Andrej Karpathy Joins Anthropic | SpaceX Files S1: How Does it Trade | Cerebras Smashes Day 1

Key points

• **Anthropic at $900B for $30B raise — tripled from $380B in February.** Rory's framing: 'If ARR multiples are still fair, 18x June revenue for a company growing 10x/yr that will clearly IPO is the best deal in the venture universe.' Why Dario does it at $900B not $1.7T: 'Sam pushes to the max because he needs infinite-capital narrative. Dario wants a fair deal in a week, not drama. The last two anthropic rounds intentionally traded at a discount to get done low-drama in days.' **Direct continuation of [Eric Ries's LTBT thesis from Issue 05](/issues/2026-05-17)** — the no-drama fundraising approach is the operating-side expression of the mission-protection governance choice.
• **The Salesforce $300M token math is the most important data point this week.** Rory's calculation: Salesforce has 20K developers out of 83K heads. $300M = **$15K/dev/year = $1.2-1.3K/dev/month** — matches a 20VC survey of 40 portfolio companies. **At ~4% of fully-loaded engineer cost ($500K), this is 'just table stakes.'** **The deeper structural insight**: 'For OpenAI + Anthropic to hit $1T of ARR, you need 5-7% of every knowledge worker salary OR 20% of every engineering salary going to tokens. Salesforce is only 1/4 of the way there. Either token consumption 4x'es from here within 2 years, or the model-company valuations are wrong.' **Direct cross-reference to [Gavin Baker's '$11B ARR in one month' framing from Issue 06 ILTB](/issues/2026-05-24)** — same math, different vantage.
• **Andrej Karpathy joins Anthropic** to head pre-training focused on recursive self-improvement. Significance: 'Karpathy is among the best researchers and engineers in the field, plus he brings massive name-brand pull for talent recruiting.' Jason's prediction: 'Anthropic will have $50B funding round at $2-3T valuation within 12 months.' Companion to [Eric Ries's mission-guardian-talent-flock thesis from Issue 05](/issues/2026-05-17) — Karpathy is the highest-profile mid-career hire so far attributable to the structural moat.
• **SpaceX S1 filed — June 12 IPO, $1.75T valuation, $75B raise, largest IPO ever.** Rory's hot take: 'The S1 will tell us everything about SpaceX as it existed in December — without xAI, without the Cursor deal, without the Anthropic deal. That's the funniest S1 ever — you're reading a story about a company that is pretty much 30-50% revenue NOT in the financials, and almost ALL the new losses NOT in the financials except for half a quarter.' **The institutional risk**: 'No forward projections allowed in an S1. Bankers will have to tell the story via roadshow.' Confirms [the Issue 05 Cerebras-IPO-window-opens prediction](/issues/2026-05-17) for >Figma-size companies.
• **The 'will SpaceX pop GameStop-style' debate.** Jason: 'Going to $5T at end of month. Robin Hood retail will go bananas.' Harry: '$3T at end of month.' Rory: under $3T — 'GameStop went 20x from a tiny float. SpaceX starts at $1.7T — Nvidia is only $5.3T. Hard to 10x from here. Also you may see a Facebook-IPO effect — flat for 1-2 days then drop 40-50%.' **Worth tracking against [Brad's June IPO date confirmation from Issue 04 Milken hit](/issues/2026-05-10)** — date held to within a week. **Price action through June 30 will be the key data point for the rest of the AI-IPO calendar.**
• **Cerebras opens the IPO window for >Figma companies, not <Figma. 68% pop on day one.** Rory: 'Better-than-Figma is the new bar. Pricing $185 from initial $115-120 range = obvious winner. Coreweave + cursor + Anthropic + OpenAI access = end-of-one positioning.' **The fickleness data point**: 'Two years ago they couldn't get the deal done [CFIUS-blocked]. People who said it was a positive tell for SpaceX last week were correct. People are very much risk-on for things that look like obvious upside.' Companion to [Andrew Feldman's 24-day OpenAI deal disclosure from this week's no-priors](/issues/2026-05-24) — the same shift in 'art of the possible' is showing up in IPO timing.
• **Sam Altman's $2M OpenAI tokens-for-equity offer to every YC startup in the current batch.** Rory: 'It's smart — Anthropic stole a march, you've got to win back hearts and minds. Yuri Milner / DST did the equivalent with early YC batches.' **Sub-text: OpenAI has surplus tokens, Anthropic doesn't.** $300M of effective free tokens per batch × 4 batches/yr = $1.2B/yr of foregone revenue at 18x = $20-30B implied valuation hit. **Implication**: 'Anthropic is so capacity-constrained that it cannot afford to make this offer. OpenAI can.' Direct cross-reference to [Gavin's '70% fewer Opus tokens per question' compute-rationing observation on ILTB this week](/issues/2026-05-24).
• **OpenAI may file S1 as soon as Friday — breaking news during the recording.** *'When Sarah Friar said we need 12 more months to start the process, what she really meant was we're going today. The last trains are leaving for Money Station.'* If true: **OpenAI + SpaceX + Anthropic + Cerebras + Cursor + xAI** = the AI-IPO calendar is now extremely front-loaded for H2 2026, completely re-rating [Issue 05's 'mid-cap IPO window opens' prediction](/issues/2026-05-17). **This is the most concentrated cohort of AI IPOs in history — and the supply/demand will dominate the public-markets AI narrative through year-end.**
• **The AI PR crisis is the central political risk.** *'No one in America other than us here in California likes the AI trend. We have people who are brilliant scientists who politically are utter morons. And the people who are utter morons at AI but brilliant at politics are going to have us for lunch.'* Eric Schmidt booed at graduation. Trump's planned AI executive order scrubbed an hour before. **Standard Chartered CEO**: 'No job losses, just job reductions in favour of the machines.' **The 'reflation hiring' charter**: Jason's argument that Big Tech needs to commit to mass re-hiring after the AI restructure to avoid social unrest. *'We'll get fit. We'll replace our workflows. We'll have AIs. Then we'll have a social obligation. Eric Schmidt can't just go to graduation and say F-you. We'll have to reflate and hire thousands per tech leader.'*
• **'8,000 ex-Facebook employees vote on the wealth tax next week.'** Rory's most-quotable framing of the political-physics problem: 'How do you think those 8,000 ex-Facebook employees will vote on the wealth tax next week? When you've been laid off by email at 4 AM because the CEO would prefer to buy $100M of machines than $100M of people — that politics is very different. You're not thinking as much. You're pretty pissed off.' **Cross-reference to [Eric Ries's $200M Pentagon turn-down framing from Issue 05](/issues/2026-05-17)** — the Anthropic-LTBT structural-mission protection IS the inverse defense against this political force.
• **The 'someone is going to wake up and pull off the AI pivot' bet from Rory.** *'One of the truly down-beaten public software companies in the next year will become up-beaten. The startups are moving faster, the models will move, but someone is going to put their best 50 people in a room and just say: build a better version of [insert hot AI app]. HubSpot has 300,000 customers waiting to buy an AISDR. If Breeze became actually good, they'd sell 150,000 instantly.'* **The implicit watchlist**: HubSpot, Drew Houston's Dropbox, Atlassian (Rovio = receipts), Figma 2.0, Twilio. **Companion to [Issue 05's data-infrastructure-layer-as-the-SaaS-exception thesis](/issues/2026-05-17)** — the founder-led + installed-base combo is the second exception.

Notable quotes

Themes

• Anthropic at $900B for $30B = 18x June revenue, 'best deal in venture universe'
• Salesforce $300M token math: $1T ARR requires 20% of engineering salary
• SpaceX S1 filed June 12 — 'funniest S1 ever' missing 30-50% of revenue + most losses
• Sam Altman $2M tokens-for-equity to YC = OpenAI has surplus, Anthropic doesn't
• AI PR crisis = political-physics risk; 'reflation hiring' social charter as the defence