SpaceX Locks in $135 Offering Price — $75B IPO Draws a Mere 2x Oversubscription

Source: 20VC with Harry Stebbings | Published: 2026-06-11T14:00:06Z

Elon Musk bypassed the traditional price discovery process to set the price directly. With only 2x oversubscription, analysts peg the odds of a first-day price break well above 10% — but long-term, every successful launch should push shares higher.

SpaceX's Pricing Gamble

In a traditional IPO, banks work to hit 8–10x oversubscription, carefully selecting a price that manufactures a first-day "pop." Even then, roughly 10% of IPOs break below their offering price. SpaceX's subscription multiple currently sits at just 2x — barely enough to close a normal IPO, let alone a $75 billion monster raise.

Rory O'Driscoll's analysis is blunt: a fixed price, insufficient demand data, a massive capital raise, and thin coverage — stack those together and the odds of a first-day break are "significantly above 10%." Jason Calacanis is even more direct: this IPO probably won't surge in the short term, but every successful satellite launch, every new commercial contract, will push the stock slowly upward.

Rory raised a more fundamental point: at 70x forward revenue, there is no historical precedent. He believes a valuation correction within the next 12 months is likely — "this incredible company might be worth one trillion instead of 1.7 trillion, but that's still a massive win."

The Real Intent Behind OpenAI's Confidential S-1

OpenAI announced a confidential IPO filing around the same time, while emphasizing it had no commitment to a timeline. Rory sees this as pure expectations management: if you commit to a date and then delay, you invite a wave of negative coverage; leave yourself flexibility upfront, and a postponement won't read as something going wrong.

The more interesting angle is the domino effect: if SpaceX's IPO underwhelms, the deepest wound may be to OpenAI. OpenAI has been extremely aggressive on both valuation and fundraising scale, and any cooling in market sentiment could force it to ratchet down expectations. Rory's read is that everyone is "rushing through while the window's still open," and the real directive from legal and finance teams is almost certainly to get it done fast — not the "no rush" they're telling the press.

The Dawn of Always-On AI

OpenAI released Dreaming V3, the biggest memory architecture upgrade since the product launched. Sam Altman has been pushing toward making AI "persistently online." Jason believes that within two years, today's model of opening a browser to use AI will feel as antiquated as the desktop era.

Rory added an economic angle that's easy to miss: memory isn't just about better experience — it reduces token costs. If the model remembers who you are and what you're working on, it doesn't need the full context fed in every time. The "harness" layer built around foundation models — memory, context management — is becoming a competitive dimension in its own right.

Apple Dropping Its Own Model Might Be the Right Call

Apple announced deep Siri integration with Google Gemini, paying Google roughly $1 billion. Many read this as "giving up on AI," but Rory sees it the opposite way: Apple's core advantage is controlling the end device and user context — calendar, contacts, usage patterns — data that enables a consumer experience no one else can replicate. Not having its own foundation model isn't ideal, but Apple's business is selling phones, not models. As long as the experience is good enough that people keep buying iPhones, spending $1 billion on a model is a bargain.

Ben Thompson wrote a line that's been widely quoted: consumers don't want to "work." Outside of professional contexts, users won't voluntarily tackle complex research or productivity tasks — they want ease and entertainment. That's bad news for OpenAI's consumer play: consumer competition is about experience, not capability, and Apple has a natural moat on experience.

What a 146-Person Company Hitting $500 Million Means

Lovable reached $500 million in ARR with 146 employees. Cursor has hit $4 billion and is targeting $6 billion by year-end. Jason said something harsh but honest: "I have nothing but contempt for startups that need to bloat — what's your excuse?"

In his view, every company does only two things: build stuff and sell stuff. If you sell through PLG, you only need people who build. The moment someone tells the board "I need 50 more headcount and another $20 million in budget," that person should be shown the door.

Rory offered an important caveat: these companies stay lean partly because 50–70% of their revenue goes to Anthropic or OpenAI API costs — small teams aren't a choice, they're a constraint. Salesforce's revenue per employee is about $300K, with token costs at maybe 1%; Lovable's revenue per employee is $3.4 million, but token costs might be 70%. They're not more efficient versions of the same species — they're an entirely different commercial organism.

Jason's core argument: if you've invented a technology designed to make humans more efficient, and that technology generates trillions in revenue, then by definition you must see trillions in efficiency gains — and efficiency gains ultimately mean fewer people doing the same work. If the companies selling AI can't use AI to become efficient themselves, what hope does anyone else have?

Elon Musk May Have Made the Acquisition of the Year

Looking back at Musk's acquisition of Cursor at roughly 10x year-end target revenue, the deal logic keeps getting clearer. Over the past 24 months, Musk built the Colossus and Colossus 2 data centers from scratch. His own model efforts failed, but because he had the nerve to pour $20–30 billion into infrastructure first, he landed right at the moment everyone needed compute. Now Anthropic and Google are jointly paying him roughly $2 billion a month in compute leases — $24 billion annualized. Cursor fills the remaining server capacity.

Rory's summary is precise: he's essentially a lowest-cost CoreWeave, layered with a high-growth software asset. On January 1st you could have said "he spent all that money on data centers with no foundation model — he's done." By June 9th he has a $24 billion outsourcing business and a software business charging toward $6 billion.

RAMP and Revolut: Stress-Testing the Growth Premium

RAMP raised $750 million at a $44 billion valuation, with ARR past $1 billion and positive free cash flow. Revolut is valued at $115 billion with $4.5 billion in revenue and $1.5 billion in operating profit, currently running a $750 million secondary sale. Both trade at 30–40x revenue, while traditional banks sit at just 12x earnings.

The logic is always the same: as long as growth holds, high valuations are defensible; the moment growth normalizes, valuations get repriced. Revolut exists for a fundamental reason: European banks are terrible — cross-border fees, FX spreads, sluggish service — and Revolut sliced right through all of it. Similarly, Nubank scaled because Brazilian banks are inefficient, and Chime is only worth $5 billion because American banks are relatively competent. A fintech's ceiling is fundamentally determined by how bad the incumbents are.

Bending Spoons: The Vista Equity of Consumer Internet

An Italian company acquired Evernote, Vimeo, WeTransfer, AOL, and Eventbrite, consolidated them into a $1.3 billion revenue portfolio, and is preparing to list in the U.S. at a $20 billion valuation. The playbook is blunt: acquire, slash headcount aggressively, kill every low-ROI marketing line, then raise prices 80%. Evernote's average annual fee went from $75 to $250.

Jason's reaction was equal parts admiration and disbelief: "If Bending Spoons from Italy can make Vimeo and Evernote grow again, what exactly were the previous management teams doing?"

Rory broke down the financials: the blended growth rate looks great at 70–80%, but it's almost entirely acquisition-driven, not organic. Individual asset growth comes mostly from price hikes — the remaining users are price-insensitive. He delivered a cutting but accurate line: "Anyone who hasn't churned off AOL by now is probably never going to." It's a phenomenal cash flow business, but whether it can support a 20x revenue valuation depends on whether they can keep finding reasonably priced acquisition targets.

Databricks Chooses to Stay Private

Databricks announced another round at a $165 billion valuation, up from $134 billion earlier this year. With Anthropic recently raising $30 billion and OpenAI raising $122 billion, Databricks can comfortably meet its capital needs in private markets — it's a software company, not burning cash training foundation models. And private markets are giving it a higher revenue multiple than Snowflake gets in public markets.

Rory sees a timing factor too: this year is too crowded — SpaceX and two major model companies are all in the queue. Next year might offer a cleaner window.

Microsoft's In-House Model: Late but Necessary

Microsoft released a homegrown AI model, but reviews were lukewarm — it doesn't even support web search. Rory sees this as marking the end of an era: outsourcing your AI strategy to OpenAI is no longer viable. The Microsoft-OpenAI relationship is somewhere "between an open relationship and already divorced."

The key question is whether Microsoft can replicate the Azure playbook — Azure was never better than AWS, but for most enterprise customers it was "good enough." Can Microsoft grind its model to "good enough" in two to three years? Rory notes that Microsoft failed to catch up in mobile and search but did catch up in cloud. Which pattern this follows is impossible to predict right now.

Where Does the Money Come From? It Comes from Courage

With SpaceX, OpenAI, Revolut, and RAMP all pulling capital from the market simultaneously, a natural question arises: where is all this money coming from?

Rory's answer is blunt: "When there's money, it's not because there's more of it — it's because people got brave. When there's no money, it's not because it disappeared — it's because people got scared." Everyone is in risk-on mode right now. Last Friday saw a mild pullback — Broadcom's guidance came in slightly below expectations, and the semiconductor index was up 100% year-to-date — but two days later everyone had already moved on.

The question is how long people stay brave. Rory says if he knew the answer, he wouldn't be sitting here talking — he'd be trading QQQ. Minsky's framework applies: everyone stays aggressive until someone gets burned.