Erik Torenberg on the state of tech media and the Hill and Valley moment for venture-backed founders

Summary

The Hill and Valley Forum 2025 drew roughly 10x the attendance of prior years, with standing-room-only crowds at the U.S. Capitol and a waiting list well beyond actual capacity. The growth reflects a genuine shift: defense tech, re-industrialization, and the tech-policy nexus have moved from fringe to mainstream, and the forum has become the clearest signal of that change.

The vibe shift is real — and raises its own question

Several investors and founders noted that investing in defense, hard tech, and government-adjacent categories no longer feels contrarian. Keith Rabois argued that while the environment has opened up meaningfully, government-only revenue is unlikely to produce many successful companies — the exception being a very small number of firms like Scale AI, which built significant federal revenue early in their trajectory. His advice to founders: understand the federal sales process, which is fundamentally different from commercial sales, and recognize that there are perhaps 40 people in the country who can sell effectively into defense and three-letter agencies. Without one of them, the technical breakthrough matters less.

Zach White from AVC flagged robotics as the next frontier for large-scale compute buildout. His observation: Boston Dynamics and Physical Intelligence are not yet talking about data centers, but the moment robotics companies need to train large transformer models on the data their machines are generating, Nvidia has its next major demand wave. Jensen Huang put that on roughly a five-year timeline.

Jensen, Nvidia, and the re-industrialization question

Jensen Huang's appearance drew the largest crowd of the day. His core message: onshore semiconductor fabrication and manufacturing infrastructure, and embrace the fact that AI will create new jobs, change existing ones, and eliminate others. The unresolved question raised on the stream — whether Jensen is a customer of re-industrialization or a reindustrializer — is a live one. Nvidia designs chips but fabs with TSMC, which is now investing heavily in Arizona. The argument is that if Jensen directs demand toward Arizona-based TSMC capacity packaged with Foxconn, re-industrialization follows almost automatically. Whether he has the tools or the mandate to go further was left open.

Defense budget: round numbers vs. precise ones

Delian Asparouhov, co-founder of the forum and a partner at Founders Fund, offered a useful read on the $150 billion reconciliation package. Round numbers, he said, come from leadership signaling intent; precise numbers come from program offices that have actually scoped the requirement. The $150 billion is a leadership-level signal, not a bottoms-up procurement plan. The $500 million earmarked for autonomous systems struck many attendees as low, but Asparouhov and Sean Maguire both argued the more important variable is trajectory — a predictably growing budget is far more useful to defense startups than a large one-year spike that collapses the following year. Asparouhov drew the analogy to VC stage-gating: no startup, however capable, is ready to absorb $5 billion in a single year.

Maguire's broader thesis on defense: roughly 10 companies in the West will emerge as genuinely large next-generation defense businesses, with several exceeding $100 billion in market cap. The defining characteristic will be vertical integration and efficiency — the opposite of Boeing and Lockheed, which subcontract most of their supply chains and capture only a slice of the value. He also predicted the defense procurement model will shift away from cost-plus, which would structurally favor the new entrants.

Josh Wolf of Lux Capital pushed back gently on unbridled optimism. He expects a defense tech bubble — too much capital chasing too few mature companies — with predictable fraud and failures along the way. His more interesting call was on maintenance as an underappreciated wave: as the cost of capital rises, CFOs will shift capex from building new infrastructure to maintaining existing assets, creating demand for preventive maintenance tech, early-detection systems, and repair robotics.

The Hill side: who actually gets it

Multiple guests independently named Richie Torres, a Democratic congressman from New York, as the Hill figure who most credibly understands the technology industry. Senator Todd Young also drew praise, particularly for his nuanced position on re-shoring — he wants to onshore a small number of genuinely strategic industries, not garments, and is comfortable with medium-strategic goods going to allied nations. Christian Garrett, the forum's co-organizer, noted that every prior Hill and Valley has produced at least one piece of legislation as a direct result of conversations at the event. The 2025 dinner featured 17 Republicans and 17 Democrats in a strictly off-the-record setting — bipartisan symmetry enforced to the seat.

Lobbying, founders, and the direct model

A former General Catalyst and Emergence Capital investor, now running the tech and venture practice at a lobbying firm and serving as a senior fellow at the Foundation on American Innovation, made the case that founders who outsource their government relationships entirely are making a mistake analogous to hiring a banker to run your fundraise. The best founders build the relationships themselves; lobbyists are most valuable as educators, back-channel operators, and intro-makers — not as proxies. He cited Blake Scholl of Boom Supersonic as a model: years of consistent presence in DC, starting when supersonic flight was well outside the Overton window.

Zach Perret of Plaid offered a fintech-specific read. The three live regulatory threads are the Genius Act on stablecoin regulation, Section 1033 of Dodd-Frank on open banking, and payments modernization — whether the federal government can move off paper checks at scale. Plaid has had a DC policy team for seven years in anticipation of the open banking rule; Perret noted the volume of tech CEOs now casually appearing in Washington buildings would have been unimaginable a decade ago.

Keith Rabois, appearing briefly before a panel with Senator Joni Ernst, made the Ramp-for-government case in concrete terms: the federal government has 4.5 million credit cardholders against roughly 2 million employees, and Senator Ernst cited 80,000 charges at casinos on federal cards. Commercial Ramp customers save 5–15% on non-payroll spend; the federal government would likely save more. Cutting 1% from the federal budget, he argued, would give the Federal Reserve room to lower interest rates by 1% without triggering inflation — a larger macro effect than most people register.

Tech media and the state of the discourse

Mike Solana of Pirate Wires offered the sharpest take on tech media's current identity crisis: outlets are trying to find a lane in a culture that has shifted under them, with center-left publications pivoting hard to anti-Trump framing while others are still searching. His own approach is to write the stories that matter to him and stay calm while others are not. On the Ezra Klein abundance book, his read was blunt: it successfully rebranded ideas that were circulating in tech circles years earlier — charter cities, housing deregulation, energy abundance — and is now selling them to a different audience. The conclusion he disputes: Klein's argument assigns blame for the absence of these ideas to the people who created them, when the actual obstacle, at least in California, is the people running the government.

On the AI race framing, Solana was agnostic. He doesn't think anyone knows whether AI is centralizing or decentralizing, whether it constitutes a race that can be won, or what happens after the paradigm shift. He noted Tyler Cowen's AGI declaration as an example of the discourse outrunning the evidence.

The Chatham House postscript

Erik Torenberg, appearing at the end of the stream, gave an update on Chatham House, the cross-partisan group chat that was leaked to the press. His position: no one killed it, though it had grown too large before the leak, and Brian Goldberg had tested its limits. The goal — a genuine left-right exchange — remains intact and the group is still active. His broader point landed as the closing thought of the stream: group chats with people who hold differing beliefs are the structural antidote to GPT-4o telling you that you're right about everything.