Harry Stebbings

Town vs Instinct vs GrokBot | Why the AI Assistant Market Is Not a Bubble

Sep 7, 2026 · 1:16:01

Jean-Denis Grèze, Co-Founder and CEO of Town, joins 20VC's Harry Stebbings to argue the AI assistant market is not a bubble and that Town can survive Google, Apple and OpenAI. He explains Town's pivot from AI tax prep to an assistant living in email and calendar, arguing the winning product will have agent-level network effects. Grèze says startups are subsidizing frontier model costs today, pricing for 20–30% margins in 18 months as prices halve every 9–12 months. He predicts people will trust agents to decide what data to share within five years, calls token maxing the wrong metric, and says Apple could lose because it lacks cloud DNA and is tied to on-device privacy. He also refuses to inflate announced round valuations, calling the practice unethical toward employees.

How to Build Your Own Data Center & Why Every Startup Should Do It

Sep 5, 2026 · 1:04:40

Cliff Weitzman, co-founder and CEO of Speechify, tells Harry Stebbings that owning NVIDIA GPUs beats renting because renting costs 1.5x the purchase price yearly, so Speechify pays $100,000 per GPU to get Rubins and B300s early. He covers inference versus training workloads, liquid-cooling and insurance logistics, and NVIDIA's 25% buyback underwriting with Blackstone, BlackRock, Apollo and Goldman Sachs. Weitzman admits his biggest mistake was staying B2C as ElevenLabs leapfrogged Speechify through B2B APIs, and defends entering B2B against ElevenLabs and Bret Taylor's Sierra. He also argues founders should hire math olympiads over coders, details agent orchestration with Claude Code and Cursor, and explains how GPU clusters help him sequence genomes to cure his brother's orphan disease.

NVIDIA Crushes Quarter | OpenAI Cuts Off Cursor | Instinct Hits $2.5B Valuation

Sep 3, 2026 · 1:20:29

Harry Stebbings hosts SaaS investors Jason Lemkin and Rory O'Driscoll breaking down NVIDIA's record $96.2 billion quarter, its $12.9 billion pursuit of Hugging Face, and 70% growth guidance signaling compute demand runs at least another 12 months. Lemkin defends OpenAI's cutoff of Cursor as rational, since coding is the model load for LLMs and Cursor competes directly. The panel argues every startup must now become a compound startup, as AI makes code so cheap to produce that teams ship 100 times more software. Lemkin lays out bull cases for Clay at $7 billion and Linear at $2.5 billion, claiming agents will consume 10 to 100 times more GTM resources than humans ever could. They also flag the Hugging Face agent hack as a wake-up call for CISOs and praise Benioff's pragmatism in embracing multi-surface access and outcome-based pricing at Salesforce.

ClickHouse CEO: AI Margins Need to Improve | Revenue Concentration Should be a Concern

Aug 31, 2026 · 1:05:57

Aaron Katz, co-founder and CEO of ClickHouse, tells Harry Stebbings the AI cycle is 'just getting started' — revenue growth is unprecedented, but the biggest risk is revenue durability because agentic apps have low switching costs. He accepts lower gross margins if expansion is possible, measures AI token spend by ROI, and predicts agents will have identities, budgets, and pick their own infrastructure. He argues open-weight models won't win the enterprise over trust concerns, and sees enterprises moving back on-prem. ClickHouse revenue went 0, 12, 50, 200, finishing north of 500 this year, with >200% NDR; Katz would take the under on $1B ARR by December 2027. He calls the $2B Series B with no product the priciest round; revenue concentration matters — AI customers are under 12% of revenue.

Only 10% of Neo-labs Will Survive | Factory CTO, Eno Reyes

Aug 29, 2026 · 1:29:23

Eno Reyes, Factory CTO, says the cheapest AI model isn't the cheapest system: outcome-based pricing can make frontier models cheaper, and predicts 99% of workflows will run on open models in three years but the remaining 1% holds most economic value. He calls 'Chinese models' a psyop by frontier labs, says frontier valuations are overweighted, sees Anthropic's $2 trillion Claude Code bet as risky due to model lock-in, and credits Microsoft with the best AI hand. Reyes says model routing is commoditized, warns 80-90% of neo labs could die in 18 months unless tied to durable workflows, and insists enterprises must own their intelligence rather than let labs control it. He rejects pedigree in hiring, wants token budgets funding outcomes, and expects SaaS to sell before cannibalization.

Anthropic's $30T Assumption & OpenAI Confirms IPO | Why Customer Service & Robotics are Overinflated

Aug 27, 2026 · 1:26:44

Jason Lemkin, Rory O’Driscoll and Harry Stebbings say the AI boom now depends on Nvidia and hyperscalers financing frontier models, as shown by Nvidia's $6B Poolside purchase and $1B at a $12B pre-money valuation. They dissect OpenAI's promised 2027 IPO, arguing it has no choice as Anthropic outraces it and 'it's all about code'; they call Anthropic's $30T TAM claim delusional and OpenAI's $200 plan a losing consumer business. Also covered: Mercor's $20B round, Hugging Face's $13B sale risk, Citadel unwinding 80% of Leopold's fund, Stripe's 41% growth, agents' data leaks, and why customer support, humanoid robotics and AI services are the dumbest categories. Their verdict: we're less than a third of the way through the cycle and addicted to tokens.

Inside Sequoia's Investment Committee | How the SpaceX & Citadel Deals Went Down | Julien Bek

Aug 24, 2026 · 1:28:28

Julien Bek, a Partner at Sequoia Capital, says Sequoia wins by hunting with conviction, not waiting for inbound calls, and shares founder-reading lessons from Don Valentine, Doug Leone, Pat Grady, Alfred Lin, and Shaun Maguire. He highlights Leone's 'worst reference' question, Lin's warning not to mistake an outlier operator for an outlier founder, and Maguire's ELO method for calibrating references. Bek also explains how asking 'why' five times exposed a fraudulent founder, how his read changes by country (French/German references get 1-2 extra points; Americans get subtracted), and why he missed Anton Osaka of Lovable. On AI, he argues agents become the new customer, AEO parallels SEO, and the next trillion-dollar company will be software masquerading as services. He tells how his mother invested in Revolut at a ~$200M valuation, now worth over $100B.

The AI Bubble WILL Burst | Should we be fearful of Chinese Open-Source | Jerry Murdock

Aug 22, 2026 · 1:10:09

Jerry Murdock, co-founder of Insight Partners, warns the AI bubble may burst by March if the Iran war continues, citing credit market complacency and Japan’s $1 trillion Treasuries. He predicts at least half of Neo Clouds disappear within 36 months, favors Fireworks over Base10, and argues open-source models plus ASIC chips will drive customization while frontier labs retain complex tasks. On security, he calls sandboxes the most underrated layer, naming Docker and E2B as leaders, and expects OpenRouter’s 5% markup to be disrupted by inference exchanges. He adds private equity is highly at risk to any financial dislocation, and in quick-fire bets Nvidia exceeds $10 trillion in five years while picking Meta as the Mag 7 short.

Stripe's $8B OpenRouter Bet | Anthropic's First Profit & The Math Behind Reaching $600B in Revenue?

Aug 20, 2026 · 1:17:47

Harry Stebbings, Jason Lemkin and Rory O'Driscoll analyze SpaceX's $60B Cursor takeover, Stripe's $7B OpenRouter deal, Anthropic's first profit, Silver Lake's $43B Workday bid and Lovable's $13.3B round, arguing all reward speed and growth over margin. Musk got Cursor at ~10x forward revenue; Microsoft, not Meta, is the real loser. OpenRouter is a niche that may vanish in five years, yet could become 20-30% of Stripe revenue. Anthropic's first profit on $11.5B Q2 revenue was inevitable at ~40% gross margin; only 2027-28 projected revenue matters for its IPO. Workday is a 5.3x-revenue system-of-record LBO baseline versus 70x trailing for OpenRouter, and Higgsfield's $5.5B and Lovable's $13.3B are compared to Cursor, with Lovable 'not radically off' as AI apps accrete moats.

Uber President on Travis, China & Self-Driving | Why Autonomy Is Existential | How to Beat DoorDash

Aug 17, 2026 · 1:09:44

In a rare interview, Uber President and COO Andrew Macdonald tells 20VC host Harry Stebbings that autonomy is existential for ride-hailing and distribution beats superior technology. He says Uber One is the most efficient long-term lever — membership beats price incentives — and admits he was wrong to resist it. On AI, Uber blew through a year's budget in four months because usage is vertical; ROI is real but hard to quantify, so he advises combining compute and headcount budgets. Recounting Uber China, Uber burned $52 million a week on subsidies before exiting to Didi, a deal he calls successful. He also explains the Delivery Hero acquisition, why Uber Eats trails DoorDash, and leadership lessons from Travis Kalanick and Dara Khosrowshahi.

How to Build a $100M Growth Engine: Lessons from Wispr Flow & Superhuman | Matt Swulinski

Aug 15, 2026 · 1:20:47

Matt Swulinski, Head of Growth at AI coworker Viktor and former growth lead at Wispr Flow and Superhuman, argues the ECOM paid-ads playbook is the right playbook for SaaS and that paid should start immediately. He details analytics setup, why 90% of SaaS companies get it wrong, and scaling creative to 400–500 new creatives monthly. He covers LTV-to-CAC, token-based economics, referral programs tied to usage limits, and paywalls right after the magic moment. He also discusses AEO as the new SEO, why X ads are polluted, and why AI-native 'systems thinkers' are the only hires that matter. He shares Viktor's affiliate program paying 10–15% revenue share as the highest-ROI channel.

Frontier Labs Threatened by Kimi? Should the US Ban Chinese Open-Source Models & Stripe Buys PayPal

Jul 23, 2026 · 1:27:39

Harry Stebbings, Jason Lemkin, and Rory O'Driscoll debate China's near-frontier open-weight models Kimi and Quan, arguing the US should not ban them but compete. They discuss OpenAI's 'AI Communism' tweet, the commoditized inference market leading OpenRouter to consider sale while Fireworks AI hits $1B ARR at $17.5B valuation. The panel notes the application layer has not arrived, with infrastructure spending far exceeding app revenues. They analyze the Stripe-Advent bid to take PayPal private at ~$60B, predicting a 35% premium deal. Finally, they touch on Databricks' $188B Series M, nuclear energy progress, and the strange inversion of private and public markets.