Episode Summary
Executive Summary: The episode centers on how AI is rapidly reshaping startups, enterprise software, logistics, and even politics. Elad Gil argues that AI is already driving major margin gains, new business models, and huge revenue ramps, while Jason highlights its effects on jobs, support, robotics, and public markets. They also discuss distributed compute, stablecoins, prediction markets, wearable AI, and the Anthropic-versus-regulation fight.
Main Topics: AI efficiency and startup adoption: The guests argue that startups adopt AI first because they are resource-constrained and can realize immediate cost savings and productivity gains. They see AI as a major sea change in enterprise software and business operations. AI-driven business model shifts and margin expansion: Examples like customer support automation and AI-powered virtual agents are framed as proof that AI can increase gross margin, reduce support costs, and boost profitability without adding headcount. Automation, robotics, and job displacement: Amazon’s warehouse automation, self-driving cars, and humanoid robots are used to explore how AI and robotics may eliminate or reduce large categories of labor, creating both efficiency gains and social disruption. Infrastructure: distributed compute, energy, and data centers: The discussion covers BitTensor, decentralized compute, and where AI training and inference will physically live, with a strong emphasis on energy costs, regulation, and global geography. Consumer interfaces for AI: wearables, voice, and devices: Sesame’s voice-first AI and glasses, Apple’s potential role, and the idea of on-device/local AI show how the next AI interface may be persistent, wearable, and privacy-sensitive. Crypto, stablecoins, and prediction markets: Stablecoins are presented as an increasingly important financial rail, while prediction markets like Polymarket are described as simplified, high-liquidity tools for event speculation and hedging. AI regulation and state vs federal authority: A major policy debate concerns whether states like California should regulate AI independently or whether federal standards should preempt state-level rules. The speakers disagree on the best route but agree the issue is consequential.
Key Arguments: AI adoption starts with startups because they have the strongest incentive to save money and move faster than incumbents. Many AI products appear underpriced relative to the value they create, which drives rapid adoption despite churn risk. Customer support is one of the first major labor areas being automated, and some companies are already seeing large margin gains from it. AI will likely displace jobs in logistics, support, and knowledge work, but the pace and social response matter more than the fact of displacement itself. Robotics is finally moving from promise to reality, with major implications for warehouses, transportation, and industrial labor. Energy costs and regulation will concentrate AI training centers in a few regions, especially the U.S. and the Gulf. Stablecoins are becoming a major buyer of U.S. Treasuries and may become core financial infrastructure for global commerce. Prediction markets simplify complicated financial/speculative products into accessible yes/no markets, which could expand participation dramatically. The AI policy battle is fundamentally about whether regulation should be federal or state-led, and whether states like California should effectively set national AI rules.
Data Points: Gross margin increase: 60% to 68% - Navan S1 example cited as AI virtual-agent automation reduced customer support costs while volume rose. Revenue growth pace: Zero to a few hundred million dollars in 2–3 years - Described as the current ramp for multiple AI companies. Amazon hires avoided: 160,000 fewer hires through 2027 - New York Times report on Amazon automation investments. Amazon hires avoided over longer horizon: 600,000 hires through 2033 - Projected savings from warehouse automation and robotics. Amazon package savings: 30 cents per package - Estimate tied to automation in fulfillment and delivery. Warehouse automation target: 75% - Amazon’s ultimate automation goal discussed from internal documents. Salesforce support savings: $100 million per year - Used as an example of automated customer support cutting costs. Tether Treasury holdings: $127 billion - Cited as of Q2 2025, making Tether one of the largest holders of U.S. Treasuries. Circle Treasury holdings: $18 billion - Referenced as among the top holders of Treasuries. BitTensor nodes: All nodes analyzed daily by a hedge fund - Elad described a small investment in a fund tracking BitTensor projects. AI voice engagement: 5 million minutes - Sesame’s first two voice models reportedly reached this usage. IPO/regulatory period: Quiet period / SEC review pause during shutdown - Discussed in relation to Navan’s IPO timing during the government shutdown.
Pivotal Quotes: "The age of efficiency is upon us." — Jason Calacanis: Used to frame the broad thesis that AI is making companies more productive and margin-accretive. "If you want to charge me 20 bucks for something that saves me $2,000 a month. Okay." — Jason Calacanis: Explaining why AI tools with strong ROI can be mispriced and rapidly adopted. "A lot of the issues with adopting AI is not the AI... it's the change management." — Elad Gil: On why organizational resistance, not technical difficulty, often blocks AI deployment.
Implications: AI is moving from novelty to core infrastructure, reshaping margins, labor, and competitive dynamics. Expect faster startup growth, more automation, more policy conflict, and a bigger role for energy, devices, and financial rails in the AI economy.
From the Episode
And it's helping the company look more profitable. So, kind of an AI story and a fintech story, a lot. The age of efficiency is upon us. Every unit in every company, particularly in startups, who are always resource-constrained, they are the ones to first use these tools because they're resource-contrained. They can turn a nickel into a dollar. A startup's going to do it. Now, big companies, like, well, we're spending $3 to get a dollar in value, and we can afford to do it because we're sitting on a bunch of cash. There's no. You know, imperative. There's no existential dread about money at Apple when you're sitting on hundreds of billions of dollars in cash. So, yeah, we'll get to it when we get to it. But at a startup, it's quite the opposite. You're seeing that as well, I assume, Lad, is startups doing really fascinating stuff with this technology already. Yeah, super exciting. And the ramp on revenue for some of these companies is out of this world, right? I mean, between going from zero to a few hundred million dollars in revenue in two, three years is something that I haven't seen in a very long time.
And there's multiple companies doing that. So I think this is the big sea change, and you see it in enterprise adoption, you see it in all sorts of things. The company that I started with Jared and Eric Wu and others that we mentioned earlier is basically working on this for big enterprise. How can you help big enterprises really make that adoption of AI? And what applications should you provide for them to be able to really thrive in this post-AI world? So I think there's an enormous amount of stuff to be done there. Yeah. And it's going to be really interesting. The cynical take on this is a lot of people say. Sampling AI solutions and trialing them, and there'll be big churn issues and, of course, competition issues. And I guess the optimist view of it, which I'm primarily in that camp, is these things are providing massive value at very low price. Therefore, why wouldn't you try it? It's like, you know, come get three hamburgers that are grass-fed beef for 20 bucks. It's like, how is that possible? Okay, I don't need to know. I'm just going to order the burgers. Like, if you want to charge me, 20 bucks for something that saves me $2,000 a month. Okay. It's just wild to see how.
Engineers, but also private equity and operations people. And if you think about it, the key for any sort of AI-driven transformation is you need, if you're doing a roll-up, you need a PE person. If you're not, you just need, you know, the sort of edict to do it. Then you need somebody to operationally implement it. A lot of the issues with adopting AI is not the AI. A lot of it's actually pretty easy to build given the APIs that we have from OpenAI and Anthropic and others or Google. It's usually the change management. It's convincing the team to adopt it, it's aligning and Of it's saying, Hey, we understand you really like this tool, but we're switching. You know, it's all that kind of stuff. Yeah, the resistance is real, right? The resistance is very real. Yeah. In some cases, maybe it's going to threaten that person's job or perceptually do so. So, um, often when you're doing roll-ups for AI, you need the PE person to be able to do good purchases, you need the ops person to really help revamp the org, and then you need the engineers to implement the AI in the workflow. And so, very few of the roll-up companies that I've seen have this magical convergence of those three skill sets into one.
About This Week in Startups
Jason Calacanis covers startups, tech, markets, media, and all the hottest topics in business and technology. He also interviews the world’s greatest founders, operators, investors, and innovators.