The smartest money is betting on physical AI and infrastructure, even if it means building data centers or moving to orbit.
📊 11 episodes across 8 podcasts
⏱ 652 minutes of intelligence analyzed
🎙 Featuring: Jason Calacanis, Cliff Weitzman, Harry Stebbings
|
The Big Shift
The next frontier for venture capital is moving beyond pure software. While AI is undeniably the driving force, smart money is now flowing into its physical manifestations and underlying infrastructure, often requiring a contrarian playbook for investment and execution.
This goes beyond just chips or cloud. We're seeing surges in "Physical AI" like robotics, orbital data centers, and even weather modification, alongside a strategic shift toward startups owning their compute infrastructure to gain a competitive edge.
"I really had to go to bat even internally to talk about, you know, suddenly I had to become a thermal engineer and try to understand how space dissipates heat and the fact that it doesn't and like how, why it's better to do."
— Caitlin Retz, Partner at 776 on This Week in Startups
This willingness to dive deep into complex, capital-intensive problems—like thermal engineering in space for orbital data centers (StarCloud) or directly purchasing and managing hundreds of Nvidia GPUs for AI training—signals a recognition that the bottleneck for truly transformative AI isn't just algorithms, but the physical constraints and infrastructure required to run them.
The Move: Investors must broaden their aperture beyond pure SaaS metrics, evaluating companies on their ability to command physical assets, secure critical compute, and tackle "infrastructure for the inevitable" (Paige Doherty on This Week in Startups). For founders, this means reassessing the build-vs-buy decision for core infrastructure, as owning physical assets like GPUs can unlock significant cost efficiencies and competitive advantages (Cliff Weitzman on The Twenty Minute VC).
The Rundown
① Physical AI funding is exploding, defying prior benchmarks.
Investment in Physical AI reached nearly $50 billion in H1 2026, significantly outpacing the combined totals of 2022, 2023, and 2024, representing an 80% increase from 2025. (Paige Doherty on This Week in Startups)
→ The signal: This capital surge indicates a major shift beyond pure software plays, identifying physical manifestations of AI as the next frontier for outsized returns and challenging traditional VC deployment models.
② Owning your AI compute is the new moat.
Speechify CEO Cliff Weitzman detailed their strategy of purchasing Nvidia GPUs (H100s, Rubens) directly, arguing it's significantly more cost-effective than renting and provides a crucial competitive advantage in AI development due to faster access and co-located memory needs. (Cliff Weitzman on The Twenty Minute VC)
→ Why it matters: Founders should re-evaluate cloud-only strategies; for capital-intensive AI companies, direct ownership of compute is becoming a key differentiator, influencing valuation and long-term cost structures. Cliff Weitzman noted that if he rented an H100 GPU, it would cost him "$35,000 to $50,000 to rent that GPU for one year. But I could buy it for $30,000."
③ AI model evaluation needs independent, evolving benchmarks to cut through hype.
VALS founder Rayan Krishnan highlighted that public AI benchmarks are saturated, with models optimizing for tests, leading to a disconnect between self-reported scores and actual performance on private benchmarks. (Rayan Krishnan on The a16z Show)
→ What to watch: The market will increasingly demand third-party, high-signal evaluations for AI models to inform purchasing decisions, manage escalating token spend, and align models with real-world intent, akin to how financial audits provide transparency.
④ Consumer health is shifting from episodic treatment to proactive, data-driven self-management.
Dr. Mark Hyman introduced Function Health, which democratizes access to comprehensive biological data through deep lab testing and AI, enabling proactive health management rather than reactive disease care. (Dr. Mark Hyman on This Week in Startups)
→ The signal: The direct-to-consumer health tech market is primed for disruption, empowering individuals with their own biometric data and shifting focus to preventive, personalized care that bypasses traditional medical gatekeepers.
⑤ Scaling to $200M ARR in a competitive market requires hyper-realistic execution and brand beyond celebrity.
Danny Yeung, co-founder of IM8 (a supplement brand co-founded with David Beckham), achieved a $200M run rate in 18 months by focusing on scientific rigor, product taste, aggressive Meta ads testing (2-3k live ads), and leveraging offline events to build trust, distinguishing it from typical celebrity endorsements. (Danny Yeung on My First Million)
→ Why it matters: Even in crowded D2C categories, rapid scale is achievable through relentless product quality, continuous marketing experimentation, and building authentic brand equity that transcends celebrity associations.
Signal Board
🔥 Heating Up
• Physical AI funding surge: Investment in Physical AI reached $47.4B across 521 deals in H1 2026, significantly outperforming prior years, signaling a major shift in capital allocation (Paige Doherty on This Week in Startups).
• Orbital data centers: Companies like StarCloud are raising significant capital ($250M at $2.3B valuation with NVIDIA/Cisco investment), validating the non-consensus bet on space-based compute infrastructure (Caitlin Retz on This Week in Startups).
• Independent, continuously evolving evaluations for AI models: Crucial for understanding true model capabilities and managing token spend, as public benchmarks are becoming obsolete (Rayan Krishnan on The a16z Show).
👀 On Watch
• 🆕 Town: An AI assistant company competing with tech giants by focusing on deep product-market fit and agent-to-agent network effects, not immediately building moats (Jean-Denis Grèze on The Twenty Minute VC).
• 🆕 Supercool Earth: Developing natural protein-based cloud seeding technology, aiming to make it rain/snow on demand, addressing water scarcity with early customer traction (Dacia Leon on The Pitch).
• 🆕 Function Health: Democratizing access to comprehensive biological data through deep lab testing and AI, shifting healthcare from reactive to proactive, data-driven self-management (Dr. Mark Hyman on This Week in Startups).
• 🆕 Nvidia GPUs: Direct purchase and ownership of GPUs by AI startups like Speechify is gaining traction as a cost-effective alternative to cloud rentals, conferring competitive advantage (Cliff Weitzman on The Twenty Minute VC).
🧊 Cooling Off
• Public benchmarks for AI models are saturating/obsolete: Models are optimizing for these tests, creating a disconnect between reported capabilities and real-world performance (Rayan Krishnan on The a16z Show).
• Deferred Interest Credit Cards ('Fake' 0% loans): Affirm's "true 0% APR" model highlights the deceptive nature of common credit products, pushing for transparency (Max Levchin on The a16z Show).
• Traditional doctors' diagnostic biases: A Harvard study found LLMs to be as or more accurate than doctors in comparable situations, indicating a shift in diagnostic efficacy (Dr. Mark Hyman on This Week in Startups).
The Debate
The conversation around AI's societal impact reveals a tension between utopian optimism and dystopian fears.
🐂 The bull case: Anish Acharya (General Partner, Andreessen Horowitz) argues that the fear of AI creating a "permanent underclass" is a "funny dark fantasy" of Silicon Valley. He believes technology fundamentally amplifies human agency and distributes opportunities, making the future more optimistic than commonly portrayed (Anish Acharya on Lenny's Podcast).
🐻 The bear case: While not explicitly stated as a "bear case," concerns about AI's potential for misuse are pervasive. Ben Horowitz (General Partner, Andreessen Horowitz) discussed the government's inclination to fear AI's capabilities in areas like biohacking and cyber hacking, necessitating independent evaluation to verify what models can and cannot do (Ben Horowitz on The a16z Show).
Our read: The market consensus leans toward AI amplifying agency, but serious investors and policymakers are actively working to mitigate existential risks and ensure responsible deployment.
The Bottom Line
The smartest money is moving into AI's physical and infrastructural layers, demanding founders own their stack, and venture capitalists embrace contrarian, capital-intensive bets to capture the next wave of value.
Episode Guide (Web Version)
1. This Week in Startups — "VC experts on why Physical AI funding is heating up | E2333"
Runtime: 80 min | Host: Jason Calacanis | Guest: Paige Doherty (Founding Partner, Behind Genius Ventures), Katelin Holloway (Founding Partner, Seven Seven Six), Caitlin Retz (Partner, 776), Caitlin Bolnick Rulien (Partner, 776 Management)
Who should listen: VCs and founders interested in the next wave of capital-intensive AI, particularly in robotics and space tech.
This episode details the explosive growth in Physical AI funding, exploring ventures from orbital data centers to military AI applications. The discussion highlights unconventional investment strategies and the increasing necessity for VCs to understand complex engineering challenges.
"I really do think that, you know, if I, if I had $100, I would much rather back a brilliant quant team in Zurich at the same, you know, at like a normal, rational price. Than like the 40th agent wrapper ripping here in Selma." — Katelin Holloway, Founding Partner at Seven Seven Six
2. Venture Unlocked: The playbook for venture capital managers — "Wins Above Replacement: The New Way to Judge Founders"
Runtime: 55 min | Host: Samir Kaji | Guest: Carter Reum (Co-Founder and Managing Partner, M13), Latif Peracha (Partner, M13), Carter (Guest, Allocate)
Who should listen: VC partners and LPs seeking to refine founder evaluation and portfolio strategy in competitive markets.
M13's Carter Reum discusses reframing risk with probability-adjusted outcomes and their "wins above replacement" framework for founder-investor fit. The episode critiques traditional VC approaches and emphasizes strategic capital deployment in the AI era.
"We didn't want to create a better VC. We wanted to create a different VC." — Carter Reum, Co-Founder and Managing Partner at M13
3. The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch — "20VC: How to Build Your Own Data Center & Why Every Startup Should Do It | How ElevenLabs Leapfrogged Us: What I Learned | The AI Talent War: How Your Hiring Process Needs to Change with Cliff Weitzman, Speechify"
Runtime: 65 min | Host: Harry Stebbings, 20VC | Guest: Cliff Weitzman (Co-founder and CEO, Speechify)
Who should listen: AI founders and CTOs considering infrastructure ownership and strategic talent acquisition in the AI talent war.
Cliff Weitzman of Speechify makes a compelling case for directly purchasing Nvidia GPUs for AI training and inference, highlighting cost-effectiveness and competitive advantage. He also shares candid lessons learned from strategic mistakes and insights into hiring for raw technical aptitude.
"If I rented it [H100 GPU] from like, you know, Azure or AWS, maybe it'll cost me $3.50 per hour. So if I multiply that times 24 hours and then times 365 days in a year, I'm actually going to end up paying 35,000 to $50,000 to rent that GPU for one year. But I could buy it for $30,000." — Cliff Weitzman, CEO of Speechify
4. My First Million — "How we scaled to $100M in under a year (ft. IM8 co-founder Danny Yeung)"
Runtime: 54 min | Host: Shaan Puri | Guest: Danny Yeung (Co-founder, IM8)
Who should listen: Founders of consumer brands and DTC operators looking for rapid scaling strategies and brand building insights.
Danny Yeung shares the journey of IM8, a supplement brand co-founded with David Beckham, which hit a $200M run rate in 18 months. The episode delves into aggressive marketing, scientific backing, and building a brand that transcends celebrity.
"We had 100 million annual revenue run rate at 11 months... now we've been the fastest growing supplement brand ever recorded." — Danny Yeung, Co-founder of IM8
5. Equity — "Apple's Ternus era begins as Nvidia bets on the whole AI stack"
Runtime: 39 min | Host: Kirsten Korosec, Sean O'Kane
Who should listen: Executives and investors following big tech leadership transitions and the strategic implications for AI and competitive landscapes.
Kirsten Korosec and Sean O'Kane discuss John Ternus's ascendance at Apple and speculate on the company's AI strategy. They also touch on Nvidia's full-stack AI bets, Andreessen Horowitz's new $1.1 billion Machine Age Fund, and the potential pivot of GoPro to AI infrastructure.
"What really matters here is how Apple is going to lead or at least keep up. Yeah, I actually think he's better set up talking about having a good first step next week with a hardware event. I think he's almost better set up to make progress on the software, if only because of how, especially on the AI side and Siri in particular, how badly Apple has lagged on that stuff over the last decade." — Sean O'Kane, Host, Senior Reporter, Special Projects at TechCrunch
6. The Pitch — "#186 Supercool Earth: Making It Rain, Literally"
Runtime: 50 min | Host: Josh Muccio, Lisa Muccio | Guest: Dacia Leon (CEO and Co-founder, Supercool Earth), Mike Ma (Investor, Sidecut Ventures), Rohit Gupta (Investor, Future Communities Capital), Michelle Kwok (Investor, Draper Associates), Jesse Middleton (Investor, Flybridge), Elizabeth Yin (Investor, Hustle Fund), Dacia (Founder, Supercool Earth), Mike (Investor)
Who should listen: Deep tech founders and impact investors interested in climate tech, especially in water management and atmospheric science.
Dacia Leon of Supercool Earth pitches her natural protein-based cloud seeding technology designed to make rain and snow. The episode covers the scientific basis, market opportunity, and the challenges of fundraising for deep tech in climate solutions.
"At Supercool Earth, we are building a natural product that is two times as effective as the traditional cloud seeding chemical that's used today. Which means we can get two times more water out of clouds." — Dacia Leon, CEO and Co-founder of Supercool Earth
7. The a16z Show — "The $100B Niches Hiding Inside Payments"
Runtime: 61 min | Host: Erik Torenberg | Guest: Alex Rampell (General Partner, a16z), Max Levchin (Co-Founder and CEO, Affirm), Max Alex (Guest)
Who should listen: Fintech founders, investors, and product managers interested in payments innovation, credit, and agentic commerce.
Max Levchin, co-founder of Affirm and PayPal, discusses the surprising durability of the credit card interface and the consistent multi-billion dollar opportunities in payment niches. He delves into Affirm's evolution and the strategic advantage of "true 0% APR" loans.
"There are no niches and payments that are smaller than a hundred billion dollars." — Max Levchin, Co-Founder and CEO of Affirm
8. The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch — "20VC: The $100 Billion AI Assistant Race: Town vs Instinct vs GrokBot | We Spend $75K Per Engineer on AI Tools | Why the AI Assistant Market Is Not a Bubble & AI Assistants Will Replace Every App on Your Phone with JD, Founder of Town"
Runtime: 71 min | Host: Harry Stebbings | Guest: Jean-Denis Grèze (Co-Founder and CEO, Town), JD (Founder, Town)
Who should listen: AI product leaders and investors tracking the AI assistant market, cost optimization, and network effect strategies.
Jean-Denis Grèze of Town discusses the competitive AI assistant market, focusing on product-market fit and agent-to-agent network effects. He highlights the economic challenges of AI development and dynamic model routing for cost-effectiveness.
"I think the product in this category that will win will have a network effect at the agent level." — Jean-Denis Grèze, Co-Founder and CEO of Town
9. This Week in Startups — "Dr. Mark Hyman on Function Health & GLP-1 microdosing | E2334"
Runtime: 58 min | Host: Jason Calacanis | Guest: Dr. Mark Hyman (Founder/Doctor, Function Health), Jason (Host, This Week in Startups)
Who should listen: Health tech founders, medical professionals, and consumers interested in preventative health and personalized medicine.
Dr. Mark Hyman introduces Function Health as a solution to democratize healthcare by providing patients with access to their own biological data. He advocates for moving from disease-based care to proactive, data-driven health management, integrating AI and wearables.
"The membership is basically a dollar a day. Now I've done these labs for years and years on patients and the retail price on these is like $15,000." — Dr. Mark Hyman, Founder/Doctor at Function Health
10. The a16z Show — "Who Grades the AI Models? | Ben Horowitz & Rayan Krishnan"
Runtime: 40 min | Host: Jennifer Li, Erik Torenberg | Guest: Rayan Krishnan (Founder and CEO, VALS), Ben Horowitz (General Partner, Andreessen Horowitz)
Who should listen: AI developers, policymakers, and investors concerned with AI safety, evaluation, and the geopolitical implications of sovereign AI.
Rayan Krishnan of VALS discusses the critical need for independent AI model evaluations due to the saturation of public benchmarks. The conversation touches on the challenges of evaluating agentic systems and the importance of evidence-gathering for AI policy.
"When Meta released Llama 4 on our held out private benchmarks, the model is actually underperforming. But on all of the major public benchmarks, it was showing incredible capabilities. So there's a huge disconnect between what was self reported based on these open benchmarks and then what we were actually finding with our higher quality, higher signal benchmarks." — Rayan Krishnan, Founder and CEO of VALS
11. Lenny's Podcast: Product | Career | Growth — "Why companies are becoming a series of loops | Anish Acharya (a16z)"
Runtime: 79 min | Host: Lenny Rachitsky, Lenny | Guest: Anish Acharya (General Partner, Andreessen Horowitz (a16z))
Who should listen: Product leaders, founders, and strategists thinking about organizational design and AI's impact on company operations and human roles.
Anish Acharya of a16z and Lenny Rachitsky discuss how company building will increasingly involve creating "loops" of AI agents handling tasks, with human intuition guiding "hill climbing." The episode explores the split between frontier and open-weight AI models and the future of AI's impact beyond productivity.
"It's a funny dark fantasy that we seem to have as Silicon Valley collectively. Like things have never been better. By almost every measure, this is a technology that really amplifies our agency." — Anish Acharya, General Partner at Andreessen Horowitz (a16z)
