Jun 19, 2024 · 1h 2m · news
Michael Eisenberg: How China Could Overtake the US in the AI Race | E1167 · 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
Venture capitalist Michael Eisenberg shares a contrarian outlook on the modern technology landscape, analyzing the AI gold rush, the structural decline of SaaS, geopolitical tech rivalries, and the survival mechanics of post-ZERP venture capital funds.
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 23.8% of the talking time here. How this is scored →
Harry as informed peer 4.8 Guest teaching 4.7 Guest disagreement 3.2 Harry pushing back 4.0
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 41:29 Forceful rejection of partial position selling
Guest strongly rejects host's argument that selling 100% of a position is a strange take, insisting that venture is a binary conviction craft where partial selling makes no sense.
Hardest push from Harry ▶ 41:03 Host calls guest's exit strategy a strange takeHost explicitly refutes guest's philosophy of liquidating full equity stakes in winning companies, calling it respectfully a strange take and explaining why he would always keep upside.
Biggest teaching moment ▶ 36:06 Guest reframes the public IPO window realityGuest corrects host's assumption that the IPO market is closed, explaining that windows are open for companies willing to take realistic market pricing rather than waiting for inflated private valuations.
Harry holds his own ▶ 18:46 Host challenges software slowdown with Accenture GenAI revenue dataHost demonstrates his market knowledge by countering guest's claim that corporate software purchasing is slowing down, citing Accenture's $2.4B in GenAI revenue as evidence of high demand.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Navigating the AI Gold Rush and Historical Tech Bubbles | 2 | 3 | 1 | 1 | Host opens by posing the debate between AI as a generational breakthrough versus an overhyped bubble. Guest harmonizes both viewpoints using historical examples from the dot-com era and fiber optic buildouts. | |
| The Pitfalls of AI Investing and Logo Chasing | 6 | 5 | 4 | 5 | Host quotes a recent interview with Perplexity founder Aravind Srinivas to argue that foundation model valuations reflect talent value rather than model depreciation. Guest counters by citing Bill Gates Sr regarding assets that walk out at night. | |
| Hyperscalers, AI Agent Gateways, and Venture Viability | 5 | 6 | 5 | 5 | Host offers a structured hypothesis on hyperscaler consolidation and queries if foundation models are a venture-grade asset class. Guest rejects the framing, asserting that venture is sui generis rather than a traditional asset class. | |
| Applied AI & Israel's Unique Strengths | 4 | 3 | 2 | 3 | Host asks how to segment the AI market and probes vertical efficiency. Guest admits candidly that he does not know the full answer, but explains Israel's focus on applied AI over foundation models. | |
| Uncharted Territories and the Risks of "Peak SaaS" | 5 | 5 | 3 | 4 | Host expresses a strong preference for non-competitive markets. Guest agrees and reframes portfolio risk, arguing that traditional SaaS is actually higher risk than hard tech due to low barriers to entry and price erosion. | |
| The Slowdown of the SaaS Market and the Custom Software Era | 7 | 5 | 4 | 7 | Host cites public SaaS multiples (Klaviyo vs Atlassian) and challenges guest's claim that software buying is slowing down by quoting Accenture's $2.4B GenAI revenues. Guest explains why consulting revenue differs from software purchasing. | |
| Transitioning from "Price-Per-Seat" to "Selling Work" | 5 | 4 | 2 | 3 | Host quotes investor Mark Evans on tech adoption timelines to challenge short-term expectations. Guest discusses shifting pricing models from price-per-seat to selling outcome-based work. | |
| AI in Geopolitical Warfare and the China Threat | 6 | 6 | 5 | 5 | Host introduces Scale AI founder Alex Wang's thesis that AI models must remain closed for national security. Guest calls this view naive, highlighting geopolitical realities such as TikTok algorithms and chip export workarounds. | |
| The Reality of Defense Investing and Patriotism | 5 | 7 | 4 | 5 | Host raises an ethical critique regarding young tech investors getting excited about defense company margins and killing efficiency. Guest grounds the discussion in personal reality, noting his children's military service and the difficulty of defense sales. | |
| The Hard Tech Shift and 'Labless' Synthetic Biology | 4 | 5 | 2 | 3 | Host questions whether generalist software investors are equipped to underwrite hard tech and synthetic biology. Guest shares his self-directed approach to re-learning high school chemistry via ChatGPT and osmotic founder learning. | |
| The Liquidity Crisis, Lina Khan, and Going Public Early | 7 | 6 | 5 | 7 | Host challenges guest's recommendation that startups should go public earlier, asking if he would really tell a $200M ARR founder to go public now. Host also cites Mark Suster's thesis on PE buyers, which guest disputes. | |
| Conviction, Selling Winners, and the Pro-Rata Fallacy | 7 | 6 | 5 | 7 | Host directly confronts guest's strategy of liquidating 100% of a winning position, calling it a strange take and sharing his own portfolio data. Guest defends binary conviction investing over public-style dollar-cost averaging. | |
| Venture Fund Mortality, TVPI Reality, and Building True Moats | 5 | 6 | 3 | 4 | Host expresses surprise that LPs still prioritize paper TVPI given recent fund write-downs. Guest explains why TVPI remains meaningful if portfolio companies possess durable competitive moats. | |
| Self-Improvement and Directness as an Investor | 4 | 2 | 2 | 3 | Host asks guest where he needs to improve as an investor. Guest reflects on his extreme directness and impatience with long board meetings, while host asks whether top founders actually prefer that directness. | |
| Zoom vs. In-Person Meetings and Relationship Building | 3 | 3 | 2 | 2 | Host asks about meeting formats (Zoom vs in-person) and domain experience. Guest strongly favors in-person relationship building and net-new naive founders over industry insiders. | |
| Quick Fire: Generational Changes and the Israeli Ecosystem | 3 | 4 | 4 | 2 | Host runs a quick-fire round covering generational change in Israel, ecosystem misconceptions, and zero-interest rate environment sins. Guest provides sharp, contrarian responses. | |
| Where Aleph and Michael Will Be in 10 Years | 4 | 4 | 1 | 2 | Host breaks down the three pillars of venture (sourcing, selecting, servicing) and asks guest where he excels. Guest reveals he has completely stopped proactive deal sourcing, relying solely on referrals. |
Statements from this episode (49)
Eisenberg: Foundation models are the fastest depreciating asset in history
“Foundation models are the fastest depreciating asset in history.”
Eisenberg advises European tech founders to relocate due to regulations
“If I'm in Europe and I'm starting a company, I would get out.”
Eisenberg: FTC Chair Lina Khan threatens American capitalism and must go
“Lena Connell said openly is, I think, a threat to American capitalism. Whether Biden or Trump gets elected, she needs to go.”
Eisenberg: The IPO market is wide open for realistic valuations
“I don't buy that the IPO window is closed. The IPO window is wide open. The question is, what's the price you're willing to take?”
Eisenberg: Investors will lose money in AI, but foundational companies will emerge
“So a lot of people are going to lose a lot of money. During this, like a lot of money. And some foundational companies and technologies will be built.”
Eisenberg calls Perplexity his favorite app on his phone
“Perplexity is my favorite app. It's on the home screen of my phone”
Eisenberg: Financial reality dictates most AI foundation model companies will fail
“I don't buy that five, six, seven, eight, nine, 10 of these are going to work. It's just not true. Financially, it's not true.”
Eisenberg: Users will rely on AI agents over direct publisher websites
“As you kind of play this forward, you're going to see a whole series of agents. Like I'm not going to go to the New York times.com to read it. I'm going to go to my AI screen and say, fetch me the articles that interest me from the New York times today.”
Eisenberg: Venture capital is a craft, not an asset class
“The venture business is completely sui generis. It's not a class. There is a small number of people in this craft who make money, and there's a small number of startups.”
Eisenberg: Only one or two foundation model startups will prove profitable
“There is, there are one or two companies. In there that will make their venture investors a lot of money and the rest of them will cost their LPs and GPs a lot of money.”
Eisenberg: Israel lacks talent density for foundation models due to military focus
“We didn't think there was a talent density in Israel specifically for foundation models. Because a lot of the Israeli technology comes out of the military and not academia.”
Eisenberg: Incumbents will find AI catch-up harder than internet adoption
“And I think what's going to be shocking about this, that's not like the internet, is it's going to be harder for legacy companies to catch up to AI companies, which is different from the internet.”
Aleph invested in an AI and synthetic biology rocket fuel startup
“We just backed a AI to synthetic biology and chemistry rocket fuel company”
Eisenberg: SaaS is riskier than deep tech due to competition and price erosion
“I don't think it's a more risky play than a SaaS thing. I'll argue that SaaS is way more risky. One, because of the competition and price erosion.”
Eisenberg: Companies get premium multiples when they are sole plays on trends
“Companies get premium multiples when it's the only way to play a future trend in the public markets.”
Eisenberg: AI will disrupt horizontal SaaS processes
“I think in general, horizontal SaaS processes are going to be disrupted by AI.”
Eisenberg: Only 1 of Aleph's last 17 deals was horizontal SaaS
“Of the last 17 investments we made at Aleph, one was a SaaS company.”
Eisenberg: Companies will build custom software with AI instead of buying SaaS
“I think you'll see a lot of companies build their own software. Or you have tools that create kind of unique business processes software for a lot of these companies. I don't know if it's tomorrow or in five years, but I think there'll be more of that.”
Eisenberg: Software is McKinsey's fastest-growing business unit
“The fastest growing business inside of McKinsey is software.”
Eisenberg: Outcome-based pricing will rapidly enter the software market
“And I think the consultants like Accenture and McKinsey are the leading edge of this. But I think it's going to come into software very, very quickly.”
Eisenberg: Strict AI regulation will further erode Europe's economic competitiveness
“I think that is going to set Europe back in competitiveness more than it's already been set back in competitiveness over the last 20 years.”
Eisenberg: US, China, and Israel will lead AI over Europe due to lower regulation
“And so when you look at kind of the key markets for AI development, where the US is first and China is second and Israel is third I think those are likely to be less regulated than most, and certainly than Europe, and so these, those will be advantaged economi…”
Eisenberg: 21st century geopolitics will be defined by AI capability
“If the 20th century was about nuclear countries versus non-nuclear countries, the 21st century is AI countries versus non-AI countries.”
Eisenberg: US Nvidia chip bans have not meaningfully slowed China's AI
“The U.S. Put a banner on Nvidia chips going to China. I don't sense that it slowed China down in a meaningful way.”
Eisenberg: China is 10 years ahead of US in TikTok AI psyops
“I think they were 10 years ahead of the U.S. On AI and psyops using TikTok. I think they were 10 years ahead of the U.S. In sophistication of how to manipulate the minds of young people.”
Eisenberg: Anduril will scale defense contracting much faster than Palantir
“Palantir took 20 years to crack this code. Andrew will crack it much faster because of Palantir.”
Eisenberg: Investors will lose a lot of money chasing defense tech
“And so people will lose a lot of money chasing the new, new thing, which today happens to be defense in addition to AI.”
Eisenberg: Great hard tech founders will reject SaaS investors' capital
“Some founders, or the great founders, are likely not taking SaaS investors' money. Because the SaaS playbook, as it's become called, doesn't apply to any of these harder technologies.”
Eisenberg: Synthetic biology is shifting to a low-cost 'labless' model
“ZERP built out a huge amount of biological infrastructure. And so now you can just rent labs for cheap. And so I think we're moving to a labless model where you do kind of AI, you know, in your office, and then send out things for testing to labs and maybe eve…”
Eisenberg: Aleph Took Lemonade Public at $60M Revenue
“We took lemonade public with sixty million of revenue.”
Eisenberg predicts more private equity software investments will see Pluralsight-style zeroes
“I'm not concerned. I'm certain. It's happening.”
Eisenberg: Secondary market buyers claiming interest are mostly fishing for startup data
“Over the last three or four weeks, we've decided to do a little experiment. We reach out to them now. We used to ignore them. But now it's interesting because, like, you discover there are actual buyers there. There aren't, ok? There are not actual buyers ther…”
Eisenberg: My biggest ZERP mistake was not selling 100% of peak-valuation positions
“One of them was not selling a hundred percent of one or two positions when the prices were high.”
Eisenberg: Aleph's largest check ever went to AI defense startup Dream Security
“We wrote this giant check to a company called Dream Security, which is AI infrastructure for protecting critical assets in countries. It was the biggest check we wrote.”
Stebbings: Fund 1's top drivers were mediocre-seeming startups, while rocket ships failed fastest
“I look at my fund one, which is a tiny, yeah, eight million fund. Actually, our best value drivers today were all in the meh. And the worst performing companies were all in the rocket ship category. They're the ones that went to zero quickest.”
Eisenberg: A generation of ZERP-era venture funds will die
“A whole good generation of people grow up during ZERP. And, you know, now reality is biting. And so, yeah, of course.”
Eisenberg: VC funds with only one strong partner will fail first
“I think the first style of funds that goes out of businesses is funds where there's three, four or five partners and only one of them is good. And then that person gets tired of holding everybody on their shoulders, and they leave and set their own thing up, a…”
Eisenberg: VCs who did not take liquidity during ZERP blundered
“If you didn't take money off the table during, you know, the ZERP era, there's something wrong. Let's just be clear about that.”
Michael Eisenberg: Board meetings should take 45 minutes, never four hours
“I've come to the conclusion that with good preparation, most board meetings can be finished in 45 minutes. There's real governance that needs to go on there. There's core strategic issues that you need to kind of pound on. But most of them can be finished in 4…”
Michael Eisenberg: Board prep should use written letters, not slide decks
“It's generally a letter and not a presentation. It's rich in data, but it up front identifies these are my core one, two, or three strategic issues that I gotta get board input on. And, ah, you prep for it, you do calls beforehand, and then the conversation is…”
Michael Eisenberg Refused to Make Venture Investments Over Zoom During Pandemic
“You know, I didn't invest over Zoom. During, during the pandemic. I just, I met people out of a farm.”
Eisenberg prefers founders who are 'net new naive' over market experts
“Oh, net new naive. Net new naive. The fact that Shai and Daniel of Lemonade didn't know anything about insurance, like helped a ton.”
Fintech investors initially passed on Lemonade for lacking insurance experience
“When we first invested in Lemonade, I called a bunch of fintech investors and said, who's the insurance guy? I said, nobody. That's the beauty. So none of them would invest.”
Eisenberg: Aleph raised an insane amount of foreign capital post-October 7
“We've raised an incredible amount of foreign money into our company since October seventh, an insane amount of money. It's wide open for business right now. Investors are on the ground every day.”
Eisenberg: Israeli founders struggle to scale because of small domestic market
“People think Israelis know how to scale things, and for the most part, when you grow up in a small country, it's harder to think of scale, and so when you're a foreign investor, you need to help Israelis think in big scale.”
Eisenberg: Board members giving product advice is complete nonsense
“Board members giving product advice.”
Eisenberg: Western VC investment in China was a ZERP phenomenon
“I think China was a ZERP phenomenon also.”