implementation expertise gap is the worst in living memory; consultants and AI-skilled implementers will be hired aggressively.

6. SaaS apocalypse status check. Atlassian, Twilio, Five9 beat. Two-prong test (monetize AI in base + add net new customers). Twilio passes both because every AI-native company runs on it. Atlassian only monetizes the base. HubSpot’s headless agent platform is the next test for legacy SaaS reacceleration.

7. Cursor at $44B and the token economics question. The key research problem: steady-state token spend as % of salary for AI-first orgs. SaaStr data point: AI VP roles run for $254/month; engineering token intensity is 2-15% in their sample. Tokens get ~10x cheaper every 18 months — be inefficient now to learn.

8. Anthropic raises $50B at ~$900B in 48 hours. Removes any compulsion to IPO in 2026. CapEx math: $3-4 of CapEx per $1 of revenue, so $1B of run rate forecasts ~$30B CapEx commitments. There is no such thing as too much balance sheet cash. Both Anthropic and OpenAI IPOs may slip to 2027.

9. Sierra raises $950M at $15.8B (105x revenue). Concerns about tam expansion being flattered by labor-replacement narrative and competitive compression at the agent layer. Counter-evidence to the “LLMs eat everything” thesis but priced for a $100B outcome.

10. Musk vs Altman trial week one. Distillation admission, $30B Brockman stake, but the case will turn on statute of limitations and DAF-standing technicalities, not the merits or theatrics.

11. Coinbase layoffs and the management thesis. Brian Armstrong: every employee must be an individual contributor; managers of managers are out. Career rule: a CMO must be able to run their own campaigns via agents. Develop AI skills or be displaced. Work-from-home Fridays dismissed as a tell for under-investment in winning.