By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.

The Twenty Minute VC

Michael Eisenberg: How China Could Overtake the US in the AI Race | E1167

Michael Eisenberg is a Co-Founder and General Partner @ Aleph, one of Israel’s leading venture firms with a portfolio including the likes of Wix, Lemonade, Empathy, Honeybook and more.

Michael EisenbergguestHarry Stebbingshost

Jun 19, 20241h 2mWatch on YouTube ↗

Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.

CHAPTERS

Jump to publisher timestamps

Chapter summaries

AI is both a revolution and a bubble: why both can be true
Michael argues that AI/LLMs are the most transformative technology he expects to see, while also acknowledging the current gold-rush dynamics and bubble-like financing. He draws parallels to the dot-com and fiber-optic booms that produced lasting infrastructure despite many losses.
- •Holding two truths: transformational tech + speculative mania
- •Dot-com and fiber-optic bubbles as historical analogs
- •Bubbles can fund infrastructure that enables long-term winners
- •Many investors will lose money; a few foundational companies will endure
Who gets hurt in an AI hype cycle: LP crowding, logo-chasing, and copycat geography bets
He outlines the non-obvious ways capital gets misallocated in hype cycles. LPs may think they’re diversified across funds but are concentrated in the same AI names; VCs chase hot logos; others overpay to clone models across geographies that won’t scale similarly.
- •Hidden concentration risk across LP portfolios
- •"Logo chasing" into the hottest AI companies
- •Overpaying to replicate a winning model in the wrong geography
- •Foundation models as "fastest appreciating asset"—but fragile moats
Foundation models as talent businesses—and why that’s not enough
Harry suggests model value may be driven by teams; Michael counters that talent can walk out the door, echoing the 'assets walk out at night' critique. He expects consolidation/pick-apart outcomes and doubts that many foundation-model companies will be durable financial successes.
- •Talent concentration doesn’t equal durable assets
- •Risks of model commoditization and team mobility
- •More "elegant" acquihires/pick-apart deals likely
- •Only 1–2 foundation model companies may generate venture-scale returns
Where value accrues in the AI stack: the new 'start page,' agents, and API layers
Rather than focusing only on who wins the foundation model race, Michael focuses on where users begin their day and how agents will mediate content, tools, and workflows. He anticipates an agent interface layer plus an underlying API layer integrated into hyperscaler stacks.
- •User entry point matters: the new AI 'home screen'
- •Agents change how people consume media and information
- •API-first model layers (e.g., Anthropic positioning)
- •AI becomes embedded in hyperscaler infrastructure
Applied AI vs. foundation models: Israel’s edge, academia vs. military, and 'AI companies vs non-AI companies'
Michael distinguishes the broader AI era from the recent LLM moment and explains Aleph’s focus on applied AI rather than foundation models. He argues legacy companies will struggle to catch up if their data foundations are poor, making AI-native companies structurally advantaged.
- •AI isn’t new; LLMs are the recent step-change
- •Israel: stronger in applied AI than foundation models (talent source matters)
- •Data readiness will determine who can adopt AI effectively
- •Legacy catch-up may be harder than in the internet era
Investing in competitive markets: execution beats competition, but uniqueness earns premiums
They debate competitive markets; Michael says competition rarely kills companies—bad execution does—yet he prefers markets requiring education and unique domain insight. He explains why public markets award premium multiples to the "only way to play" a future trend (e.g., NVIDIA).
- •Competition vs. execution as the true failure mode
- •Preference for market education and hard-to-copy domain knowledge
- •Examples beyond SaaS (synthetic rocket fuel)
- •Premium multiples come from uniqueness: "only way to play" a trend
Peak SaaS and pricing model disruption: from seat-based fees to paying for outcomes
Michael argues SaaS is riskier than commonly assumed due to competition and price erosion, and he believes horizontal SaaS will be disrupted by AI. He predicts a shift toward value-based pricing—buyers paying for work/outcomes—highlighting consultants’ success integrating AI for enterprises.
- •"Peak SaaS": growth slowing and markets saturating
- •AI enables more bespoke internal software creation
- •Enterprises often lack AI integration know-how (Accenture/McKinsey)
- •Move from price-per-seat to paying for outcomes/value delivered
Adoption cycles and regulation: fast micro adoption, slow macro deployment—especially in Europe
Michael agrees both that AI progress is rapid and that economy-wide adoption will take longer than people expect. Regulation and compliance-heavy sectors (finance, pharma) will slow deployment, with Europe likely to regulate more heavily and risk falling behind the US, China, and Israel.
- •Overestimate 1-year impact, underestimate 10-year impact
- •Visible efficiency gains already in customer service/fintech
- •Regulated industries will constrain deployment speed
- •Europe most likely to over-regulate; competitiveness implications
AI as geopolitical power and defense technology: openness, chips, and China’s real capabilities
The conversation shifts to AI’s role in national defense and warfighting, including "closing the kill chain." Michael doubts AI can be kept closed from adversaries and argues the US must remain technologically ahead, warning against underestimating China’s sophistication (e.g., TikTok psyops).
- •AI will be a decisive military advantage (kill-chain automation)
- •Skepticism that export controls/closed systems can truly contain AI
- •Need to assume adversaries are ahead to stay competitive
- •China’s algorithmic influence operations as a warning signal
The ethics and reality of defense investing: patriotism, maturity, and the difficulty of selling to governments
Harry challenges the MBA-style enthusiasm around defense margins; Michael responds from lived experience of war and argues weapons and defense tech carry moral responsibility. He also emphasizes defense is hard to sell—regulations, procurement, long cycles—and predicts money will be lost chasing defense as a trend.
- •Moral weight of defense: responsibility and values
- •Patriotism and the historical tech–defense alliance
- •Defense procurement is a specialized craft (Palantir/Anduril examples)
- •Many investors will lose money treating defense like a fast SaaS playbook
From SaaS to hard tech: 'labless' biology, new infrastructure, and how investors adapt
Michael describes a middle ground between pure SaaS and extreme hard tech: semiconductors, synthetic biology/chemistry, and lab outsourcing enabled by prior infrastructure buildouts. He argues the classic SaaS playbook won’t translate and that great founders may avoid SaaS-only investors.
- •Hard tech shift risks: software investors may be out of depth
- •"Labless" model: outsource labs/manufacturing using excess capacity
- •AI accelerates experimentation and iteration in bio/chem
- •Founder–investor fit: why non-software founders may avoid SaaS investors
Liquidity in venture: M&A headwinds, IPOs are open (at the right price), and the mirage of secondaries
They discuss exit conditions: Michael criticizes Lina Khan’s approach to antitrust as harmful to M&A and argues most startups never reach liquidity anyway. He insists IPO markets are open if companies accept realistic pricing and warns many secondary 'buyers' are simply fishing for data.
- •M&A constrained by antitrust posture; implications for venture returns
- •IPO window is open—pricing expectations are the blocker
- •Go public earlier and build in public markets (Reddit/Lemonade examples)
- •Secondaries often lack real buyers; beware information-harvesting
Selling strategy and portfolio construction: why he’ll sell 100% and rarely does pro rata
Michael explains his conviction-driven approach: venture is binary, so act decisively when you have an information advantage, including selling out completely when returns are compelling. He describes Aleph’s ownership philosophy—buy in seed/A, use big initial checks, and limit follow-ons.
- •Biggest ZIRP mistake: not selling enough at peaks
- •Venture as a binary/outlier business—avoid incremental averaging
- •Sell 100% when you have unique insight and conviction
- •Rare pro rata: secure ownership early with larger initial checks
Sustainable growth, moats, and fund survival: TVPI vs DPI, competitive advantage, and the new normal
Michael argues TVPI still matters but must be evaluated through the lens of moat depth and sustainability of cash flows, especially as AI disrupts existing SaaS portfolios. He discusses the shift from hypergrowth to durable 30–40% growth, outcomes for 10–20% growers, and why many funds will disappear.
- •Moats drive duration of cash flows; multiples reflect durability
- •AI may undermine many overvalued SaaS portfolios
- •30–40% growth is excellent; 10–20% growth can trap venture-backed companies
- •Fund shakeout: teams with one strong partner and weak bench are vulnerable
Quick-fire and personal operating style: directness, better boards, and time trade-offs
In the closing segment, Michael reflects on being very direct, his impatience with long board meetings, and preference for in-person interaction over Zoom. He shares beliefs about first-time founders, skepticism of expertise, networking as a key advantage, and the personal cost of time allocation.
- •Direct communication as an "acquired taste"—and why he doesn’t soften it
- •Board meetings should be letter-driven, data-rich, and focused (often 45 minutes)
- •In-person beats Zoom for decision-making and reducing pontification
- •Work-life 'balance' reframed as how others experience your time choices

Publisher timestamps

Get more out of YouTube videos.

Read along on your phone, search it beside the video, or ask your favorite AI app what was said.