Jul 2, 2026 · 1h 18m · news
Coinbase Cuts AI Spend by 50% | Kalshi's $40B Valuation & Impending IPO | The Year for SaaS Roll-Ups · 20VC with Harry Stebbings
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
This episode of 20VC features Harry Stebbings, Rory O'Driscoll, and Jason Lemkin as they analyze major shifts in the tech landscape, focusing on the pressure for real AI return on investment, the rise of SaaS roll-ups, and the evolving macroeconomic realities of venture capital funding.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 10.1% of the talking time here. How this is scored →
Harry as informed peer 3.9 Guest teaching 3.9 Guest disagreement 4.0 Harry pushing back 3.0
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:51 Jason's rant against performative CEOs
Jason aggressively rejects Brian Armstrong's AI spend post, expressing extreme burnout over non-AI CEOs sharing performative data on Twitter while their core revenues shrink.
Hardest push from Harry ▶ 1:03:22 Harry rejects criticism over Series A benchmark tweetHarry forcefully refuses Rory's suggestion that VCs need more compassion when setting expectations, telling sensitive founders to stay out of the startup arena if honest feedback ruins their day.
Biggest teaching moment ▶ 34:14 Rory dissects Microsoft's core AI vulnerabilityRory educates Harry on why his Microsoft holding is dropping, explaining that despite heavy CapEx, Microsoft lacks an owned frontier model while Anthropic and Cursor eat its core knowledge worker and developer franchises.
Harry holds his own ▶ 1:01:54 Harry asserts venture capital benchmark realitiesHarry demonstrates strong venture expertise by defending his decision to reject a startup growing from $1.5M to $5M ARR, explaining how the opportunity cost of cash in the AI era has permanently raised Series A hurdles.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Coinbase's 50% AI Spend Cut & The Search for AI ROI | 4 | 3 | 6 | 5 | Harry introduces Brian Armstrong's memo on cutting Coinbase AI spend by 50% through token routing and open source. Jason vehemently dismisses non-AI tech CEOs posting performative metrics, while Rory defends Coinbase's discipline as cost management 101. Harry pushes back on Jason's cynicism and steers the conversation to Anthropic's aggressive revenue targets. | |
| Anthropic, Chinese Distillation, and Regulatory Capture | 3 | 5 | 5 | 3 | Harry prompts the panel on Dario Amodei's allegations that Chinese AI models are stealing IP via distillation. Rory highlights the hypocrisy given foundation models' training on copyrighted books, framing Anthropic's Senate testimony as strategic regulatory capture. Jason and Rory debate whether protectionism, tariffs, or open market competition best serves US AI interests. | |
| Microsoft's AI Strategy & Decelerating Azure Growth | 2 | 6 | 2 | 1 | Harry admits to holding Microsoft stock through a 16% monthly decline and asks the panel to explain what went wrong. Rory educates Harry on Microsoft lacking an owned frontier model while tools like Claude Code eat away at its core developer franchise. Jason adds that Azure's decelerating growth from 40% to 37% acts as a canary in the coal mine for market expectations. | |
| Kalshi's Valuation & The Growth of Prediction Markets | 4 | 3 | 2 | 2 | Harry asks whether Kalshi's rumored $40B valuation signifies the ultimate casino-ization of retail investing. Rory breaks down Kalshi's TAM, noting sports betting and financial perpetuals drive revenue while political prediction markets remain a niche hobby. Harry contributes context on consumer stock-betting startups like FOMO. | |
| Bending Spoons & The Year of SaaS Roll-Ups | 5 | 3 | 3 | 3 | Harry examines Bending Spoons' impending IPO at a high 8-9x forward revenue multiple despite acquiring legacy consumer software assets. Harry sets up a roleplay scenario asking Jason which three B2B companies he would acquire for a B2B roll-up model. The guests discuss Marketo, PagerDuty, and Asana as prime targets for aggressive cultural and operational turnarounds. | |
| Chamath's $135M AI Startup & The Realities of Part-Time Founders | 2 | 2 | 6 | 2 | Harry introduces Chamath Palihapitiya raising $135M to lead AI software factory startup 8090. Rory commends Chamath for stepping into the arena as a founder, but Jason strongly attacks the premise of part-time VC CEOs, arguing that non-full-time commitments lead to failure. | |
| Series A Benchmarks & The Brutal Honesty of Twitter Backlash | 8 | 4 | 6 | 7 | Harry discusses the Twitter backlash he received after posting that growing from $1.5M to $5M ARR is no longer sufficient for a Series A in the AI era. Rory argues Harry came across as insensitive, but Harry aggressively pushes back, stating founders must handle blunt market realities. Jason validates Harry's underlying venture benchmarks while critiquing the framing. | |
| Claude Tag & The Future of Autonomous Agents | 3 | 5 | 2 | 1 | Rory prompts Jason on Anthropic launching Claude Tag inside Slack. Jason explains how cross-platform AI agents capturing organizational context could render traditional SaaS front-ends obsolete. Rory concludes that massive foundation model players operate at a scale where traditional software vendors risk getting trampled. |
Statements from this episode (30)
Lemkin: Struggling non-AI CEOs post performative AI metrics on Twitter
“I am getting burnout on struggling CEOs on Twitter sharing performative AI data when they're not AI companies. Like, show me the money.”
O'Driscoll: Unprofitable mega-cap tech companies should scale back ambitions
“If you can be the largest tech company on the planet and still not make money, you might have oversized your ambitions a little and it might pay to come back a bit.”
Rory O'Driscoll: AI-Era Software Companies Are Either Accelerating or Irrelevant
“Software companies in the age of AI are either accelerating or irrelevant.”
Lemkin: Portfolio SaaS companies see no revenue lift from AI coding
“Even any of our port, many of our portfolio companies that are doing well, that are not purely reselling tokens, they're coming to the conclusion that I'm not quite sure what the hell I know I want to do this. I can't put the genie back in the bottle. AI is gr…”
Lemkin: AI features helped Box return to double-digit growth
“It has led to a boost at Box. It hasn't turned Box into a hundred percent grower. It's gotten it back to double digit growths.”
O'Driscoll: Anthropic Settled Book Copyright Lawsuit Over IP Scraping
“All the foundation models, including Anthropic, were trained on other people's IP. To the point where Antropic recently settled litigation with a whole bunch of copyright holders on books because they had, quote, unquote, unfairly, to use Dario's word, applied…”
O'Driscoll: Anthropic Letter Tells Senate Chinese AI Firms Distill Their Models
“Antropic the company said actually in a letter to the Senate Banking Committee recently that basically the Chinese open source companies are bootstrapping their development of their state-of-the-art models by effectively breaching the terms of service of Antro…”
Lemkin: >10% Chance US Bans Chinese AI Models for US Companies
“I, let's, is it on Kalshi? It seems to be more than a 10% chance it's gonna happen.”
O'Driscoll: Protecting AI Frontier Models Harms Economy by Raising Intelligence Costs
“Because what you're saying is keep up the price of AI as an input to the rest of the economy by protecting, protecting these foundation models pricing structure so that they can get a trillion dollars in revenue. And the loser, as always is the case with trade…”
Lemkin: Leaving oligopolies triggers massive price erosion in AI models
“When you leave oligopolical stages, there's massive price erosion because you're no longer competing on features, you're competing on price, and that's exactly what we're seeing.”
O'Driscoll: Without open source, Coinbase's AI costs would have sextupled
“The counter counterfactual is if those open source product didn't exist, the Brian Armstrong quote would tweet would have been, We were paying ten million dollars six months ago for our AI intelligence. Now we're paying 60.”
O'Driscoll: Microsoft lacks compelling proprietary AI products beyond OpenAI stake
“So when you zoom out a million miles, Microsoft, which is in the software business, unlike Apple, and therefore can't afford to stay apart from the great AI wars, other than its equity ownership in open AI doesn't have anything compelling to bring to the table…”
Lemkin: Anthropic stock will be extremely volatile post-IPO
“You know, it'll be interesting if Anthropic really does IPO in the coming months, right, if it stays on track, I have to imagine it's going to be one of those volatile stocks out there, right? Even a hint of news will send it up and down math.”
O'Driscoll: Intercontinental Exchange owns ~20% stake in Polymarket
“ICE as in the owners of NASDAQ, the, I think, Intercontinental Exchange has a big ownership stake in Polymarket. Why that's relevant is that's a company that runs ICE that runs stock trading and, you know, real financial markets taking a significant, I think, …”
O'Driscoll: Political prediction markets are too small to build $100B companies
“I don't think predicting who's the next Who's going to win the general election in the UK or who's going to be the next president in the US is a huge, enormous, ginormous business. I love it. I find it fun. It's actually the most interesting part of Polymarket…”
O'Driscoll: Bending Spoons grows via price hikes and cost cuts, not users
“When you read the S one, you realize they don't get organic growth. They don't get new users. They have raised prices, cut costs, and suddenly that's being valued at You know, eight, nine, 10 times revenue”
Lemkin: B2B SaaS will see Bending Spoons-style roll-ups revive legacy software
“And I think there's also going to be another bending spoons in B to B that's going to nail this. They're going to buy these horrific products like Marketo and others, put a few smart people in it and just boost the NRR.”
O'Driscoll: Traditional PE cost-cutting fails in B2B SaaS without AI re-engineering
“If you buy a B to B software company today, it's pre-AI, it is highly unlikely that simply optimizing and pressing the buttons will be enough... Unless you're kind of generating new revenue from AI and significantly reengineering the company... It won't be suc…”
Lemkin: Declining public software companies will accept acquisition premiums as low as 15%
“I think any public company in decline today that gets an offer, a premium of even 15%, they have a fiduciary duty to take that very seriously... Management with their earn out, they're going to take that deal in a heartbeat.”
Lemkin: Asana product feels terminal and fits roll-up target criteria
“This is easy for me to, I don't want to be too much of an armchair quarterback from a product that feels terminal, but it's probably a good one. I think it's a good candidate, right?”
Chamath Palihapitiya raises $135M for AI startup 8090 as CEO
“Chamath Palihapitiya raises a hundred and thirty five million for his AI startup, eight zero nine zero or 8090, whatever he calls it. But he also is now CEO.”
Lemkin: Wealthy VCs who become startup CEOs run out of energy
“When wealthy folks, especially VCs want to be a CEO, but they're not working at The insane rate of a traditional founder CEO. I just find those run out of energy.”
Lemkin: Investments in multi-project founders yielded zero financial returns
“I've written small checks into successful founders doing multiple things just for fun, and they're all zeros.”
Lemkin refuses to invest in Chamath's 8090 without full-time commitment
“So I, I'm out. If literally Chamath drops everything, but 80, 90, everything. And I see the sweat from his brow and I see all, and I see that cause the dude's fit and he's looking good. Right. He just had, A massive, the massive largest win of his career as an…”
Stebbings: Growing $1.5M to $5M ARR is no longer enough for Series A
“They were finishing the year at one and a half million ARR, finishing next year at five million ARR. Today, brutal as it is, this isn't good enough to raise a good Series A.”
O'Driscoll: Preeminent Series A startups grow from $1.5M to $15M ARR
“The deals that are getting swept off are going one and a half to 15, right?”
O'Driscoll: Initial startup growth rate only modestly correlates with long-term success
“There is a correlation, but it's modest between initial growth rate and overall outcomes and companies that have grown slowly at the start have been huge at the end. Procore was a slow grower and then became a huge outcome.”
Lemkin: Autonomous AI agents could reduce Salesforce and HubSpot to dumb databases
“If the agent can run 24 seven autonomously, take all your data and build all the analytics, build all the dashboards, run autonomously out of it then your data can flow between apps and you won't even care where it lives and all the fears about headless become…”
O'Driscoll: AI agents lurking in Slack capture crucial context to automate work
“Capturing that context is really useful because it allows you, it allows an AI to automate that work. And the truth is a lot of that context exists in Slack. So if you're watching people interact on Slack, and if you watch it autonomously for weeks and months …”