Elon’s Empire: SpaceX, Tesla, Neuralink After the Storm & Anduril’s $2.6BN Power Move

Summary

• Circle’s IPO is the strongest since 2020 — and a cautionary tale about pricing. Priced at $31 and trading at $80 two days later, with over half the deal being secondary sellers (~20M shares), roughly “a billion dollars that went to the buyers not the sellers.” Jason’s warning: Circle and CoreWeave are “at the edge of meme stocks” — good companies with meme value layered on — and “at the start of a run of IPOs you have to be careful on the meme stocks and toward the end of the bull run you price the hell out of them.”
• The IPO process itself is indicted and acquitted in one line: “a wildly flawed process to which we can find none better.” Bankers hold the information asymmetry (“at 31 bucks you get Fidelity and T. Rowe, but at 35 you only get a bunch of hedge funds”), oversubscription figures are bogus game theory (10x isn’t enough; you want ~30x to pop hard), and the alternatives are mixed — SPACs failed, direct listings suit companies that don’t need primary capital, and Dutch auctions worked for Google. The median $2B company still can’t afford to innovate on process.
• The window is always open at the top — “there’s always room at the top” — so Databricks and Stripe staying private is choice, not constraint. Figma’s confidential filing is read as a revenge win after the collapsed $20B Adobe deal and secondary at 10: “This time I’m getting my freaking win.” CoreWeave, which cut its range at pricing, is up ~2.5x and used the pop to raise fresh debt that erased its existential repayment risk — proof “it’s so random, it’s so outside your control.”
• The unicorn liquidity math is brutal: ~1,500 unicorns, Rich Wong of Accel says 20% will simply fail, Rory’s distribution puts only ~20% ever getting public and 50% stuck in a merged/PE middle — and only the top 10–20% can even run tender offers. The 4-year path to a bell-ringing became 12 years, so liquidity becomes a recruiting weapon: “How does your tender offer process work? That might be my first question in the interview.”
• Founders Fund put $1B into Anduril’s $2.5B Series G — its largest check ever, and its prior largest was also Anduril. The panel reads it as a 20-year national-security thesis (they co-founded Palantir in 2003–04) plus stage-appropriate math: if Anduril goes from $13B toward $100B and they keep 20%, “the partners clear a billion dollars personally off a one-day decision,” with Lockheed’s $150B cap as the benchmark.
• “Stuff money into your winners” sounds smart and mostly doesn’t survive the math: in a 20-deal fund only ~4 are stuffable, and by the C/D round even those price to 3–4x. Harry’s example: bought DocuSign at $1–2, passed on the round at $19, stock at $80 — but the honest lesson is TAM discipline, not “always double down,” because Chime looked stuffable and the price turned out to be 12–15, not 25.
• The SaaS slowdown is arithmetic, not cyclical: at 40% of workloads moved to the cloud after 20 years of ~30% growth, three more years takes you to ~80% — maturity was inevitable — while AI eats the budget (Cursor did ~$500M in revenue that “sucked up” dollars from Okta and Salesforce). The open question is whether AI is additive TAM via labor replacement; Jason’s contact-center data point cuts against the hype: customers replaced 40–50% of contact-center humans and ACV rose only 50%.
• On Elon post-firestorm: SpaceX is hate-proof, Tesla isn’t. “The definition of a great business is when your customers can hate you and still do business with you” — the government has no other rockets, and SpaceX’s former largest customer is no longer its largest as Starlink has become such a big business — whereas Tesla’s EV subsidies and consumer brand are exposed. The government episode was “a management failure of massive proportions… everyone should play the position where they can score and win,” but the panel’s base case is that in a year “maybe no one cares.”

Deep dive

Circle IPO Pricing

• The whiplash Rory flags first: “we went from the window was shut four weeks ago to the window was open to oh my god we underpriced this thing” — with no intervening period of gratitude. Circle filed, raised the range, and still opened at roughly 2.5x, from ~$31 to $80 within two days.
• The underpricing sting is unusually acute because over half the IPO was secondary: ~20M shares sold at $31 that traded $50–60 higher days later — “that’s a billion dollars that went to the buyers not the sellers.” Great outcome, but the sellers “feel amazing about the outcome but oh my god that’s a lot of money on the table.”
• The recent cohort is broadly healthy — all recent IPOs except SailPoint up, averaging +76.8% (with the caveat that “averages are confusing and misleading”). Mountain Hinge Health and likely eToro are the un-hyped names trading nicely; Circle and CoreWeave are the outliers.
• Jason’s classification, qualified by Rory: these are meme stocks — but good ones. “Something like GameStop was just a meme stock. Both CoreWeave and Circle are exceptionally good companies in big-ass meaningful industries” — Circle carrying heavy interest-rate sensitivity, CoreWeave as the listed AI proxy — with meme value on top that makes pricing nearly impossible.

IPO Process Tradeoffs

• Some pop is structurally necessary — no prior price history means buyers “got to get paid something for the volatility you’re incurring.” The judgment call is retail vs institutional demand, and the bankers’ claim that cheap anchor allocations create the retail demand strikes one camp as plausible and the other as “a total violation of the efficient market hypothesis.”
• The banker’s real edge is repetition: “you do this once in your life… and a banker’s doing it every week.” They’ll tell you “at 31 bucks you get Fidelity and T. Rowe, but at 35 you only get a bunch of hedge funds and they’re going to flip it” — so you take 31, it pops to 70, Fidelity flips it anyway, “and you feel like you’re a sucker.”
• Oversubscription numbers are theater: 5x isn’t enough, 10x (Chime’s reported book) isn’t really enough — “you really want to be like 30x oversubscribed to pop hard” — because buyers pad orders expecting cutbacks, so “the demand is entirely theoretical.” And the final allocation is decided not by the relationship bankers but by “some person from Equity Capital Markets that crawls out of the hole in New York” declaring who will and won’t flip.
• The alternatives have all been tried: SPACs “have been a disaster,” direct listings only work if you’re amazing and need no primary capital, and Dutch auctions seemed to work for Google. The median $2B company “can’t afford to get it wrong — this is a one-time debut,” which is exactly why it’s discouraged from innovating on process.

Figma and CoreWeave

• Rory’s core frame: “the window was and always has been open for Databricks and Stripe — they just don’t want to go through the window… there’s always room at the top.” Windows open and shut for the $2B market cap, never for the $50B one.
• On Figma — “the low end of really amazing like Stripe, but the high end of more than amazing” — the psychology matters: after deciding to sell to Adobe for $20B, losing the deal, and doing a secondary at 10, “if you nearly had the win and it was taken away right at the last minute, you’re like, ‘This time I’m getting my freaking win. I am ringing the bell.’”
• Jason’s momentum caveat: when everything trades up, “every meeting starts to be about should we go public now… when people are on the edge, they just kind of go forward in this environment” — even though this data shows the window was arguably open last year too.
• CoreWeave is the humility lesson: it had to cut its filing range four months before trading 2x up, then used the strength to raise additional debt ($2 million as spoken) and kill the ticking debt-repayment risk: “a triple hat-trick… from teetering on the edge to set for the better part of a decade.” Rory’s takeaway: “pretending you have this a priori knowledge of when the window is going to open — you just have to internalize it’s so random,” do the preparation, and accept the timing. Jason’s confession: “You could have bought all the CoreWeave your little heart desired… I didn’t buy any. What kind of idiot am I? Where we rank on the omniscience factor is probably a two or three out of 10.”

Wise Lists in America

• Wise following Deliveroo off the London market isn’t a ding on the Brits, per Rory (an Irishman who’d enjoy one): “it’s just about how freaking awesome the US capitalist system is. The United States has 4% of the world’s people, roughly 23% of the world’s GDP, and 67% of the world’s market cap. We won.”
• The market agreed instantly — Wise popped 8% on the announcement, “basically like saying you can make 8% free money just by listing in the US.” For any international business the logic is one-way: “why wouldn’t you go where 70% of the cap is and just join the team?”
• Jason’s underrated addendum: lay folks overestimate tech-stock liquidity. At the $2–5B level “there’s no analyst coverage, the institutional buyers are not there in the single-digit billions” — so if you’re at the edge of liquidity, being anywhere but the deepest market is untenable.

Unicorn Liquidity Squeeze

• Should thin-liquidity $2–5B companies even be public? Rory’s answer: yes, eventually — “even the thin liquidity of a public market is better than the liquidity in the private markets.” His employee test: would you rather have equity accessible “once or twice a year with the approval of management” or “freely tradable every day of the year? It’s pretty obvious.”
• The distribution math: ~1,500 unicorns per Crunchbase; Rich Wong of Accel says 20% will just fail — they won’t limp along. Rory’s rough cut: ~20% (plus or minus) good enough to eventually get public, ~50% “meaningful enough to be valuable but not so meaningful they’ll get public” — merged, combined, PE’d — and a low-end tail of 20–30%. “Some companies once worth a billion dollars can go to zero, easily done, especially if you have debt and a high cost structure.”
• The employee-comp squeeze is what forces the issue: “it used to be work really crazy hours for four years, we’ll go public, you’ll ring the bell. Now it’s sign up for four years, that turns into 12 years, and at the end we’re still trying to put a tender offer together.” Only the top 10–20% can even run tenders — so for most unicorns equity comp “is notional and not accessible.”
• Jason’s talent-market conclusion: “if I was a top-tier engineer, I might only join someone with a perfected tender offer program. That might be my first question in the interview.” Rory generalizes it: monetizable stock carries a liquidity premium in recruiting — one more force pushing scaled companies toward just going public.

Anduril Draws Concentrated Capital

• Founders Fund put $1B into Anduril’s Series G — its largest check ever, and its prior largest was also Anduril. Jason’s route to conviction ran through a jab: Sam (likely Lessin) said Anduril was the only really important company he could think of besides OpenAI — “just cuz you think someone’s being a jerk doesn’t mean they’re not right… my learning is it’s probably an even better company than I realize.”
• Rory refuses to call it surprising: “they told you to do this back when they started the firm” — concentrated bets, plus a 20-year national-security thesis (they co-founded Palantir in 2003–04, “not Johnny-come-latelies rushing in to catch up”), plus founders at “the Maslow hierarchy stage where minimizing risk for your investors is not your one, two, three, or four priority.” You cite your Lord of the Rings and push it across the table.
• The payoff math makes it rational, not romantic: “if that goes from 13 to 100 and they keep 20, the partners clear a billion dollars personally off a one-day decision.” Harry’s comp: Lockheed Martin at $150B market cap makes a 3.5x visible if Anduril is the next-generation prime. And a $1B check “is much more ambitious than throwing 50 million into the last round of Anthropic.”

Sizing Follow-On Bets

• Rory runs a 10% concentration limit, ~20 deals averaging 5%, and answers with self-knowledge: “I’m probably more risk-averse at the margin… it’s the ‘know what’s in the box’ thing.” His verdict on the hypothetical: “if everyone was set free to do whatever they want, for the median firm it would be value destructive.” Though he concedes he’s wrestling with whether “a little more standard deviation in your bet sizing” fits this market — “up for grabs, thinking about it.”
• Harry’s answer: every dollar into every winner up to a billion — seed, then every round A through E, maximizing ownership and capital, if the later-check risk could be offset into SPVs. He sees exactly this behavior on his hottest cap tables: funds with “for all intents and purposes unlimited capital” stuffing every round.
• Rory’s dismantling is the keeper: in 20 deals, ~30% fail (don’t put a dime in), ~50% are 1–5x (the next round is a 2.5x or worse), so only ~4 deals are stuffable — and what was 10x from the A and B prices is “by definition three and 4x from the C and D prices.” The forced adaptations: right-size the opportunity fund (nobody does), drift into general growth investing, or triple the A/B at-bat count to find the 5-billion-plus outcome the math now requires.
• The scar tissue: “If I knew DocuSign was going to compound to where it did, I would have done the round at 19 bucks too. We did the round at a buck and two bucks… it’s now at 80 and I passed at 19.” But the intellectually honest question is “what other deals did I have that looked equally promising” — Chime looked stuffable and the price turned out to be 12–15, not 25, “and stuffing didn’t work.” The venture lesson: “be visionary but relentlessly honest about TAM,” and remember 2021 was “a fake signal — even your okay ones got highly valued.”

Entry Growth Has Limited Predictive Power

• Harry’s confession sets it up: 18 months into his first fund he predicted his top five returners — “you had your Hoppins, your BeReals, your Clubhouses” — and none of the five outperformed; the real winners were always in the middle bucket. Roger [likely Ehrenberg] at IIA told him he’d seen exactly the same.
• Rory says that’s true at seed but shifts at A/B: his mental model is a ~20% going-in probability of a 5x-plus outcome, but “if after two years the company has done what we said it would do — it’s ramped — the probability goes up to around 60 or 70%.” Paying post-product-market-fit means the data comes back quicker.
• His firm’s quartile study is the counterintuitive core: ranking every deal by growth rate versus peers at time of investment, there’s “very little correlation between great outcomes and being top versus second quartile.” Bill.com was “always second quartile” and compounded into “a force of nature”; HubSpot was second quartile for a couple of quarters. “It turns out that would be a very bad rule… you want to be top-half growth, but then capital efficiency, time to market, entrepreneur, persistence” decide. His friend’s line: “If it was factory work, they’d pay you factory wages.”
• The live disagreement: Harry says a vertical SaaS company going 1 to 7 million over three years is “not attractive — no one’s going to touch that”; Rory reframes (“it’s not you, it’s me” — great business for the founder, wrong for the venture model, “you never want to diss the entrepreneur”); and Jason breaks ranks: “I might do that deal” — if the founders are incredible, the true TAM is large, and the price gives him time, citing his Pipedrive bet at $16M pre with similar metrics. He backs it with a cited report finding that velocity to 100 hasn’t fully correlated with success at scale.

SaaS Spending Slows

• The cited H1'25 analysis shows SaaS spend growth declining again after the presumed mid-2024 bottom and reacceleration. Jason’s read: “AI is sucking up budget — here’s a real example of it happening” — Cursor did almost half a billion dollars of revenue in the period, dollars “that would have gone to Okta and Salesforce.” “Even if the 2024 days are behind us, it doesn’t appear to be any easier.”
• Rory’s structural claim is that this was inevitable: the “we’re only 40% of workloads moved to the cloud” narrative was “horrific news — you idiot — because you’ve compounded from 1% to 40% share in 20 years at 30% growth. Three more years of 30% and you go from 40 to about 80.” These are mature, served markets; what follows is “bundling, consolidation, grinding out the weak.”
• His extreme example: “Who the hell do you think is left in 2023 who doesn’t have a freaking Zoom account? If you didn’t buy one in ‘21, you’ve hit complete TAM.” Same for CRM: “you’ve had 20 years to buy the damn thing — if you haven’t bought it now, you are a trailing-edge adopter.”
• The load-bearing hope, stated as a condition: if AI merely replaces CRM with AI-CRM, “it really is a knife fight for limited resources”; if it takes over labor dollars, it’s additive TAM. Jason’s check on the groupthink — notable from someone who’s “been more apocalyptic”: “I want to believe, but I don’t think it’s a slam dunk that the B2B TAM goes 5x because we attach to human budgets. We haven’t proven that.” Practically, “there’s no brownie points for taking on a SaaS conversion in 2025; there’s a lot of brownie points for doing something in AI.”

Contact Center TAM

• Jason’s portfolio data is the sharpest evidence in the episode: at what is likely Gorgias, which dominates Shopify contact centers, “their average customer’s replaced 40 to 50% of their humans with AI — and their ACV is only up 50%.” Trading a $50–60K fully-burdened human for a low-hundreds-per-year software bill: “I don’t know that there’s enough TAM appreciation” for the compounding math to be exciting.
• Rory’s counter-framework, imported from robotics: labor gets replaced at a two-for-one arbitrage — “you’re spending 100 grand on labor; if we can do it for 50, they’ll do the deal” — and the vendor keeps half the savings. Contact-center software is $10–15B a year, the labor pool at least $150B; on the 2:1 rule that’s arm-waving toward $75B. He won’t claim AI eats it all — “there’s going to be humans on phones for the foreseeable future” — but sees “at least a 2x, 3x TAM expansion.”
• The pricing bifurcation both land on: enterprise automation pays because the quantum is big (2,000 agents at $50K each), while at true SMB, S-tier AI gets bundled free with limited upsell — “Agentforce will try to charge massive amounts of money… and we may look at SMB products and say, wow, look what I get for free.”
• Jason’s deeper worry about venture price points: “when the cursor for sales comes out for real, it might not be a traditional sales process charging 50 grand — it might be 30 bucks a month… when the underlying COGS approaches zero, I’m not sure the price points we hope for are sustainable.” Rory’s sequencing rejoinder: this is SaaS in 1999 — obviously the future, still took 20 years, and value-ordered (CRM first, accounting last). “Pick the spots where it works now and avoid the spots that take five more years.”

Elon Companies Face Uneven Exposure

• The panel’s cold-blooded read on the government episode: “most investors would have preferred to skip the whole thing” — staying supportive from the sidelines (the Peter [likely Thiel] play: halo without the line of fire) would have captured the December-January “amazing for Tesla” trade without the pain. Tesla is the most impacted: EV purchase subsidies and emissions-credit sales to GM can be withdrawn by Congress “with no obvious political cost,” and the consumer brand is alienated, worst in Europe. SpaceX is different in kind: “the definition of a great business is when your customers can hate you and still do business with you” — the government has no other rockets, and SpaceX’s former largest customer is no longer its largest as Starlink has become such a big business.
• Jason’s structural verdict: “it’s a management failure of massive proportions to hire the guy who did that” — Tesla, SpaceX, OpenAI founding, Neuralink — “for something political. Everyone should play the position where they can score and win.” Jason’s timeline call: “this may sound crazy, but in a year we will not have forgotten about it, but maybe no one cares. I don’t even know if Trump cares anymore.” Harry’s open meta-question — “at least 50% of the tech billionaires on X are unhappy; can they still innovate at that stage?” — gets Rory’s Gresham’s-law answer: social media “forces a persona… bad opinionated people drive out good, boring people,” so you can’t read happiness off X. (Jason, on Chamath: privately “very humble and kind.”)
• Kalshi quickfire, Sundar leaves Google this year: Jason says no; Harry offers “some version of yes,” but the most likely outcome is no. Jason’s analogy from his Adobe VP days — the board “would grab Shantanu [likely Narayen] by the jacket” before letting him leave — and Jason’s sharper point: Google has gone from “the world is ending” to shipping good models, but the innovator’s dilemma stands (“I start on ChatGPT to do my research, not Google anymore”) and swapping CEOs doesn’t solve it. The wrinkle worth watching: Sergey is “back back” and publicly energized.
• NYT vs OpenAI: Jason gives 80% to win-or-settle and argues the suing strategy will be validated versus the outlets that cut $20–30M deals. His game-theory nuance from a reference call: LLMs need modern news, but “do I need the third or fourth marginal news source? Maybe not” — which reprices media content and favors unique catalogs. Jason wants his cut (“ChatGPT scrapes a lot of our content… I get three or four grand a month from Twitter — I want $500,000 a year from likely SaaStr”) and notes that if it doesn’t settle, the Supreme Court decides what fair use means in the AI age — “probably a 48% gamble.” On Linda Yaccarino: Jason, judging only the public persona, calls her “the VP that maybe I’ll upgrade this year”; Harry counsels the returning Elon, “do you just want the heartache, dude? Just let it run. Don’t be a hero.”