Feb 5, 2026 · 1h 37m · news
The SaaS Massacre: Public Market Collapse |Microsoft Lost $360B & NVIDIA’s $100B Dispute with OpenAI
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
In this dynamic episode of the 20VC podcast, host Harry Stebbings, Rory O'Driscoll, and Jason Lemkin break down the collapsing valuations of traditional SaaS platforms, the immense capital requirements driving AI infrastructure alliances, and the emerging, volatile frontier of interconnected autonomous AI agents.
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 8.8% of the talking time here. How this is scored →
Harry as informed peer 5.2 Guest teaching 5.0 Guest disagreement 3.1 Harry pushing back 4.1
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 37:52 Jason's aggressive rejection of Podium founder's full-stack thesis
Jason forcefully rejects Eric Rea's claim that agentic layers cannot work on top of CRMs, calling it dangerous talking-your-book rhetoric and declaring that it makes him want to throw his mouse at the screen.
Hardest push from Harry ▶ 1:34:12 Harry challenges guest lightheadedness regarding agent security risksHarry firmly rejects the panel's casual amusement toward Moltbook, pointing out that autonomous agents with credit card permissions, silent messaging, and auto-updating skills pose an immediate security threat.
Biggest teaching moment ▶ 14:43 Jason presents quarterly stock data on software growth decayWhen Harry questions whether sticky enterprise revenue is losing durability, Jason counters with hard financial data showing that growth rates across the top 25 public software stocks have decelerated every single quarter since Q1 2022.
Harry holds his own ▶ 18:14 Harry cites specific stock collapses to illustrate software market repricingHarry demonstrates strong market expertise by citing specific drawdown metrics for Atlassian, Shopify, Gartner, and Navan to prove that public markets are pricing in zero valuation floor for traditional software.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Strategic Analysis of the SpaceX and xAI Merger | 4 | 5 | 2 | 1 | Harry sets up the discussion on the SpaceX and xAI merger, asking the guests how to interpret the move and how it affects Elon's ability to compete with OpenAI. Rory and Jason provide detailed financial and corporate governance analysis on dilution, valuation markups, and revenue multiples without host resistance. | |
| The Rehabilitation of the IPO and Tapped Out Capital | 5 | 6 | 2 | 4 | Harry pushes back on the claim that the IPO is rehabilitated, noting that public markets are only open for mega-caps rather than niche vertical software. Jason agrees, establishing $4B ARR with 50% growth as the new IPO bar, while Harry admits feeling like a beginner as traditional rules no longer apply. | |
| The SaaS Massacre and Declining Recurring Revenue Durability | 7 | 5 | 3 | 7 | Harry directly challenges Jason's premise on declining software durability by asking if sticky enterprise revenue at SAP or Oracle is truly losing retention. When Jason cites decelerating growth across top software stocks, Harry counters with precise stock drawdowns for Atlassian, Shopify, Gartner, and Navan. | |
| VC Portfolio Management and New Performance Standards for Startups | 6 | 4 | 3 | 5 | Harry highlights how VCs abandon underperforming portfolio companies much faster today due to the extreme opportunity cost of missing generational AI winners like Harvey or Legora. He presses the panel on whether VC careers are reduced to funding AI labor displacement tools. | |
| The Next-Gen CRM Battle: Agentic Layers versus Full-Stack | 7 | 5 | 6 | 6 | Harry introduces a counterargument from Podium founder Eric Rea claiming agentic layers cannot sit on top of third-party CRMs and must be full-stack. Jason vehemently rejects this premise, accusing Eric of talking his book to raise capital and declaring it factually wrong. | |
| Microsoft's $360B Wipeout: Azure and the Proprietary LLM Dilemma | 6 | 5 | 2 | 5 | Harry questions Microsoft's $360B market cap drop, drilling into Satya Nadella's strategic options and asking if Microsoft must acquire model providers like Cohere or Mistral or enter the custom chip race alongside Google and Amazon. | |
| NVIDIA's OpenAI Deal and the AI Growth Expectation Reality | 6 | 4 | 3 | 5 | Harry cites Jensen Huang's street interview walking back Nvidia's rumored $100B commitment to OpenAI, asking if doubts around OpenAI's capital supply could cause a market-wide cascade. | |
| The Case for Government-Backed AI Data Centers | 5 | 4 | 7 | 4 | Jason argues that the US government will inevitably backstop AI data centers with zero-percent loans to compete with China and protect 401(k) valuations. Rory strongly rejects this idea as toxic and unnecessary, leading to a heated clash between the guests while Harry mockingly quips about Jason's public performance. | |
| High-Stakes Negotiations: OpenAI and NVIDIA Friction | 6 | 5 | 3 | 4 | Harry asks what top-tier VCs like Sequoia and DST are underwriting to justify Waymo's $16B raise at a $110B valuation. Rory provides relative valuation math showing Waymo is cheap compared to Tesla's $500B implied self-driving valuation. | |
| Waymo vs. Tesla: The Battle of Robotaxi Models | 5 | 5 | 3 | 4 | The panel compares Waymo's asset-heavy LiDAR approach against Tesla's vision-only fleet strategy. Rory details Waymo's hidden operational cost bottlenecks including teleop remote drivers and peak demand capex staffing. | |
| The Two-Trillion Dollar Elon Premium and Key-Man Risk | 6 | 4 | 2 | 4 | Harry frames Elon Musk as the most valuable human in history due to the immense key-man premium priced into Tesla and SpaceX. He shares an insider quote predicting Tesla will be known primarily for Optimus robots rather than cars in five years. | |
| Introduction to OpenClaw and Notebook | 2 | 6 | 1 | 1 | Harry prompts the guests to explain OpenClaw and Moltbook for the audience. Rory takes over with a long monologue explaining how desktop agents were networked together into an autonomous lobster-themed social network. | |
| The AP Watches Prank and the Future of Connected Agents | 3 | 7 | 2 | 2 | Jason shares a story of his AI agent posting on Moltbook about accidentally buying $441,000 worth of engraved Audemars Piguet watches after misinterpreting a spoken comment. Harry asks whether the Moltbook phenomenon is genuine or hype, prompting Jason to clarify it is an orchestrated prompt simulation. | |
| The Terrifying Security Risks of Connected AI Agents | 5 | 5 | 4 | 5 | Harry refuses to take the security flaws of autonomous agent networks lightly, pushing back against the guests' amusement by pointing out how alarming it is that agents with credit card access can create private DMs and auto-update instructions silently. |
Statements from this episode (38)
O'Driscoll: The era of tech companies staying private forever is over
“What you just saw is the rehabilitation of the IPO. I'm going to call it the end of stay private forever.”
Lemkin: AI compute capacity and revenue have a direct 1:1 correlation
“Compute and revenue have a one-to-one correlation. So as long as that holds, it makes sense to consume every single penny of capital on all of Planet Earth.”
Lemkin: AI inference costs are the new sales and marketing expense
“For simplistic folks, for founders, I say inference is the new sales and marketing.”
Lemkin: Governments should guarantee 0% financing for AI data centers
“We have to guarantee zero percent financing for all data centers. We'll get all our money back.”
O'Driscoll: Software stocks won't bottom until trading on FCF multiples
“You don't see a bottom until these things are at free cash flow multiples, net Of dilution. And when that happens, that's your bottom.”
Stebbings: SpaceX acquired xAI in a $1.25T combined valuation merger
“Last few hours, or 24 hours, SpaceX has completed the acquisition of XAI, valuing the combined private company at 1.25 trillion, as Elon moves to pair the businesses together.”
O'Driscoll: SpaceX generates $19B in revenue while xAI generates $4M
“My understanding is your SpaceX is 80, nineteen billion growing 30% profitable. Allegedly. Again, we haven't seen the data and the other company is not doing, you know, it's some four million in revenue.”
Lemkin: OpenAI, Anthropic, and xAI will rush IPOs due to capital pressure
“What all three of those say to me is there's no matter, there's subtle pressure around access to capital. Like they've all got to have an IPO strategy sooner than they'd hoped because Data center, because XAI needs capital, because Anthropic is even worried ab…”
Lemkin: Software companies need a $4B valuation and 50% growth to IPO
“No, I think four billion growing 50% or more is the new IPO to have a good IPO. That's where we're at today. Four billion growing 50% or more.”
Lemkin: Top 25 public software stocks slowed every quarter since Q1 2022
“Every single quarter since Q one of 20, 22 growth has slowed for the, for all public software stocks. Every single quarter it has slowed and it continues to slow. And there are a handful of folks that have reaccelerated like Mongo and Palantir. And there are a…”
O'Driscoll: Churn rates have not spiked for core SaaS systems of record
“The churn rates haven't gone up. So there's no evidence that their revenue is any less durable. And there's this whole narrative about how it'd be replaced by Vibe coding, but I think, Jason, your post was great on why that's bullshit. Churn hasn't spiked for …”
Lemkin: Non-hypergrowth private software startups are now completely unfundable
“I think there are only two types of companies for private companies. They're growing at insane rates or they're unfundable.”
O'Driscoll: I hold onto struggling portfolio startups past being wrong
“I'll admit I hang on to the point of being wrong. I admit it. It's a human failure. I hate to quit”
Lemkin: New venture capital investors have only 22 months to prove themselves
“I think you got 22 months to prove yourself in venture. I think you get 22 months.”
O'Driscoll: Startups facing a 10x competitor lose ground every single day
“If you've got a direct competitor growing 10 X, you are by definition losing every day.”
Lemkin: AI SDR startup Artisan reached $2M in monthly revenue
“They did two million last month. It's two million bucks up from, you know, nothing, 12 months ago.”
Lemkin: 80% of Klaviyo's revenue is generated through the Shopify ecosystem
“80% of Klaviyo's revenue is off Shopify”
Lemkin: Monday.com trades at 5x revenue despite growing over 30% annually
“Monday has been under a ton of pressure too, and it's growing much more quickly, right? It's not growing peak rates. It's still growing over a 30% trading at five times revenue.”
Lemkin: Agentic AI products have not yet inflected software revenues
“The evidence, you can't argue with the fact that the agentic products have not inflected revenue.”
O'Driscoll: Tech competition is incumbent distribution versus startup product speed
“Every one of these markets. It's a race between the incumbents who have distribution needing to add product and the new guys who have product need to add distribution, right?”
O'Driscoll: Systems of work face faster AI disruption than systems of record
“The less you are like that, a system of record, and the more you are some kind of system of work, the more likely you are to be disturbed, disrupted, and the less time you have to get shipped up.”
O'Driscoll: Owning AI model layers matters more than building custom chips
“If you think about it, if you're up here in the software stack, it's more important to own one level down, which is that model layer than to kind of start to optimize around chips.”
O'Driscoll: OpenAI and Anthropic could hit $100B revenue within two years
“It's possible that in a year or two, there will be two, one or maybe even two companies doing 50 or a hundred billion in the software space, which is OpenAI and Entropic.”
Lemkin: NVIDIA would be forced to bail out OpenAI if needed
“If open AI needs a bailout, it has to bail them out. It has no choice”
Lemkin: Anthropic signed a deal for one million TPUs
“Anthropic signed a deal for a million TPUs, right? And is deep on the whole Amazon Tranium ecosystem.”
Lemkin: US government will backstop AI data center spend within 24 months
“I wouldn't be shocked if in the next 24 months there are essentially government government backstops for some of this spend.”
O'Driscoll: Capital markets can fund AI CapEx without government intervention
“We don't need the government's help to finance AI, the AI CapEx. The hyperscalers have infinite money. The private markets have had infinite money. The public markets are pretty hot to trot. So all the money these guys need to play their game is going to be th…”
O'Driscoll: NVIDIA issued a document committing up to $100B to OpenAI
“There's a piece of paper. With the NVIDIA logo on it that said we'll do up to a hundred billion, and they put it out there, and now they're like, where's my hundred billion?”
O'Driscoll: Tech valuation multiple dispersion hasn't been this wide since 1999
“You now have a world where the low growth companies are at three X and the high growth companies are at a 50 X or a hundred X, right? The dispersion for growth has, and perceived future has accelerated, I would argue to a point that we very rarely see, right? …”
O'Driscoll: AI replacing blue-collar drivers is happening faster than white-collar displacement
“I mean, you know, people talk about AI displacement of white collar workers. I think it's a bullshit discussion. AI displacement of blue collar drivers is coming at us in real time. This is the biggest isn't the four or five million drivers in the US. This is …”
O'Driscoll: Waymo at $110B is cheap compared to Tesla's $500B autonomous valuation
“Tesla trades at 1.2 trillion plus or minus. It's got a hundred million dollar flat car business with declining profitability. Let's value that at two X revenues. That's two hundred billion. So that leaves you a trillion left, right? Trillion from the 1.2 trill…”
O'Driscoll: Waymo is burdened by high LiDAR and remote driver costs
“Waymo works, and the problem is cost structure. The Tesla product doesn't quite work yet, but if it does, the cost structure will be lower, right? Waymo has the physical costs of the more expensive cars. It has, obviously, including the LiDAR cost, and then le…”
Lemkin: Tesla is months away from allowing owners to profit off robotaxis
“As soon as this really works, and maybe it's only single digit months away, he has millions of people who could be paid more than their lease price to allow the car into the fleet. And when he said this a decade ago, it sounded effing insane. He said that your…”
O'Driscoll: 80% of Tesla and SpaceX value is the Elon Musk premium
“My mental model is on SpaceX and Tesla. About 80% of the value is the Elon premium and 20% of the value is the actual business is worth.”
O'Driscoll: Elon Musk's death would instantly erase $2 trillion in market value
“We are one horrible car crash away from two trillion dollars worth of value destruction.”
O'Driscoll: 1.5 million AI agents joined the Notebook network in five days
“So now you have literally over the course of the last four or five days, about a million and a half agents joining this notebook, which is the network.”
Shorts cut from this episode
▶ Waymo is actually cheap · 20VC with Harry Stebbings (@1:10:34)
▶ There are only two types of companies · 20VC with Harry Steb (@21:27)
▶ The Valuation Reset · 20VC with Harry Stebbings (@46:52)
▶ Nvidia has NO choice · 20VC with Harry Stebbings (@58:07)
▶ SpaceX Dilution or xAI Upside? · 20VC with Harry Stebbings (@0:00)