Thrive Raises New $10B Fund | OpenAI Buys OpenClaw | Stripe at $140B: Is Adyen Wildly Undervalued?

Summary

• Anthropic’s $30B raise at $380B post (upsized from $10B) is “the only play in venture,” per Jason Lemkin — with public software down ~20% this year, every multi-stage fund needs its 0.01% just to show up in the one thing working. Rory O’Driscoll’s caveat: “in the short term momentum massively outperforms value,” and Anthropic’s last two 2025 rounds ($60B, then $160B) each roughly 2x’d, so nobody fights the tape.
• The growth is genuinely unprecedented — and so is the fragility. Rory plowed through early Microsoft, Google and Compaq: “you’ve never seen a company grow 10x in GAAP revenue… year-on-year for 3 years… you’re leaning into the singularity here.” But those companies were wildly profitable; Anthropic is structurally closer to a semiconductor company, and Dario himself frames under- vs over-investing in compute as a bankruptcy-adjacent tightrope. At ~$4.5B GAAP revenue last year, maybe $15B this year, “30, 40 times GAAP revenue is a little pricey” once growth slows.
• Anthropic went from 5% of OpenAI’s revenue to 64% in 14 months — “that is stealing budget.” Enterprise is deciding, not evaluating: Jason’s core call is that Corporate America will “will AI into existence” regardless of ROI — Klarna’s Sebastian went 6,000→3,000 employees and wants 2,000 — meaning 1-2 years of mega AI budgets on “the presumption of success,” then, per Rory, probable over-investment and retrenchment two-plus years out.
• Public SaaS has entered the “dead zone”: the cohort is approaching 10% annualized growth, and if AI shaves another 30% off that, “it’s all just price increases and suing customers.” Jason can’t see a bottom because nobody wants more seats for humans “who complain and whine and quit every 3 months”; Rory’s rebuttal — “price clears all markets” — there’s a level where these compound at 12-15%, just not one holders will like.
• Even the anointed aren’t safe: Figma IPO’d in July as “the best of the best,” then Replit and Lovable took ~$350M each in ARR from product-prototyping — revenue Jason says Figma should own outright (“Figma Make is a failure”). And Shopify — which all three call oversold — carries a real agentic-commerce bear case: Tobi has done more code commits in 60 days than in his history because the software “in 2 years may be obsolete.”
• Stripe ($130-140B) vs Adyen ($40-50B): Rory buys Adyen — half Stripe’s revenue ($2B vs $5B), ~50% operating margins, 21% H2 growth, and public data you can actually value — while Jason takes Stripe for private-market flexibility. Both agree Adyen’s leadership “wildly miscommunicates” while Stripe tells a brilliant narrative. Rory’s prediction: Anthropic, OpenAI and SpaceX dive for the public line by year-end because capex-hungry companies have exhausted private capital.
• OpenClaw’s sale to OpenAI (Harry guesses ~$100M, not the rumored $1B) marks “a freaking movement in AI”: a product designed to break guardrails that ignited developers, forced Anthropic into a cease-and-desist it may regret, and made autonomous 24/7 agents inevitable — “the horse has left the stables.” Downstream trades: agent-first security products, and inference managed via Haiku 4.5 at ~1/20th the cost of Opus 4.6.
• Monday at ~$3.5B (down 51.3% YTD, $1.25B revenue growing 27%, ~$1B cash) is “the greatest buy of the value stocks” — if you still believe the December durability thesis. The tell that conviction is gone: Jason planned to buy $200K of four beaten-down names for the show, Shopify first — it crushed earnings, the stock didn’t move, and he couldn’t pull the trigger: “I couldn’t see the floor.” A live four-stock, $200K bet is now on for a coming episode.

Deep dive

1. Anthropic at $380B: the only trade in venture

• Jason’s read on the $30B round (upsized from $10B): “if you’re not in that, you’re not of interest to 98% of venture” — people he didn’t even know were on the periphery of the deal shocked him, and with public software stocks down ~20% this year, “this is the play… there’s almost no other play other than showing up to demo day.”
• Rory’s momentum logic: the last four or five rounds all worked, and the two 2025 rounds — “the $60 billion round at the start of the year, the $160 billion round at the end” — each delivered roughly a 2x. Once that’s true, a multi-stage fund “just got to put some money in and get your 0.01% ownership,” because “in the short term momentum massively outperforms value.”
• The supply-side symptom, from Jason: a $30M Series A with $1.2B of demand — “I’ve never seen so much money chase seemingly such a small concentrated number of companies.”

2. Unprecedented growth, semiconductor economics

• Rory went back through early Microsoft, Google and Compaq, even adjusting for GDP and inflation: “you’ve never seen a company grow 10x in GAAP revenue… year-on-year for 3 years” at this scale. “So, you’re leaning into the singularity here.” The asterisk: those comparables were wildly profitable while compounding; Anthropic is still losing a ton of money.
• The structural point: these aren’t software companies with free cash flow — “they are much closer to semiconductor companies,” with tens to hundreds of billions in capex ahead. Rory cites Dario’s “long and awkward podcast”: under-invest and miss a cycle, growth dies; over-invest and “you are really long compute” — with the implicit message that Anthropic is trying to be “more circumscribed than the other major player OpenAI.”
• The valuation trap Rory flags: getting to free cash flow assumes slowing growth, and once you slow, ~$4.5B GAAP last year and maybe $15B this year means “30, 40 times GAAP revenue is a little pricey.” The bear-case canary: enterprises saying in a year or two, “That was great, but we’re going to slow down on spend… we didn’t make the labor savings.”

3. From 5% to 64% of OpenAI’s revenue — “that is stealing budget”

• Jason’s distinction: it’s too early for Anthropic to be stealing enterprise deals (contracts are one-year and freshly signed; Anthropic’s 100K customers are up 7x in a year), but going from 5% of OpenAI’s revenue to 64% in 14 months “is stealing budget. You can’t argue with it.” Even with global software spend up an unprecedented 14% this year, “it’s still a fixed pie even if it’s growing.”
• Rory’s nuance — worth keeping: the two are “slightly separate companies with an overlapping Venn diagram.” Most of OpenAI’s revenue is consumer ChatGPT; Claude is less than 20% consumer, very much focused on enterprise and coding (Claude Code from $1B to $2.5B, per Jason — “that’s escape velocity”). The narrative isn’t Anthropic-wins-so-OpenAI-loses; it’s that capital “just loves AI and don’t love anything else.”
• Jason’s kicker on Sam’s internal memo: “You thought code red was about Gemini? I guess it was, but maybe it’s like missile command, they’re coming from everywhere.”

4. Corporate America is willing AI into existence

• Jason’s thesis — and he says this changed in the last 60-90 days: AI “does not miraculously replace 10,000 employees in an hour,” so enterprises must make a bet, and they’re choosing to also want smaller companies. “Enterprises are going to will this into existence. They want it to be true.” His exhibit: Klarna’s Sebastian went 6,000→3,000 employees and told 20VC “in two years I want to be at 2,000.”
• Rory converts it to a timeline: “Corporate America has decided they’re going to make this bet. The zeitgeist is making this bet. It’s unstoppable now” — meaning one to two years of mega AI budgets on “the presumption of success,” not proven ROI. “My gut would be people will over bet, over invest, and you will have a retrenchment period two plus years from now.” Which, he notes dryly, makes the next two years “a good time to access the capital markets.”
• On 2030 software spend: Rory’s frame is that software has grown GDP+200-300bps for decades; whether it stays at GDP+400-500 “is all about massive labor and efficiency and productivity savings” actually materializing from AI.

5. The SaaS gravity well: is there a bottom?

• Jason’s metaphor: “gravity’s almost gone up to Jupiter levels in tech” — Navan down to a $2B market cap, everything sucked into the well while two or three companies achieve escape velocity. His pessimism is arithmetic: public software as a cohort is approaching 10% annualized growth, and if AI cuts another 30%, “we’re in the dead zone… it’s all just price increases and suing customers.” Crucially, SaaS was sick before AI: “Dropbox growing minus 1% a year is not crushing it, is it?”
• Rory’s counter: he agrees on sentiment but rejects “there is no bottom” — “in the end, price clears all markets.” SaaS has flipped from the presumption of success to the presumption of failure, but there’s a price at which these compound at 12-15% a year. “The other side will overshoot… it’s just not a number you like.”
• His larger frame: Wall Street “fell in love with AI and to do that had to fall out of love with SaaS.” Public SaaS CEOs are now stuck trying to disprove a negative — “I think you might be displaced by AI. How do I disprove that?” — and until the momentum trade dissipates, “you would be a buffoon to get in the way of a momentum trade.” His grounding stat: after 20-30 years as the best industry in the world, software is ~4% of US GDP. “Everybody should get a little bit of a grip.”

6. Shopify: oversold, but the agentic bear case is real

• Rory picks Shopify as the baby thrown out with the bathwater: “I don’t think there is a credible end story that replaces a shopping website for shopping and a payments mechanism” — 60-70% of the business — and if running a long/short book, “maybe we put 5% of the fund into Shopify.”
• Jason agrees it’s oversold but insists the bear case exists: ServiceNow, Salesforce and Shopify “all could be abstracted away into a database.” If e-commerce becomes conversational and happens off-platform, Shopify keeps the plumbing short-term but loses the future — and Tobi himself “has done more code commits in the last 60 days than in the rest of the history” because the software “in 2 years may be obsolete.” “If Tobi thinks it… what hope is there for mere mortals?”
• Harry’s pushback: conversational commerce “is the UI of today” only for “the most productive people in the world — Sam, Elon, Dario”; consumers like browser-based discovery. Jason’s synthesis: “every platform that is open is at risk that an agent is better than the native platform” — he hasn’t logged into Salesforce in six months, but his agents use it every minute.

7. Define your terms: which SaaS is actually dying

• Rory’s taxonomy — Cloud is SaaS, Harvey is SaaS, Salesforce is SaaS, so “SaaS is dead” is useless without splitting: model companies, post-GPT “new school” (pejoratively wrappers), and old-school pre-2022 SaaS, which is what everyone actually means. The hardest place to be: “horizontal workflow-related software built pre-2022 that’s not public with critical mass.”
• Jason’s portfolio as the lab: his last fund is at 4.5x, but only two companies are “massively benefiting” from the current world; one nine-figure business is a push, “and everything else is struggling because of AI.” Companies at 8-20M ARR growing 80-100% are concluding “we’ve raised our last venture dollars” and stepping off the venture train. “It’s hard to remain the optimistic” when February updates show no AI boost.
• Rory’s public-company version, echoing Mike Cannon-Brookes: make the model work financially while investing in AI, because if you only harvest, “the good news is you have five years of 30% free cash flow, and the bad news is in year six, your revenue goes to zero.”

8. Figma missed the generation it should have owned

• Jason’s brutal audit: the number-one at-scale use case for both Replit and Lovable is product teams building prototypes — “that is all revenue Figma should have. Figma Make is a failure.” Both are at ~$350M independently; “they missed 400, 300 million of growth in their core for AI… I’d be pretty critical.” Base44 inside Wix is crushing it too. And the timing stings: Figma IPO’d in July 2025 as “the best of the best,” and Replit and Lovable took that market “maybe even since September or October because the product didn’t even really work until” then. “Let’s not view any islands of stability.”
• Rory’s valuation ladder: at ~$12B and roughly 10x revenues with 30-40% growth, Figma is “probably just getting valued correctly in the absence of narrative lift” — it’s moved “from the voting to the weighing part of the capital markets.” With another $200-300M of ARR it’d trade at 20-25x; without getting on top of this, “that 10x revenues is going to go down to five and four.”
• The generalization Jason draws: “If Figma was too slow, what hope is some of these other public companies?”

9. The six-figure deal an agent booked on day one

• Jason’s live evidence on pace: Monaco, his AI SDR investment (his fifth SDR agent, alongside Agentforce, Artisan, Qualified), launched last week and “booked us a six-figure deal the first day” — on its own it chose targets, reached a VP at a leading hyperscaler, and booked the meeting for a six-figure sponsorship. “None of these agents could do this before February… that’s just progress in 60 days.”
• His own caveat is the meta-lesson: “everyone’s promiscuous with agents. We will switch” — this one is so much better that he’d throw out an agent bought two months ago, “which shows you how fragile everything is.” No agent vendor is safe from disruption either.
• Rory’s continuum, pushing back on uniform doom: disruption speed tracks AI-adjacency — individual creativity and coding are “extraordinarily close” (Canva included), while “80% of the value in an accounting software package is independent of AI. It’s just debits and credits and a good UI.” Go-to-market software has so far been less acutely eaten than creative, coding or support. Jason’s grim reframe: on a terminal basis it may not matter — “if you’ve gone on to hospice care, does it really matter how long you stay in hospice?”

10. Sequencing is the edge: be six months ahead of the posse

• Rory’s history check on velocity: Salesforce went public in 2004, ServiceTitan in 2024 — the SaaS transition took ~20 years. AI won’t be that slow, “but I also don’t think it’ll be 2 years” — which is why picking which markets tip next matters. Legal and healthcare doctor-information have been amazing categories precisely because they had no compelling old-school incumbent: “legal was a miserable category in SaaS land. It’s been amazing in AI land because LLMs manipulate language, and lawyers manipulate language, too.”
• On strategy, Jason poses the fork — invest in maximum disruption today (Anthropic) or in the slowest-changing categories with the best founders and buy yourself “two, three, four years instead of two weeks.” Harry’s objection: the contrarian route only works with lifecycle funding, given the opportunity cost of capital. Rory’s resolution, via Henry Luce: “My job… is to be 6 months ahead of the American public. Not 2 years, not a week” — credit to Spark and Menlo looking at Anthropic in May ‘23.
• Harry’s ground truth from dating a lawyer: big law firms are willing AI in because “they are competing on price just like us as venture investors” — an existential adoption driver SaaS-era Atrium never enjoyed. Rory: pre-2022 you hid the AI in a sales cycle; post-2022, “if you don’t have an AI strategy as a corporate leader, you’re a buffoon.”

11. Stripe vs Adyen: value the data or buy the flexibility

• Rory went looking for a narrative mispricing and found mostly rational math: Stripe at ~$130-140B does ~$5B revenue to Adyen’s ~$2B at ~$40-50B, so adjust for the 2.5x size gap and they’re “a lot closer together than one would think.” Adyen is wildly profitable (~50% operating margins, 21% H2 2025 growth) but slowing; Stripe’s profitability and growth rate aren’t public. It’s “the eternal venture question: how much extra revenue multiple do you pay for how many extra points of growth?”
• Forced to choose, Rory takes Adyen — “the likelihood of it being mispriced to the upside is lower… you have access to the data” — while Jason takes Stripe for the flexibility of staying private: “it’s so stressful being public,” citing Cannon-Brookes’s must-be-more-profitable-while-massively-investing bind. Harry’s addendum: Adyen “wildly miscommunicates” while private Stripe ironically tells a brilliant strategic narrative — “Do they even have a podcast at Adyen management?”
• Jason’s prediction: back end of this year, “the big three dive for the public line — Anthropic, OpenAI, and SpaceX,” because capex-hungry companies have “sucked up all the private capital.” His structural gripe: public markets now only work “when you’re pristine and your narrative value is high” — “if every time you have to deal with a problem you have to go private to fix it, that strikes me as a little absurd.”

12. OpenClaw: the guardrail-breaking “movement” OpenAI just bought

• Jason’s initial take on Peter Steinberger’s OpenClaw was dismissal — “it’s designed to break guardrails… to run pseudo-autonomously 24/7, which Anthropic or OpenAI could do but chose not to” — and an agent already deleted his entire database last summer. But “it has ignited the developer community like something we have not seen in a long time”: one of the fastest-growing stars on GitHub, every cool engineer buying Mac minis and Mac Studios to run it nights and weekends. On price: “I think it’s about a hundred” million — he doesn’t buy the billion — and Steinberger probably turned down more from Zuck, who then cloned it on Manus the next day, proving “the technology itself is not that differentiated.”
• His verdict on Anthropic’s cease-and-desist (sent after OpenClaw and likely-Moltbook published users’ private keys and passwords): not a fumble — “it was hard to tell at the time… it was too early to know it would be a movement. Maybe the legal department could chill on the next one.” Harry concurs: a safety-branded company being half-named by “some cowboy” had to react. But it’s over: “the horse has left the stables for these autonomous agents.”
• What’s coming, per Jason: Replit’s V4 will spawn agents overnight — “when you log in the next day, it will have built three to four features for you on its own.” Inference math works if you run Haiku 4.5 (~1/20th the cost of Opus 4.6) every 20 minutes rather than Opus 24/7. And on responsibility: whoever initiates the agent gets fired when it goes wrong — CISOs included — so “even as we speak there’s people building really compelling agent-first security products.” Rory: “And we are about to invest in one.” Jason: “Of course you are.”

13. Thrive’s $10B and the GP exodus: “just show me the carry”

• Rory’s fund-size math: calling Thrive “early stage venture is just a mistake” — when four or five private companies are worth $100B+ and one may hit a trillion, a typical 5% position is $5B, “it’s just math.” Thrive’s model is radically simple: pick winners (Stripe, Databricks, OpenAI — monogamously OpenAI over Anthropic), then do every round at maximum size. Jason: “It’s a very calming model. The partner meeting’s very simple.” Harry’s aside: in a more sensible world these would all be public and someone could buy Fidelity mid-cap growth for 50 bips.
• On the same week’s departures — likely Arif Khan Mohammad and CRV’s Max Gazola (starting Striker, as heard) — Jason’s cynical opening: “Who wants to get back to early stage? I want to do the Anthropic round… Just show me the carry.” But he concedes Arif is post-economic, bored of big-firm bureaucracy, and wants to pick founders again. Rory: at a multi-stage firm, 90% of decisions are “do you lead the $60B pre at Anthropic,” not “do you put $10 million into this early-stage founder.”
• Jason’s structural warning: walking away isn’t free — carry now vests over 10 years, some firms backload it to penalize leavers, “Chamath, all his ex-partners have sued him”; even at a kind firm you’re “walking away from a couple hundred million of carry” and starting an 18-year clock from scratch.

14. Workday’s boomerang and the workforce that won’t do the work

• Aneel returning to run Workday after only ~8 months out is, per Harry, the tell: “You would think Workday would be one of those spaces that isn’t going to be disrupted overnight, and Aneel had to come back.” Rory — who usually resists founder mysticism — comes around: a hired executive can run cost/go-to-market playbooks, but when “the core thing you built has to be changed for a new way of building it,” only the founder knows the tradeoffs made 15 years ago. “What you need is massively specific knowledge and skills and courage.” On whether the stock works: Rory has no developed opinion — “I don’t think there is a magic pixie dust” for a mature category’s growth rate — while Jason expects faster change but admits “a lot of these bets won’t work.”
• Jason’s deeper diagnosis of incumbents: “no one wants to do the work in the age of AI” — VPs at public tech companies “all want to do the same job of 2023,” and the smart ones are quitting for hot AI companies hiring recycled SaaS executives. Mark, by contrast, is “driving more change in the last 8 months than the decade before that, and who but a founder could drive that level of stress?”
• The 11 Labs data point: 20x sales quotas, $4M quotas (vs a typical ~$1M SaaS quota). Rory’s decode: “It’s not that they’re hiring 4x better salespeople… when people want to buy your [product], it’s easy to sell it. Do invest in companies that are in markets that are exploding right now. End of complex analysis.” On who boomerangs next: the question is ruled out of order — Moskovitz “left the keys on the table,” Lawson’s doing fusion — though Jason salutes Drew and Aaron for grinding while “blabbermouths like us on podcasts are saying nothing you do matters.”

15. Monday’s math, and the $200K bet Jason couldn’t place

• Harry’s numbers on Monday: ~$3.5-3.8B market cap, down 51.3% YTD, $1.25B revenue growing 27%, 2026 guide of $1.45B, $175M non-GAAP operating income (14% margin); Jason guesses they have ~$1B in cash. Rory’s frame: “10 times cash flow for something growing sustainably at 20% is wildly cheap” — if the product roadmap survives the age of AI; if not, “you have to price the thing as if it can go away entirely.” Jason: if you believed durability in December — and everyone did — “this is the greatest buy that there is… you must be saying we believe that none of this revenue is durable anymore.”
• Jason’s durability hierarchy: Monday churns faster than HubSpot, HubSpot faster than Salesforce, ServiceNow slowest (99% GRR, five-year contracts) — but “it’s just delayed churn.” And durability has a definition: 100%+ NRR “for real, not just based on price increases and threats… Durable doesn’t count as dial-up at AOL shrinking every year.” He’s long Salesforce as an agentic platform but notes 10% growth built on acquisitions and price hikes “doesn’t suggest high durability, does it?”
• The confession that says everything about sentiment: two weeks ago Jason planned to buy $200K across four beaten-down stocks for the show, Shopify at the top — waited for earnings, Shopify crushed them, the stock didn’t move, and he still couldn’t buy. “I couldn’t see the floor.” Rory: “It is telling that you went to write the check and just couldn’t.” The stakes are now set: Harry and Jason each pick four names at $50K apiece and show receipts; Rory, traveling, gets two weeks.