# Stripe Buys OpenRouter for $7B, Anthropic Turns Its First Profit, and Silver Lake Circles Workday at $43B: 20VC x SaaStr > Source: > Published: 2026-08-23 12:42:24+00:00 *Plus the math that has to be true for AI to reach $600B in revenue: $100K of tokens per engineer, and 30% fewer engineers.* A heavy week on 20VC x SaaStr with Harry Stebbings, Rory O’Driscoll and Jason Lemkin: - SpaceX closed its $60B all-stock takeover of Cursor - Stripe bought OpenRouter for around $7B, four months after OpenRouter was valued at $1.3B - Anthropic turned its first profit - Silver Lake circled Workday at $43B - And Higgsfield and Lovable raised at $5.5B and $13.3B on roughly the same revenue. Here’s what we got into: ## #1. By the time the Cursor deal closed, $60B wasn’t an expensive price. **Rory’s math:** At roughly 15 times current revenue, or under 10 times the $6B the company is tracking toward by year end, this is a normal multiple for the fastest-growing asset in the largest AI category. SpaceX itself trades around 40 times revenue. Buying at 15 with paper that trades at 40 is accretive the day it closes. **Jason’s read:** The best AI deals can often look absurdly expensive at signing and by closing, look almost conservative at closing. Don’t look at ARR multiples, look at the multiples of forward revenue and forward growth.That’s the bet. ## #2. Cursor was declared almost dead on this show not 12 months ago. Going multimodel is what saved it. **Jason’s read:** We’ve been doing this show about 70-something weeks and there have probably been three different Cursors in that window. It rocketed to half a billion in revenue. Then Claude Code came out and it genuinely looked over. Nobody’s portfolio companies were using it. I’m not even sure Claude Code had launched when we started the show, which is insane to say out loud. Cursor went multimodel very early, and that one call turned into a $60B outcome. This was not linear progress. Many teams would have quit on that journey. Being agile isn’t enough at this rate of change. You have to be able to absorb a competitor shipping something better than you and keep executing your own plan anyway. ## #3. The gross margin criticism was correct. It just stopped being the deciding factor. Early Cursor was selling a dollar of tokens for well under a dollar. That was the standard VC mockery when we started the show. **Rory’s read:** The negatives you can cite along the way tend to be true. They were real. It’s easy to sound financially sophisticated and say everything eventually has to generate free cash flow, so a business without gross margins is worth nothing. That’s wrong when the market underneath is exploding, because the buyer looks past it. Coding is the biggest AI market there is, and a large share of Anthropic’s value is predicated on it. If you own the number two asset in that category and it’s compounding, margin structure is a second-order issue. ## #4. Elon bought a gross margin problem he was already positioned to convert into revenue. The cleanest line of the episode, and the actual logic of the deal: your gross margin problem is my revenue opportunity for my Colossus cluster. **Rory’s read: **The set of buyers for whom this works is tiny. You need enormous compute and no obvious business sitting on top of it. That describes two companies, Meta and SpaceX. One of them bought. Cursor’s worst line item was inference paid to someone else. Owned by a buyer who owns the compute, that cost becomes internal revenue. The same P&L looks completely different depending on who holds it. ## #5. How you take a company off a round it already has. Cursor was closing roughly $2B at a $50B valuation from Andreessen and others. Elon bid $60B, added a $10B breakup fee if the deal failed, and granted autonomy. That combination removed the objections. **Jason’s read:** Zuck did Instagram and WhatsApp in about an hour on the back of a napkin and paid high. This is the same move executed better. When a company is mid-round at a known price, you don’t run a process against it. You put a number on the table that makes the round irrelevant and you underwrite the seller’s downside if you can’t close. Very few people on earth have the authority and the currency to move that fast. ## #6. Why Meta (potentially) passed, and why the seller may not have wanted the Meta deal anyway. **Harry’s read:** Meta is building model capability with Alex Wang and company and is missing the enterprise piece. Cursor would have solved that. Three reasons it went the other way: **Deal certainty.** SpaceX filed and got antitrust clearance quickly. A seller would have far less confidence a Meta deal clears given the scrutiny Meta attracts. Certainty is a term, and sellers price it.**Necessity.** SpaceX built AI capability out of an existential need to use compute it already owned, and it now rents that compute to Anthropic and Google. Meta’s AI push is driven by interest rather than necessity, because a $100B+ advertising business already works.**Who you want to work for.** For a founder already running a $50B company, the calculus is who executes. Mass layoffs and pressure to produce a frontier model is a different offer than what Elon put in front of them. The broader founder advice on the show: in M&A, ignore brand reputation and the job description you’re promised, because the business will look nothing like it in 24 months. ## #7. Microsoft is arguably the real loser in the Cursor deal. **Rory’s read:** owning a developer platform is existential over the medium term for exactly one company, and it isn’t AWS and it isn’t Meta. Microsoft lost its connection to developers, the group that made it in the first place. GitHub is now a trailing-edge product. That’s a serious medium-term loss even though the company is operationally fine. **Also worth flagging, per Jason:** the popular VC theory that one big acquisition triggers an industry-wide land grab is mostly wrong. Buyers don’t panic-buy the second or third choice because a rival moved. They buy when they realize they have less time than they thought. ## #8. Stripe paid roughly $7B for OpenRouter, four months after a $1.3B round. OpenRouter, led by Alex Atallah, is the leading LLM routing layer. Stripe’s business has always been absorbing complexity in exchange for a share of the flow. Model selection and API integration is the same shape of problem. Stripe already benefits from AI growth indirectly, taking a cut every time somebody pays OpenAI or Anthropic with a card. **Rory’s read:** current revenue is close to irrelevant to the price in deals like this. The acquirer is buying future extractable value, and can usually generate cash from the asset faster than the standalone company could have. In a market moving this fast, the buy-versus-build analysis that assumes a five-year build window gets abandoned. A few billion in cash and stock buys you a live capability inside a week. This is what venture returns are made of, and it happened repeatedly in the early internet buildout. ## #9. The bear case on OpenRouter is that enterprises only want two or three models for most workflows. And they may want their router as part of the harness. **Jason’s read:** routing is genuinely useful for developer tools and chatbots, where availability matters more than perfectly consistent output. It’s much less useful for high-reasoning B2B workflows, where switching models introduces drift and forces constant re-testing and QA. Rippling picked a small set of models on price, performance and speed rather than managing a large menu, and that’s the more common enterprise posture. Databricks, Replit, Lovable and Vercel all do their own routing. The counterpoint: enterprises that don’t want to be captive to one frontier lab use a router as plan B, and having plan B is what keeps the lab honest on price. **Rory’s read:** it’s a niche product today. It could become 20% to 30% of Stripe’s revenue. It’s also entirely possible the product doesn’t exist as a product in five years, absorbed into Stripe infrastructure or reduced to a checkbox. Call it a one-in-three chance it becomes a real AI routing business, which is a fine bet at this size relative to Stripe’s market cap. ## #10. Stripe is running public-company M&A while private, and buying two different things. OpenRouter is a TAM expansion bet. The reported interest in PayPal is a consolidation bet: use scale, strip general and administrative expense, get consumer wallets. Same acquirer, opposite logic, and both funded partly with private stock as currency. ## #11. Anthropic’s first profit was arithmetic, not discipline. Anthropic turned its first profit on $11.5B of Q2 revenue. Gross margins went from negative in earlier years to roughly 30% last year, with an expectation of about 40% by year end. **Rory’s read:** when revenue moves from $4.5B a year to $10B+ in a quarter and gross margin goes from negative to 40, you can’t add expenses below the line fast enough to stop yourself from making money. Headcount and operating expense simply cannot scale that fast. Don’t expect a base case of continued profitability in an IPO filing, because compute costs keep rising, including the price increases on hardware bought from Elon. ## #12. Nobody is going to care about the accounting. They’re going to care about 2027 and 2028. Anthropic heads toward an IPO with off-balance-sheet compute commitments and enormous stock-based comp. **Rory’s read:** those numbers look ugly and everyone will look through them. The only thing that matters is the growth rate in projected 2027 and 2028 revenue. The compute commitments are investments that get justified by growth, and become a millstone only if growth slows. On SBC, the distinction is real: at Workday or Salesforce, RSUs are cash comp by another name, because remove them and you have to raise salaries. At a hyper-growth company, most of that SBC number is appreciation, not recurring cost. Someone hired on a $1M package in 2023 ended up with $51M four years later. That’s not a repeatable expense line. The leniency on margins, off-balance-sheet liabilities and SBC lasts exactly as long as the growth does. ## #13. The math behind $200B, and why $600B in revenue for Anthropic is a much harder claim. The bull case has Anthropic-class revenue reaching $200B by 2028 and $600B the year after. The show did the arithmetic. The “one billion knowledge workers” framing doesn’t survive contact with global income distribution. The US is roughly 50% of the world’s high-end knowledge worker software budget and 25% of global GDP, and global software spend runs about double US spend. Inside the US there are roughly 83 million knowledge workers, but most of them are nurses and teachers, not AI replacement targets. The actual core market is the roughly 5 million people in software-adjacent roles, and they earn about $600B a year in aggregate. **Rory’s read:** $200B of revenue means capturing about a third of every salary dollar paid to US software workers. That’s the real bar. The variable that decides everything is the steady-state ratio of salary dollars to AI spend. ## #14. The steady state might be $100K of tokens per engineer and 30% smaller teams. **Jason’s read:** running 10 agents in parallel around the clock costs something close to $100,000 a year. So the model that’s emerging is $200,000 in fully loaded wages plus $100,000 of AI per person, and a team that’s 30% to 40% smaller. That’s not just developers. It’s systems administration and QA too. Run that across the US software workforce and you land at roughly $200B, call it $350B globally. Two things that make this feel real right now: a lot of companies have started capping AI budgets after spend went exponential over the last 60 days, and CFOs are starting to use the $100K per head number as the justification for holding engineering headcount flat. **The dispersion is enormous.** Ramp data has the top 1% of tech-forward spenders at about $7,000 per month while the median sits near $100. At the top end, AI spend reaches 50 cents for every dollar of salary. ## #15. Pre-2023 companies shipping 2x to 3x faster is the signal that this isn’t token maxing. **Jason’s read:** the fastest companies are already working on their 2027 roadmaps. That’s the tell that AI spend has crossed from performative into productive. If you’re still running quarterly and monthly planning cycles in an agentic environment and you haven’t finished this year’s product goals, you’re behind, and the gap compounds. **Rory’s caution:** the surveys show wide dispersion, with many companies spending far less per head than the leaders. The bet is that spend correlates with output. That correlation is a hypothesis, not a proven fact, and it’s the single most important open question in the $200B math. ## #16. Going first is worth something in the IPO race. **Jason’s read:** Anthropic can go public with a profitable quarter and an enterprise leadership position. That’s a good first impression, and going later is a risk if either the company’s growth or the public market decelerates. **Rory’s read:** OpenAI has a harder path after executive turmoil and a slower first half, and probably prices more precisely and less enthusiastically as the second one out. Long run the ordering doesn’t matter much. Short run it does, because both companies have enormous ongoing capital needs, and even pre-IPO OpenAI can sell stock at a discount to its implicit valuation. Price clears every market. The terms are what’s uncertain. Related, and a good signal on how AI companies now think about GTM: Denise Dresser is out as OpenAI’s CRO and Dali Rajic is in. AI companies are hiring Salesforce operators to scale the way the last generation hired out of Oracle, but the preference at the top is for leaders forged in hyper-competitive environments like Wiz rather than classic B2B motions like Slack. ## #17. Silver Lake at $43B for Workday is the most honest valuation datapoint of the week. Workday stock jumped 18% on the news. The structure is a large equity check plus substantial debt, using roughly $3B of annual cash flow on $10B of revenue at about 35% operating margins to service it, aiming at a 20% IRR over five years. **Rory’s read:** this is a mature-phase transaction, not a growth one. Unlike venture, where entry price barely matters if you’re right, this is precise financial engineering. Overpay by 20% to 30% and your IRR falls to the low teens. The base case for a system of record growing 13% is about 5.3 times revenue. Software that isn’t a system of record, or that grows slower, prices closer to 2.7 times. Set that against OpenRouter changing hands near 70 times trailing revenue and you have the entire market in two numbers. Aneel Bhusri already came back after hiring a successor, which tells you the board thinks AI changed the job. If Silver Lake runs an agentic transformation with the founder leading it, the upside case is real. ## #18. Closed systems of record are safer right now. Open ones are more valuable and more exposed. **Rory’s read:** Workday is closed, which buffers it from agents siphoning value. Salesforce is open, which is why Gong, Outreach and Salesloft exist on top of it and why nothing equivalent exists on Workday. Open ecosystems get more innovation and give more of the value away. In the end, the money in an ecosystem gravitates toward the general ledger. **Jason’s read:** running Salesforce headless, with agents doing the interacting, makes the software far more useful and simultaneously far easier to abstract away. High logo retention is not the same as safety. The environment is grow or die, and holding customers hostage inside a system of record does not produce growth. CIOs are actively trying to pay 80% or 90% of last year’s bill even on contracts they’re locked into. **Where this bites:** private equity buying at 10 to 12 times revenue with no growth plan ends in a death spiral against the debt and a mid-teens IRR. Bending Spoons can triple prices, lose 20% of customers and come out ahead. That’s a strategy, not a business. Companies with no AI in the product and a thick middle management layer are the most exposed group in software right now. ## #19. Higgsfield at $5.5B and Lovable at $13.3B on roughly the same revenue. Higgsfield raised at $5.5B on roughly $700M ARR. Lovable raised from Menlo at $13.3B on something in the $600M to $700M range. Same revenue, more than double the price. **Jason’s read:** Lovable and Replit were mediocre products when this show started and are now generationally good. Both added automatic deep security penetration testing. Cursor is expanding into things like Origin and could plausibly replace the GitHub workflow. Adjusted for platform complexity, Lovable’s price is comparable to what Cursor commanded. Coding is also simply a deeper market than enterprise video, which is real but smaller. **What both did right:** start with high-volume consumer demand, website building and coding for one, video for the other, then layer mid-market and enterprise on top. Classic PLG, executed in an AI cycle. **Rory’s read on moats:** in new software markets, moats start light and accrete with execution. Netscape and MS-DOS both started simple. Every feature you add makes a viable alternative harder to build. Model development and bioinformatics are genuinely IP-dependent. Most software is not, and wins on being faster and better. These companies are also now talent magnets, and that becomes its own barrier. ## #20. $700M at $21B, three weeks after a $10B round. One company raised $700M at $21B from Jane Street, Kleiner Perkins, Sequoia and Andreessen, three weeks after closing at $10B. **Jason’s read:** you can now raise on one good month. Two years ago you needed three or four. **The related story:** the DOJ is looking at Andreessen over overlapping board seats, triggered during the Fivetran and dbt merger review and extending to Databricks and Fivetran. Section 8 of the Clayton Act, which dates to the early 1900s, bars a person from sitting on two competing boards. Whether two different partners from the same firm counts is much less clear. Precedent, including Thoma Bravo’s, is that the firm removes a director and everyone moves on. Low drama, low consequence. ## What Ties This Week Together: Speed Four transactions, one behavior. SpaceX paid $10B over an existing round to compress a build into a week. Stripe paid roughly 70 times trailing revenue for routing infrastructure it could have built. Silver Lake is paying 5.3 times for predictability instead. And Anthropic is spending against compute commitments that only make sense if 2027 revenue arrives on schedule. For founders, the practical version is narrower. The buyers who pay these prices are the ones for whom your worst cost line is their existing asset. If that buyer doesn’t exist for your category, the gross margin criticism you’re brushing off is going to be the thing that decides your outcome. Cursor had that buyer. Most companies do not. ## Quotable Moments ### Jason Lemkin “We’ve been doing this show about 70-something weeks and it almost seemed like Cursor was dead. Nobody’s portfolio companies were using it. I’m not even sure Claude Code had launched when we started.” “There have probably been three different Cursors since we started the show. This was not linear progress to $60B, and most teams would have given up on that journey.” “Zuck did Instagram and WhatsApp in an hour on the back of a napkin and paid high. Elon did even better.” ### Rory O’Driscoll “Your gross margin problem is my revenue opportunity for my Colossus cluster.” “The negatives you can cite along the way tend to be true. They were real. The positives in terms of market trajectory just outweighed them.” “Pessimists sound smart, optimists die rich.” ### Harry Stebbings “SpaceX closes the $60 billion all-stock takeover of Cursor, minting thousandfold returns.” “I’m intrigued why Zuck didn’t buy it. He’s building the model capabilities with Alex and co. He’s missing the enterprise capabilities, and this would have solved that the same way it solved it for Elon.” “It ended up not even being that expensive by the time the deal closed.” *This post is part of the ongoing 20VC x SaaStr collaboration with Harry Stebbings and Rory O’Driscoll.*