Jan 8, 2026 · 1h 27m · news
Groq’s $20BN NVIDIA Deal | Why Sam Altman Doesn’t Care About Dilution & Invisible Unemployment 2026 · 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
In this dynamic episode of 20VC, host Harry Stebbings sits down with tech veterans Jason Lemkin and Rory O'Driscoll to dissect massive AI acquisitions like NVIDIA's $20 billion purchase of Groq and Meta's buyout of Manus, while exploring founder-VC misalignments, Sam Altman's unique dilution strategy, the rise of persistent AI companions, and the grim reality of 'invisible unemployment' in the 2026 job market.
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 10.5% of the talking time here. How this is scored →
Harry as informed peer 4.1 Guest teaching 4.3 Guest disagreement 2.6 Harry pushing back 3.0
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 56:52 Rory dismisses host's question line
Rory aggressively shuts down Harry's line of questioning about investing in real estate data centers, saying 'I thought you were in danger of asking an interesting question there, but you missed it.'
Hardest push from Harry ▶ 21:39 Harry challenges early founder acquisitionHarry forcefully refuses Jason's premise that Manus founders were smart to sell for $2.5B, insisting that at $100M ARR and triple-digit growth, he would be actively fighting the founders to stay independent.
Biggest teaching moment ▶ 46:34 Rory's LiveScribe hardware lessonRory surprises both Harry and Jason by revealing his historical investment in smart-pen company LiveScribe, delivering a masterclass on consumer hardware unit economics and writing behavior adoption hurdles.
Harry holds his own ▶ 16:40 Harry reveals unannounced Manus metricsHarry asserts high domain expertise and insider authority by reciting exact, non-public acquisition figures for Manus ($2.5B price tag, $100M ARR, 25x multiple, 5x ARR growth in 8 months).
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Montage of Key Video Highlights | 0 | 0 | 1 | 0 | This is an introductory montage clip reel and host introduction. Because it is a monologue intro segment without host-guest interrogation, host expertise and pushback scores are zero. | |
| Evaluating Training Versus Inference Workloads and Groq's Premium | 2 | 5 | 1 | 1 | Rory explains technical distinctions between training and inference compute workloads, using his portfolio company Tavis as a real-world example. Harry listens quietly while Rory lays out the product positioning. | |
| How Game Theory Shaped the Groq Acquisition Structure | 2 | 6 | 1 | 1 | Rory breaks down the game theory pricing mechanics behind Nvidia acquiring Groq for twenty billion dollars to protect its gross margins. Harry gives space for the guests to analyze the M&A dynamics. | |
| Jensen Huang's Fast Deal Execution and Chamath's Smart Bet | 5 | 3 | 2 | 3 | Jason details Jensen Huang's rapid deal execution while Harry interjects with specific financial data regarding Groq's revenue run rate. Harry then shifts the discussion to how the deal impacts rival Cerebras. | |
| How Landmark Deals Reset Psychological Barriers For Corporate Acquirers | 4 | 2 | 1 | 3 | Jason explains corporate M&A psychology and how landmark acquisitions clear mental barriers for corporate buyers. Harry catches and corrects Jason's verbal slip from million to billion. | |
| The Historical Realities of the Semiconductor Venture Capital Winter | 3 | 5 | 1 | 2 | Harry highlights Groq's non-traditional cap table and funding history, prompting Rory to detail the extended VC winter in semiconductor startups from 2000 to 2016. | |
| Breaking Down Meta's Two Point Five Billion Dollar Manus Deal | 7 | 2 | 1 | 2 | Harry demonstrates strong insider market knowledge by sharing unreleased deal terms for Meta's acquisition of Manus, including a 25x ARR multiple and a 5x growth rate in eight months. | |
| Why AI Founders Choose Local Maximum Acquisition Exits | 6 | 3 | 4 | 7 | Harry strongly challenges the premise of selling Manus at a two point five billion valuation, arguing he would push founders to hold out for higher valuation given their growth rate. Jason responds by pointing out founder tax and local maximum incentives. | |
| Navigating the Misalignment Between VCs and Founders on Exits | 5 | 6 | 6 | 7 | Harry plays devil's advocate regarding young founders taking early life-changing buyouts. Rory directly counters Harry's framing, arguing that VCs are diversified while founders face local risk maxima and shouldn't be forced to stay. | |
| How Multi-Billion Dollar Exits Impact Venture Capital Portfolios | 5 | 4 | 3 | 4 | Jason argues returning one-third of a fund is uninteresting to him, but Harry pushes back by noting Benchmark's fund returns. Harry then pivots to Yann LeCun's controversial FT interview regarding Meta and Llama benchmarks. | |
| How Spite and Frustration Are Driving AI Breakthroughs | 4 | 4 | 3 | 3 | Jason presents his theory on 'spite startups' driving AI innovation, while Harry provides corporate structure context on Yann LeCun and Alex LeBrun's new venture. | |
| The CEO Spite Driving a Return to Intense Corporate Cultures | 4 | 3 | 3 | 3 | Jason expands on executive spite against post-2020 remote work culture, while Rory agrees and discusses Zuckerberg's aggressive AI investment posture. Harry offers light banter on Larry Ellison. | |
| Why OpenAI's Sam Altman is Completely Indifferent to Dilution | 6 | 6 | 2 | 3 | Harry introduces detailed compensation metrics showing OpenAI spends 46% of revenue on stock-based compensation. Rory educates on the nuances of GAAP RSU accounting and amortized costs versus real economic transfer. | |
| Masayoshi Son's Forty Billion Dollar Conviction Bet on OpenAI | 4 | 3 | 1 | 1 | Harry prompts discussion on Masayoshi Son closing a forty billion dollar OpenAI investment. Rory and Jason analyze Masa's singular leverage risk tolerance and paper gains. | |
| Venture Lessons From the LiveScribe Smart Pen Era | 5 | 7 | 5 | 5 | When Harry brings up OpenAI's hardware pen device, Rory schools the host by recounting his own historical VC investment in LiveScribe. Harry playfully pokes fun at Rory's takeaways, leading to lively back-and-forth banter. | |
| The Shift Toward Persistent and Personalized AI Companions | 3 | 5 | 1 | 1 | Jason explains the strategic rationale for dedicated AI hardware in a 24/7 inference world, sharing how his Claude instance proactively named itself 'Ren' over the holidays. | |
| The Competitive Advantage of Feeding Personal Data to AI | 3 | 4 | 4 | 2 | Rory agrees on the value of permanent context accumulation for knowledge workers but explicitly pushes back on Jason's characterization of AI pseudo-sentience. | |
| Why Software Investors Avoid Heavy AI Infrastructure and Real Estate | 4 | 8 | 7 | 6 | Harry asks Jason why he isn't investing in physical data centers if he believes in continuous inference. Rory forcefully interrupts and mocks Harry's question framing, claiming Harry missed asking a truly interesting question. | |
| How AI Companions Prevent VCs From Lowering Standards | 4 | 5 | 2 | 4 | Jason describes integrating AI into his investment funnel to maintain deal discipline and filter cold inbounds. Harry pushes back with a witty joke about Jason's missed opportunities in unread emails. | |
| The Reality of Public Markets and Navan's IPO | 6 | 4 | 2 | 3 | Harry introduces Navan trading at 4x ARR post-IPO. Jason and Rory analyze private versus public market multiples, capital structure debt obligations, and mega-scale private companies like Revolut and Databricks. | |
| Invisible Unemployment and the 2026 Tech Job Market | 5 | 5 | 3 | 3 | Jason presents his thesis on 'invisible unemployment' in tech for 2026. Rory shifts his previous stance to agree with Jason's diagnosis of entry-level and senior executive labor market dislocations. |
Statements from this episode (55)
Lemkin: People will carry AI 24/7 once they believe it is alive
“When most people believe their AIs are alive, you will take it with you, 24 seven.”
Stebbings: Nvidia acquired Groq for $20B cash at 3x valuation markup
“Grok being acquired for twenty billion in cash. It happened just before Christmas. Chamath obviously coming out as one of the big winners. Price is three X more than the last round price.”
Lemkin: Nvidia will crush earnings in 2026 despite surging competitor attacks
“Nvidia's numbers are gonna crush this year, but we're gonna see all the daggers really coming out.”
O'Driscoll: NVIDIA operates at 75% gross margins and $100B cash flow
“They have a small number of customers Whom they charge 75% gross margins to and kick off a hundred billion a year in cash, right?”
O'Driscoll: Nvidia bought Groq to eliminate one of six margin threats
“There's only five or six people that can exert margin pressure at all on Nvidia. This was potentially one of them, and let's get it off the table.”
O'Driscoll: Groq generated $4M in 2023 and ~$40M in 2024 revenue
“As I said, four million dollars in 23, forty-ish in 24.”
Lemkin: Jensen Huang closed NVIDIA's $20B Groq deal in two weeks
“He reached out to Jonathan Ross literally just weeks ago, and the deal closed for twenty billion within two weeks. And he told his whole team, I want it done before Christmas, and it was done ahead of time.”
Lemkin: Offering 3x last round valuation removes M&A deal objections instantly
“When you come in hot to buy a company and take it off the table, three X is a traditional way to remove objections. And close the deal instantly.”
Stebbings: Groq is generating $175 million in revenue
“They're at a hundred and seventy-five million in revenue, so that would be quite an AR multiple to get to where they are.”
O'Driscoll: Bankers Will Use Groq's $20B Deal as Cerebras Valuation Comp
“So from the banker process of, you know, you use comps to justify value, now you've got the world's best comp. Right? And believe me, they'll be using it, right?”
O'Driscoll: Nvidia-Groq Deal Is Marginally Net Positive for Cerebras
“My guess is marginally net positive, but kind of, it's like the real, it's one of those weird things where you ask, is it a positive or negative? The truth is there's a big embedded positive, the comp, and there's a big embedded negative, the musical chair pro…”
Lemkin: Nvidia will likely deprecate Groq's product line post-acquisition
“One, they could do, they could just deprecate the entire product line and do something brand new. They probably will.”
O'Driscoll: Nvidia's $20B Groq deal normalizes mega-acquisitions for tech buyers
“There's this, everyone, you know, six months ago, if you said, I think we should buy Cerebus or Grok for 20,000,000,020 billion, you're at one of these other companies, you know, five people would dump at you and say, idiots. Now you say it and they go, well, …”
O'Driscoll: VC semiconductor startup exits were non-existent from 2003 onward
“And if you look at the semiconductor exits from 2000 about three on, almost none, right? In venture land, right? A very hard way to make money.”
O'Driscoll: Groq acquisition won't yield more than two major semiconductor outcomes
“I don't think there's going to be 10 more semis. There might be two.”
Stebbings: Meta acquired Manus for $2.5B at 25x ARR
“The price was two and a half billion dollars. It was a 25 X current ARR. They were at a hundred million of ARR, doing a hundred and twenty-five million run rate, including consumption.”
O'Driscoll: Manus is not an obvious product fit for Meta
“It's not obvious why a kind of B to B or individual knowledge worker work tool, which is what Manus is, is an obvious product to roll out, you know, three billion users on Facebook, most of whom don't do knowledge work.”
Stebbings: Manus team retained 80% equity prior to acquisition
“So the team had 80% of the company still. They took very little dilution.”
O'Driscoll: VCs should never force founders to reject exit offers
“And what you'll discover pretty quickly is trying to persuade a founder to hold on when they don't want to is a very hard thing to do, right? And arguably you shouldn't even try. Right? In fact, not even arguably, you shouldn't even try. In the end, you know, …”
O'Driscoll: Blocking Founder Acquisition Offers Creates Toxic Board Dynamics
“You can't leave a founder in a deal Where he didn't want to sell, where he wanted to sell, and you stuck him in the deal, and now he's bitter. And if the thing goes wrong, you have no idea how toxic that will be.”
Lemkin: Startup founders should always sell when offered an acquisition
“I tell all founders to sell. That's my advice.”
O'Driscoll: Corporate AI imperatives conflict with leaders skeptical of LLMs
“Once it became a corporate imperative for Meta to be competitive with OpenAI and Anthropic in the LLM space, having a guy running that product, running that project, whose first line is, I don't think this is important. We should do something else. Just causes…”
Lemkin: Venture investors must seek out spite to back winning AI startups
“Like, I think if you want to make money in venture, you gotta search out Spite. This is driving the greatest AI companies of our generation is Spite.”
O'Driscoll: Silicon Valley was founded on spite, from Fairchild to Intel
“First of all, Silicon Valley was founded on spite in the sense of if you look at, you know, Fairchild, if you look at Shockley, all those guys spun out the fair and did fair and Fairchild, which was a spite deal. And then they left Fairchild because they could…”
Lemkin: Investing in part-time, non-executive chairman founders is flushing money
“I believe, and don't get me wrong, this particular deal may work, but in my life experience, You're just flushing your money down the toilet. When a great seat founder reaches out to you, I've got this one I want to do. And then you dig a little deeper. I'm th…”
O'Driscoll: Non-Executive Chairman Role Is Ideal for Non-Commercial Academic Founders
“I might argue a different, I actually, it may well be that's actually the correct configuration if you have a highly talented academic who isn't a commercial driver.”
O'Driscoll: Skeptical the Market Can Support Ten Separate AI Research Labs
“It will be interesting to see, can the world support 10? Research labs coming up with 10 different variants of AI, all of which have to build either commercial or consumer businesses. I think on aggregate I'm fairly skeptical, but it's less a quality of, I mea…”
Lemkin: Mark Zuckerberg is investing heavily in AI out of spite
“It's CO spite about how they were forced to run companies in 20, 21. I think there's versions and I'm a big Sheryl Sandberg fan, but I think Zuck is spiteful. He was forced to run The company in a, in, in a certain way. I know a lot, I mean, Owen from Intercom…”
Lemkin: Longevity will be the dominant tech theme of 2027
“This will be the year of twenty-four-seven AI, but I think longevity will be the theme of twenty-twenty-seven.”
Lemkin: OpenAI's comp is 3-4x higher than Anthropic's
“This is like three to four X the comp of Anthropic effectively.”
Lemkin: Anthropic has done far fewer secondary sales than OpenAI
“They've had a very limited number of secondaries compared to open AI.”
Lemkin: Sam Altman's zero equity stake leaves him unbothered by dilution
“Sam just wants to build the biggest, greatest AI planet on planet Earth, and if he dilutes everyone, 99%, it doesn't impact him at all as a shareholder.”
O'Driscoll: Anthropic's per-share growth is under 5x despite 15x valuation jump
“In the year when, you know, when Entropic raised money at four or five billion, then there is a 14, then they raised at 60. Your rough math would say four billion to sixty billion is a 15 X. It's just under a five X on a per share basis because you're issuing …”
Lemkin: OpenAI retains only around 60% of its AI researchers
“They still only have like 60 something percent retention in researchers at OpenAI.”
O'Driscoll: Masayoshi Son negotiated a $40B OpenAI investment deal deadline
“Massa had negotiated a deal to put in, I think, forty billion into OpenAI, but it had to close by December 30th of this year.”
Lemkin: Masayoshi Son is OpenAI's only double-digit shareholder
“He's the only double-digit shareholder.”
O'Driscoll: Startup failures usually stem from bad ideas, not complex causes
“Failure rarely has complex idiosyncratic shit. It just failed because it was a dumb idea and it didn't work.”
Lemkin: OpenAI's upcoming hardware device with Jony Ive will be successful
“I think calling what I think this pen is going to be eventually pretty successful, but I think calling it a pen may be confusing. I don't think this is a writing instrument.”
Lemkin: Aims to invest in one AI agent startup per quarter
“I've got to work on AI agents. That's where I get so much because of all this SaaS stuff, I get so much inbound on AI agents, whether it's GTM or otherwise, I got to pick the best thing I can once a quarter and just do that deal and just try to do a 50 X or a …”
Lemkin: Completed a VC investment recommended by an AI VC tool
“I've done one deal cause my AI VC recommended it, which is, that's exactly it. So that's one, and that's just a start. That was just 20, 25.”
Lemkin: I will never lower my investment standards again
“I'm no longer ever going to lower the bar. If it means I do one deal every two years, don't care. Never gonna lower the bar again, not once.”
Lemkin: VCs can identify top 0.1% founders without speaking to them
“You can identify a point, a top, oh.1% founder Without talking to them. I don't think you can do it without any interactions, but without talk, I do, I a hundred percent believe it's possible.”
Lemkin: Most of my best deals originated from cold emails
“Most of my best deals have been from cold inbounds and I have begun that journey by analyzing the in inbound, analyzing the intelligence and the quality of the founder, analyzing this. It's not an, and I've done multiple billion dollar exits from cold inbound.”
O'Driscoll: Navan is significantly undervalued at 4x ARR
“The fundamentals are this is a 27% growth company trading at four times revenue that's cashflow positive and operating income positive on a non-GAAP basis. Actually, after this call, I made a mental note to follow Andreessen and buy some, right? Perfectly good…”
Lemkin: Navan was forced to IPO due to debt obligations
“I think Navon had seven hundred million in debt and only two hundred million in cash. It had to IPO to pay off its debt.”
Lemkin: Tech startups below Figma's caliber will struggle to IPO
“If you're not Figma or better, it's going to be rough out there. And that all these VCs are saying you can IPO at a hundred million and the IPO, the Vaughn says no to me. It says it's just barely cracked opened.”
O'Driscoll: Late-stage VC has bifurcated into sub-$400M and mega-private tiers
“There really are two different late stage businesses now. It used to be late stage was, I remember Meritech would say, beyond 10 Million is late stage, and you'd think 10 to a hundred million was quote late stage. Now, late stage is 10 to four hundred million,…”
O'Driscoll: Public markets have a core problem if founders prefer remaining private
“And if the answer keeps coming back, being private is the best way to achieve your dreams. The public markets have a problem and they need to figure it out.”
Lemkin: Going public enables Databricks to execute tens of billions in M&A
“Maybe at some point, even if you're Revolut, if you really want to do deals for real, it's, there are limits to private stock in cash, and if Databricks goes public, and it's worth two to three times Snowflake, it can buy, it may be able to buy the parts of it…”
Lemkin: Invisible tech unemployment will grow significantly throughout 2026
“I call it invisible unemployment, and it's all around us, and it's gonna grow this year. It doesn't show up in the government numbers yet but it's everywhere.”
Lemkin: Workforce reskilling is a total delusion that never succeeds
“Reskill is a delusion. Reskill is something we say to make everybody feel better. Reskilling is like when you get, when you, they do layoffs and they bring you in the room and they give you a packet of jobs potentially. It's just to make people feel better. No…”
Lemkin: 2026 will mark the end of entry-level sales jobs
“There'll be, this is the year where we will see the end of so many entry-level sales jobs. We still need AEs. We still need people knocking on doors. We do not need two-year-old kids sending emails. Those jobs will disappear.”
O'Driscoll: The employee quit rate is the best metric for labor weakness
“One of the real tells on unemployment is quitting rate. People don't quit their job when they know they won't get another job. And if you look at the quit rate, That's a really good tell that, oh my God, I know where I am is good, and I ain't leaving, right?”
O'Driscoll: Young founders are the most ruthless judges of young talent
“The scariest judge of young talent is young talent, right? And I think one of the really interesting things is, I mean, you talk to any of your founders who are 25, they're marked to market on their team and on other people. They know because that's, you know,…”