Ben Horowitz on the Global Race for Tech, Power, and Influence

Summary

• AI models are becoming geopolitical infrastructure because they will mediate nearly every product while encoding contested histories, ethics, and values. Horowitz’s warning that “the models are not objective. They have opinions” makes model choice a sovereignty decision, not just a benchmark or cost comparison: the defaults inside cars, education, and household systems will project somebody’s worldview.
• Deterrence increasingly rewards innovation velocity, not only military scale, pulling private AI, autonomy, and cyber vendors into the core of allied security. Neuberger points to the Strait of Hormuz and Red Sea: adversaries can field cheap, software-built systems, while “those technologies today are not being built by governments.” Government-private-sector access and allied interoperability therefore become strategic assets.
• AI and APIs globalize product demand before startups have the organizational capacity to serve it, creating a distribution bottleneck for venture-backed companies. Raghuram contrasts the old threshold of “a few hundred million dollars in revenue” with today’s earlier international pull, but stresses that local relationships and market structure still matter. In top-heavy economies, five or 10 companies plus government can matter most, so access may outperform a premature full-country rollout.
• An anchor customer can reverse the economics of market entry: a $5 million or $10 million opportunity can justify the same scale of upfront country investment. Horowitz argues this is a16z’s edge in allied, AI-forward, relationship-heavy markets, where government, business, investors, and adoption are intertwined. The target map includes Japan, Korea, the Middle East, Mexico, and Canada, while Raghuram notes Japan, Korea, and Taiwan contain 15–20% of the Forbes Global 2000.
• Cybersecurity is the clearest dual-use AI opportunity and the hardest policy trap: finding a vulnerability enables both patching and exploitation. Neuberger believes the models may help defense “far more,” because defenders cover a broad expanse while attackers need one opening; AI can “jiggle every doorknob continuously and at scale.” Yet Horowitz warns that restricting vulnerability discovery could also prevent defenders from auditing and patching their own code.
• Silicon Valley is not software that can simply be copied online; its moat combines technical talent, entrepreneurship-friendly rules, and a culture that grants status to risk-taking. Horowitz calls the internet-as-distributed-Valley thesis “happy talk” and warns the culture is “so easy to destroy.” The investable corollary is that tax, property, hiring, and social incentives can expand—or abruptly shrink—the founder and growth-capital pipeline.

Deep dive

1. Technology has become the arena of national power

• Horowitz grounds a16z’s international expansion in an America-first mission: a person with only an idea has their “best shot” in America, whose economic, cultural, and military leadership depends on technology leadership. Bringing allies into that mission produced three linked functions—government relationships, global capital relationships, and international go-to-market work that drives adoption.

• Neuberger’s view from years in the intelligence community and White House is that technology moved from a diplomatic tool to “the central arena of both economic growth and national power”: chip and drug-component supply chains can become leverage, while vulnerabilities in power or water systems are no mere IT problem. Deterrence now includes “the pace of innovation” against cheap software-built systems originating largely in private AI, autonomy, and cyber companies.

2. AI exports values along with capability

• Horowitz’s core premise: when AI is “the interface for everything,” systems from cars and refrigerators to education will be built on models whose judgments differ by history, culture, ethics, and law. Because those histories and values differ by vantage point, “the models are not objective”; model deployment also deploys values.

• Kha’s formulation—that it is important for the world to run on “American and Western tech”—and Horowitz’s cyberspace-colonization framing make the issue one of trust and political agency. Horowitz prefers projecting rule-of-law and opportunity-oriented values to receiving defaults he believes offer people less opportunity.

• The sharp commercial specimen came from Mexico: Horowitz says a16z connected TelevisaUnivision CEO Alfonso de Angoitia, whose company had rich Spanish-language content but faced heavy competition from global streamers, with ElevenLabs. Within a couple of months, actors’ intonation, laughter, and performance could travel into new languages across France, Poland, and elsewhere; he says the success then helped produce a big Netflix deal.

• Neuberger tied the example back to censorship: using an American model, they did not have to worry that a sensitive reference to Tiananmen Square or the Chinese government would be altered during translation. Her caveat matters: a16z “deeply value[s]” open source because there are different models and price points; the issue is knowing whether the infrastructure preserves what was actually said.

3. Product goes global before the company can

• Raghuram’s compressed law is “products and technology travel faster than companies can.” APIs make software globally discoverable immediately, and AI makes localization easier, so founders can no longer wait for a few hundred million dollars of revenue before looking overseas; both B2B and consumer products encounter international pull earlier.

• His counterweight: “the way you do business hasn’t changed that fast.” Scale still requires people in-region who understand structure, partnerships, and trust, especially in relationship-based rather than transaction-based economies; being digitally available is not the same as having a repeatable local sales motion.

• The market-entry math is unusually concentrated. Raghuram says five or 10 companies often matter most, with government a major influencer and buyer; Horowitz estimates entity setup, hiring, and localization can cost $5 million or $10 million before a first deal. If the first opportunity itself is worth $5 million or $10 million, founders can rationally enter earlier instead of funding blind expansion.

4. Allied markets reward concentrated access over blanket expansion

• The geographic scorecard combines US interest, allied status, AI-forward modernization, concentrated buyers, and a business culture where trusted access has high marginal value. Raghuram says Japan, Korea, and Taiwan hold 15–20% of the Forbes Global 2000; Horowitz wants simultaneous links to government, business, and investors because “those things are not separate.”

• Neuberger uses Japan as the defense case: it has moved beyond its postwar pacifist policy and wants a military for today’s wars. She situates that effort within the broader shift toward fewer, smaller exquisite platforms alongside more numerous, cheaper ones. Japan seeks to build—not merely buy—autonomous ships and AI for an integrated battlespace while retaining domestic capabilities and protected supply chains.

• The same lens changes market priority. Horowitz wants a more connected Western Hemisphere supply chain with Mexico. Neuberger connects Mexico’s public-safety and national-security needs—including autonomous ships along its vast coastline—to US interests, including fentanyl entering via Mexico. Saudi modernization and AI infrastructure create openings such as introducing Adam Neumann’s Flow to government, businesses, and investors in Riyadh.

• Horowitz distinguishes this from VC expansion whose “product is the money.” a16z waited until it could offer company-building networks, aims to fill growth-capital and globalization gaps rather than displace seed or A-round investors, and may back local winners after understanding their ecosystem; foreign presence begins with portfolio-company acceleration, not primarily asset gathering.

5. AI can harden cyber—and industrialize attack

• Neuberger’s cyber example captures AI dual use: both attacker and defender first find a vulnerability, then one exploits and the other patches. She personally believes models help defense “far more,” precisely because defending a broad estate is harder than finding one entrance, though the near-term attack surface remains vast and fragile.

• Her policy problem is distribution: when capability fits on a thumb drive, “export controls are a joke,” so government and model builders need joint safety design. Horowitz’s pushback is that restricting release of cyberhacking capabilities could also prevent defenders from finding vulnerabilities in their own code; as soon as a model can find vulnerabilities, it can also enable attacks.

• Neuberger’s metaphor is attackers able to “jiggle every doorknob continuously and at scale.” The highest-risk areas include public, third-party, and unmaintained open-source dependencies, and companies are now taking more responsibility because the cost of leaving code insecure has risen while the cost of addressing it has fallen. Horowitz’s longer arc—cyber moved from vandalism to theft to war—leaves a “20-year legacy of code” built before AI as both risk and opportunity.

6. Silicon Valley’s fragile moat is culture plus policy

• Asked whether the internet itself could become a distributed Silicon Valley, Horowitz answers, “I don’t think that’s true…that’s happy talk.” Countries still differ in law, policy, and governance despite being online for decades; he wishes more US policymakers heard foreign leaders ask for Silicon Valley before trying, in his view, to weaken it at home.

• His first two ingredients are build-capable talent—often visible in a strong technical university—and rules that make extreme risk survivable. Entrepreneurs must be able to form companies, hire and fire, face workable taxes, and own valuable illiquid assets; talent alone is common in countries that never form a Valley.

• Horowitz says Norway introduced an unrealized-capital-gains tax that “effectively ended the tech ecosystem there.” The harder ingredient is cultural status: do capable young people find it admirable to found or join startups, or does rhetoric equate billion-dollar success with theft? That norm is “so hard to build” and “so easy to destroy.”

• Kha closes with a World Cup analogy: whether America can win matters less than the fact that America thinks it might—“a feature, not a bug.” The closing exchange invokes “I think I can,” with Jen endorsing the idea that in America people can believe they can.