The capital markets for AI infrastructure are hardening in a way that’s rewriting the power law of venture, making strategic allocation to compute and specialized talent paramount.
The Intake
📊 11 episodes across 6 podcasts
⏱ 676 minutes of intelligence analyzed
🎙 Featuring: Harry Stebbings, David Morehead, Anish Acharya, Lenny Rachitsky
|
The Big Shift
The power law of venture capital is becoming even more extreme, driven by the unique dynamics of frontier AI where capital directly compounds advantage. Unlike traditional startups where excessive funding can be a liability, for top-tier AI companies, capital translates directly into compute power, which directly fuels their competitive edge. This has led to an unprecedented acceleration in value creation; Aram Verdiyan, Partner at Accolade Partners, noted on The a16z Show that AI achieved $100 billion in revenue in just four years, a feat that took SaaS fifteen years.
"For the first time in my career, you can take capital and throw it at a company and it compounds their advantage. And this is a thing like how do you screw up a startup? Well, throw too much money at it..."
— David George, Host at a16z
This shift extends beyond software, as AI is now attacking every facet of GDP—transportation, labor, services, and capital coordination—representing a potential total addressable market 10x larger than traditional tech sectors. The strategic implication is clear: accessing and deploying capital effectively for compute and specialized talent is now the ultimate differentiator, further stratifying returns among venture funds. The race isn't just for innovation, but for the fundamental resources that enable it, redefining what it means to be a "core" asset in a portfolio.
The Rundown
① AI prototypes are stalling at 90% failure rate before production.
The primary hurdle for AI prototypes reaching production isn't model quality, but the absence of "harnesses" for durable execution, essential for managing AI agents in complex distributed environments (Samar Abbas, CEO at Temporal on This Week in Startups).
→ The signal: This reveals a critical infrastructure gap, indicating a new bottleneck for enterprises trying to operationalize AI, creating opportunities for platforms that offer reliability and visibility for AI agent workflows.
② LPs are prioritizing capital velocity over extended fund durations.
David Morehead, CIO at Baylor University, articulated a shift in LP perspective, emphasizing that "velocity of capital" and the absolute dollar pile for students are more critical than optimizing for higher multiples over longer holding periods, which GP marketing often prioritizes (on The Twenty Minute VC).
→ Why it matters: This suggests LPs are pushing for faster liquidity and shorter fund cycles, potentially reshaping fund structures and investment strategies to prioritize distributions for immediate impact, rather than just long-term paper gains.
③ Companies are fully delegating coding to AI engineers.
Eight Sleep's engineering team has "stopped coding," instead relying on hundreds of AI engineers who code for them, indicating a radical, immediate shift in software development paradigms (Matteo Franceschetti, Co-Founder at Eight Sleep on The Twenty Minute VC).
→ What to watch: This signals a fundamental reorganization of engineering teams, where human engineers become AI orchestrators, dramatically altering talent needs and the types of tools required for enterprise software development.
④ Open models are consuming the majority of enterprise tokens, not budget.
While open models are projected to consume 80-90% of enterprise tokens, they will only account for 10-20% of the budget due to their lower cost, challenging traditional assumptions about AI spend allocation (Jeffrey Morgan, Co-founder and CEO of Ollama 🆕 on Y Combinator Startup Podcast).
→ The signal: Enterprises will adopt a hybrid approach, using expensive frontier models for critical, hard tasks and open models for widespread, cost-sensitive applications, driving demand for efficient open-source infrastructure and tooling.
⑤ AI "doomer" warnings might be a strategic flex for valuation.
An Anthropic researcher's public resignation over AI safety concerns, coinciding with IPO preparations, prompted discussion among hosts that such warnings could inadvertently boost company valuation and perception of advanced capability (Kirsten Korosec, Transportation Editor at TechCrunch on Equity).
→ Why it matters: This suggests a sophisticated PR strategy within the AI ecosystem, where public safety concerns are leveraged to enhance a company's market position, making it harder for investors to discern genuine risk from strategic messaging.
Signal Board
🔥 HEATING UP
• Adaptive reuse of cannabis cultivation facilities for data centers 🆕: This novel concept is emerging as a viable solution for the data center compute crunch, leveraging existing infrastructure for rapid deployment (Jason Calacanis on This Week in Startups).
• Capital compounding advantage for frontier AI 🆕: The ability to directly translate capital into compute power for AI companies is creating an unprecedented competitive advantage, reshaping venture capital power laws (David George on The a16z Show).
• AI diffusion beyond coding and knowledge work 🆕: AI is increasingly impacting tasks across various industries (transportation, labor, services), signifying a vast new addressable market beyond traditional software applications (Aram Verdiyan on The a16z Show).
👀 ON WATCH
• Ollama 🆕: This company is gaining traction as a crucial "operating system" for open AI models in enterprise, standardizing deployment and driving widespread adoption (Jeffrey Morgan on Y Combinator Startup Podcast).
• Deepseek Flash model class 🆕: The emergence of ultra low-cost, efficient "Flash" LLMs is poised to enable "unlimited tokens" for enterprise use cases, changing the economics of AI deployment (Rahul on Y Combinator Startup Podcast).
• Coding Agents 🆕: AI agents that can generate code are becoming a significant driver of explosive token usage growth and a key focus for consumer AI opportunities (Anish Acharya on The a16z Show).
• AI safety warnings and doomer narrative 🆕: The public discussion around AI risks is intense, with some suggesting it's a strategic flex for AI companies, while others warn of its economic impact (Scott Galloway on Pivot).
🧊 COOLING OFF
• AI-driven 'permanent underclass' fear 🆕: This concern is viewed as misplaced by some, arguing that AI decentralizes opportunities and accelerates economic diffusion rather than concentrating power (Anish Acharya on The a16z Show).
• LP incentive misalignment in venture capital 🆕: The current LP structure incentivizes loss avoidance over maximum gains, creating a systemic bias against truly innovative, high-risk venture investments like those in AI (Jen Kha on The a16z Show).
The Debate
This week, the venture ecosystem debated the sincerity and strategic implications of AI "doomer" warnings from leading frontier labs.
🐂 The bull case: Some argue that warnings from top AI executives are genuine, reflecting legitimate concerns about the technology's risks, especially given the accelerating pace of development. Scott Galloway, Professor of Marketing at NYU Stern, while critical of the lack of regulation, acknowledged that top executives might be genuinely concerned about AI's potential societal impact on Pivot.
🐻 The bear case: A counter-narrative suggests these warnings are a strategic maneuver. Kirsten Korosec, Transportation Editor at TechCrunch, questioned on Equity if it's "a weird way of flexing to show how far advanced their company's, like, AI model," inadvertently boosting valuation. Similarly, Kara Swisher, Host at New York Magazine, posited on Pivot that "The motivation to slow down is not altruistic... they're trying to protect their businesses so they can dominate and nobody else can."
Our read: While genuine concern undoubtedly exists, the timing and impact of these public warnings on market perception and competitive dynamics suggest a strong strategic component. Both can be true, but the latter is increasingly difficult to ignore.
The Bottom Line
The new VC power law is all about capital access for compute and specific AI talent, making every investment a strategic play in an increasingly concentrated market.
Episode Guide
1. This Week in Startups — "90% of AI prototypes never reach production (w/ Temporal's Samar Abbas) | AI Basics"
Runtime: 19 min | Host: Jason Calacanis | Guest: Samar Abbas (CEO, Temporal)
For the Builder: This episode provides critical insight into the often-overlooked challenges of moving AI prototypes into production, focusing on the need for robust infrastructure.
Samar Abbas, CEO of Temporal, unpacks why most AI prototypes fail to scale, pinpointing a lack of "harnesses" for durable execution rather than model quality, highlighting Temporal's role in providing reliability and visibility for AI agents.
"I have an awesome idea, I got started and built an application through a coding agent. But like 90% of those ideas die after a POC. Essentially they never see light of the day." — Samar Abbas, CEO at Temporal
2. The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch — "20VC: How LPs Allocate to Venture in 2026: What They Want, What They Do Not Want | Why Fund Multiple Does Not Matter Without a Timeline | Why Velocity of Cashback is the Most Important Thing with David Morehead, CIO @ Baylor"
Runtime: 69 min | Host: Harry Stebbings | Guest: David Morehead (Chief Investment Officer, Baylor University Office of Investments)
For the LP & GP: This is a must-listen for understanding the evolving priorities of sophisticated LPs like university endowments, especially regarding capital velocity and market timing.
David Morehead, Baylor University CIO, reveals their endowment's focus on capital velocity over mere multiples, tactical software investments during downturns, and a unique approach to hiring and mission-driven LP engagement, challenging conventional VC metrics.
"What we're really after is the velocity of capital, not just returns on capital. There's a rule in our office that you're not allowed to talk about returns without also talking about time." — David Morehead, Chief Investment Officer at Baylor University Office of Investments
3. The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch — "20VC: 7 Predictions for How AI Changes the World: Labour, Engineering, Social Media, GrokBots Buying Cybercabs and more with Matteo Franceschetti, Co-Founder @ Eight Sleep"
Runtime: 76 min | Host: Harry Stebbings | Guest: Matteo Franceschetti (Co-Founder, Eight Sleep)
For the Growth CEO: An essential listen for founders and executives navigating the rapid integration of AI into product development, marketing, and organizational structure.
Matteo Franceschetti of Eight Sleep shares how his engineering team "stopped coding" due to AI engineers, how his company now models CAC with proprietary AI, and predicts successful AI companies will evolve into diversified holdings akin to Chinese conglomerates.
"Our engineers stopped coding around a year ago. What they have is hundreds of AI engineers that they code for them." — Matteo Franceschetti, Co-Founder at Eight Sleep
4. Y Combinator Startup Podcast — "Open Models Change The Economics of AI"
Runtime: 57 min | Host: Y Combinator | Guest: Jeffrey Morgan (Co-founder and CEO, Ollama), Rahul (Co-founder, Ollama), CEO of Ollama (CEO, Ollama)
For the Product Leader: This episode offers a deep dive into how open models are redefining enterprise AI adoption, focusing on cost, customization, and deployment challenges.
Jeffrey Morgan, CEO of Ollama, discusses the enterprise shift to open AI models driven by cost and control, the surge in token usage, and the rise of efficient "Flash" models, highlighting Ollama's role as an operating system for these distributed AI architectures.
"Cost is by far the largest pain point that open models can jump in and solve. But every business has a vision of getting better control over AI and customizing it for their business, and that's really their North Star." — Jeffrey Morgan, Co-founder and CEO of Ollama
5. Equity — "An ex-Anthropic researcher’s doomsday warning comes at a very interesting time"
Runtime: 40 min | Host: Kirsten Korosec, Anthony Ha, Sean O'Kane | Guest: Host-led discussion
For the Board Member: This offers a nuanced perspective on AI safety warnings, distinguishing between genuine concerns and potential strategic messaging in the competitive AI landscape.
Hosts discuss an Anthropic researcher's AI safety resignation and Apple's AI-first strategy, analyzing whether AI doomer warnings are genuine or a strategic move to flex AI capabilities and influence valuation.
"Is this a weird way of flexing to show how far advanced their company's, like, AI model? I mean, that sounds very cynical, but it does achieve that purpose." — Kirsten Korosec, Transportation Editor at TechCrunch
6. This Week in Startups — "Ask Jason: AI Extinction, Grok's Rogue Meme Coin & Weed Farms as Data Centers | E2337"
Runtime: 71 min | Host: Jason Calacanis | Guest: Jim Pfaff (Conservative Activist, Midterm Republican Convention), Michael Lee (Founder, Gondola Partners), Lyda Liber Polu (Caller), John Wright (CEO and Co-founder, Nuviz), Ray (Founder, Vortex Consilium), David Rosenberg (Student, NYU), David (Student), Mo (Ex-cannabis industry professional)
For the Strategic Investor: Provides a candid look at the immediate and long-term implications of AI, from liability and regulation to innovative infrastructure solutions.
Jason Calacanis tackles AI extinction fears, liability for AI-created financial products, and innovative data center solutions like repurposing cannabis farms, while also criticizing OpenAI/Anthropic's communication strategies around AI risks.
"I'm going to put blame on the platform that doesn't have safeguards for human in the loop. This is where regulation matters... at least two people, maybe three, need to put their driver's license in." — Jason Calacanis, Host of This Week in Startups
7. The a16z Show — "Why Companies Are Becoming a Series of Loops | Anish Acharya on Lenny’s Podcast"
Runtime: 78 min | Host: Lenny Rachitsky | Guest: Anish Acharya (General Partner, a16z)
For the CEO & Founder: This episode delivers a forward-looking framework for reorganizing companies around AI, focusing on where human judgment remains critical and new opportunities in consumer AI.
Anish Acharya, GP at a16z, introduces the "loops" concept for AI-human collaboration, dispels fears of an AI-driven "permanent underclass," and emphasizes consumer AI's potential for life enrichment over mere productivity, highlighting specific market opportunities.
"The loop will help you climb to the local maxima, but then it plateaus. You need human intuition. You need somebody to actually help you land at the base of the next hill." — Anish Acharya, General Partner at a16z
8. Pivot — "AI Panic: Dario’s Warning, Trump’s Dismissal, and OpenAI’s IPO Delay"
Runtime: 58 min | Host: Kara Swisher, Scott Galloway | Guest: Host-led discussion
For the Policy Maker: A critical examination of the current state of AI regulation, contrasting US inaction with global efforts and proposing concrete regulatory frameworks.
Kara Swisher and Scott Galloway dissect the lack of AI regulation in the US, the political response to AI risks, and OpenAI's delayed IPO versus Anthropic's plans, suggesting regulatory models and highlighting economic fragilities linked to AI over-investment.
"Right now, America is engaging in an experiment. It's the following: We have never in the history of the modern economy had a trillion dollar industry that is totally unregulated." — Scott Galloway, Professor of Marketing at New York University Stern School of Business
9. The a16z Show — "How AI Is Rewriting the Power Law of Venture Capital"
Runtime: 49 min | Host: Jen Kha, David George | Guest: Aram Verdiyan (Partner, Accolade Partners)
For the GP: This is a definitive discussion on how AI is fundamentally reshaping venture capital's power law, requiring a recalibration of investment theses and portfolio construction.
David George of a16z and Aram Verdiyan of Accolade Partners argue AI is intensifying VC's power law, as capital directly compounds advantage for frontier AI. They discuss AI's rapid revenue growth, its vast addressable market beyond software, and the critical importance of LP portfolio construction.
"AI is attacking every facet of the gdp. Transportation, labor, services, capital coordination. There hasn't been a technology paradigm that hits on 30 trillion in GDP at the same time." — Aram Verdiyan, Partner at Accolade Partners
10. This Week in Startups — "Did OpenAI Steal the Navier-Stokes Solution? | E2335"
Runtime: 81 min | Host: Jason Calacanis, Lon | Guest: Yohei Nakajima (General Partner, Untapped Capital), Ben Lerer (Co-founder, Lerer Hippeau), Rebecca Lynn (Co-founder, Canvas Prime), Jason (Host), Ben (Guest), Yohi (Guest), Rebecca (Investor)
For the Founder: Essential for understanding the IP risks associated with frontier AI labs and the critical need for data privacy and ethical development.
Jason Calacanis and guests debate AI doomerism and warn founders against sharing proprietary data with frontier AI labs due to IP theft risks, discussing the OpenAI-Navier-Stokes controversy and the inherent trust issues with large tech platforms.
"I would not trust them with anything that's important or proprietary. This idea that you're going to give instinct or Zuckerberg your Gmail, then your notion, then your docs, then your databases means they're going to steal your ip, train it, and then give it to your competitors." — Jason Calacanis, Host
11. The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch — "20VC: Jensen Huang Declares AGI Has Arrived | GPT Astra and Fable 5.1 Accelerate the Model Race | Tesla Launches Cybercabs | Index Pulls Out of Town & Anthropic Pulls From Descartes Acquisition"
Runtime: 78 min | Host: Harry Stebbings | Guest: Rory O'Driscoll (Partner, Scale Venture Partners), Jason Lemkin (Founder, SaaStr), Jason (Guest, The Twenty Minute VC), Rory (Co-host/Commentator, The Twenty Minute VC)
For the Operator: Provides direct insights into the impact of generative AI on coding and legal industries, highlighting the need for human judgment even as AI capabilities accelerate.
Harry Stebbings, Rory O'Driscoll, and Jason Lemkin discuss rule-breaking AI agents, the impact of AGI on the legal sector, and the acceleration of the model race with GPT Astra and Fable 5.1, emphasizing human judgment's continued role amidst rapid AI advancements.
"The only thing that mattered for the last two years is LLMs do code. And code is a half a trillion dollar industry. Focus, people." — Jason Lemkin, Founder of SaaStr
