America’s AI Race Requires Silicon Valley to Build for National Interest

Ben Horowitz frames a16z’s scale as carrying obligations beyond capital allocation. After the firm raised more than $15 billion across new funds, which David Ulevitch described as the largest fundraise in a16z history and the firm as “the largest venture capital firm in the world,” Horowitz cited advice he said he once received from Andy Grove: when you lead an industry, “the size of it, the ethics of it, the morality of it kind of depends on you.”

For Horowitz, that responsibility now runs through a national strategy question: whether the United States wins the next technological revolution. He ties America’s current military, economic, and cultural influence to having “won the Industrial Revolution” through superior technology. The relevant comparison now is artificial intelligence. If AI is the next industrial platform, then the issue is not only which companies succeed, but whether America retains the influence needed to preserve a system that, in his telling, gives more people “a chance to contribute.”

Horowitz does not claim America offers equal starting conditions. He says explicitly that it does not: some people are born rich, some poor; some have better genetics; some grow up in dysfunctional families. His narrower claim is that America is unusually good at giving people a shot, and that chance is central to human progress because contribution is how people create impact. Most countries, he argues, give people either no chance or a “very small” one.

That is why he describes a16z’s mission in national terms. The firm’s role, as he puts it, is to help America win technologically: through what it invests in, how it helps technology integrate with government, how it makes government systems more accessible to citizens, and how it works with allies. American Dynamism, in this framing, is not merely a sector label. It is the firm’s attempt to organize venture capital around the premise that technological leadership, state capacity, and the ability for people to contribute are connected.

Ulevitch pushes the claim further by asking why American technological dominance would be good not merely for America but for the world. Horowitz’s answer rests on the American theory of rights. He points to the Declaration of Independence’s “we hold these truths to be self-evident,” emphasizing that the rights described there are treated as coming from a higher source rather than from government. That matters, he argues, because if government did not grant those rights, government cannot simply take them away.

The example he uses is free speech. Other countries, in his account, can reconsider whether speech rights are “such a good idea anymore.” The American structure makes that harder. Horowitz does not claim the system is perfect or uncontested; he says there are people who “would like to change some things” and believe they know better. But he argues that the durability of rights in the United States is unusually strong and is central to why American power is preferable to the alternatives.

David Ulevitch locates American Dynamism at the intersection of advanced software, hardware, national security, manufacturing, and the defense industrial base. He asks Ben Horowitz what has surprised him most about investing where national security and venture capital meet, especially after the firm built a larger presence in Washington.

Horowitz says the surprise has been the speed of the catch-up. Before ChatGPT, he says, the conventional wisdom was that China had a major AI lead over the United States. ChatGPT complicated that view by making it appear that the United States might instead be ahead in frontier AI. But Horowitz says one part of the earlier assessment was correct: China was far ahead in integrating AI into government, including the military, bureaucracy, and other state functions.

By that measure, he says, the United States began “very far behind.” What changed his outlook was the number of entrepreneurs willing to help and the government’s willingness to engage. He describes U.S. officials as increasingly open to new technologies and new companies, and even willing to change rules where necessary. That willingness, he says, has been “amazingly fruitful,” and he claims it is already visible in current conflicts.

Ulevitch calls this a moment of optimism: a sign that systems can shift quickly when urgency and “changemakers” align. In Horowitz’s argument, model performance and private-sector invention are not enough on their own. The United States also has to adopt, procure, and operationalize new technology quickly enough for public institutions and defense systems to use it.

David Ulevitch raises the recent Anthropic controversy directly, referencing Emil Michael, identified in the discussion as Undersecretary of War for Research and Development and CTO of the Department of War. Ulevitch says the issue had reached mainstream news in the previous week but had been ongoing before that.

Ben Horowitz’s interpretation is blunt: the deal did not fall apart because of philosophical differences. It fell apart, he says, because Anthropic wanted out.

His reasoning is based on leverage. In software deals, he argues, a vendor that is already deployed and whose customer, in his phrase, is “about to go to war” has extraordinary leverage. If Anthropic had wanted to remain in the arrangement, he says, anything reasonable it asked for would likely have been granted, “even probably beyond what they should have.”

Horowitz points to his account of the sequence — Anthropic finding something it could walk away on, not returning Emil Michael’s calls, and a call to the administration — as evidence that the company wanted out. He says he does not know why: it could have been “an employee thing” or “Dario thing” or something else. But he rejects the public framing that the core issue was an ethical conflict over AI use.

Horowitz’s counterargument is that if a company wants its AI used safely, the U.S. government, and particularly the Department of Defense, is the environment with the most rules. Speaking to a Washington audience, he adds that if those rules were broken, the violation would almost certainly leak to the press. “There are no secrets in Washington,” he says.

The lesson he draws for founders is that companies should not casually substitute internal sentiment for the judgment of institutions responsible for national security. He recalls telling the firm when American Dynamism began that employees could hold whatever geopolitical views they wanted, but the world still contains war and crime. Given that reality, he asks whether a technology company is really prepared to decide that people serving in the military, who risk their lives to protect others, should be denied the best technology because company employees believe they are ethically superior to the State Department, Defense Department, or intelligence agencies.

Ulevitch agrees, describing boardroom versions of the same problem: founders or CEOs calling to say employees do not want to do business with a particular country. His response is that companies should not act as if they are the State Department. Horowitz calls many of these internal geopolitical judgments “dime-store morality” and “vibe geopolitics,” while acknowledging that the issues themselves are complicated.

American Dynamism, David Ulevitch says, is about America and “her allies,” not America in isolation. He asks how the United States can export startup culture, manufacturing capacity, defense-building energy, and broader “American dynamism” to allied countries, especially those connected through supply chains and geography.

Ben Horowitz says he does not have a complete answer. But he identifies one reason America is difficult to copy: entrepreneurs need to believe that if they bet their lives on an idea, the government will not arbitrarily take it away. In his view, very few countries offer that confidence. Sweden and Israel are exceptions he names as places with strong technology entrepreneurship, but he presents the broader condition as rare.

Still, allies bring capabilities the United States needs. Horowitz singles out Mexico for high-quality manufacturing expertise. He says American cars made in Mexico are, in some respects, better than Chinese manufacturing on many fronts. Mexico has talent, wants to work with the United States, and is a partner. But because the U.S.-Mexico border is “a thin border,” he argues that helping secure the country is part of the strategic relationship.

Japan is the other major example. Horowitz says Japan has long loved robots and is strong in manufacturing, while the United States has a “big deficit” in the robot supply chain. He also says Japan “just went from 0% of GDP to 3% of GDP spent on defense,” and that Japanese and American interests are highly aligned with respect to China. For a16z, its portfolio companies, and the country, he sees opportunities in that alignment.

Ulevitch adds that recent progress in robotics suggests the world may be on the edge of a robotics revolution, one in which a16z companies and U.S. allies will both have roles. AI, in this account, is not a purely digital race. The strategic frontier includes robotics, manufacturing, defense production, and supply chains — areas where allies may hold essential pieces of the system.

David Ulevitch cites a recent deep dive by Packy McCormick that described a16z as a power broker using capital and networks to shape markets and influence. He asks Ben Horowitz how the firm should balance that power with authenticity toward founders and institutions.

Horowitz accepts power as part of what the firm offers. Entrepreneurs may have a major invention or important technology, but they often lack the ability to get the right meeting: with Congress, with a regulator, with a major CEO customer, or with someone who can help a market form around their product. In that sense, he says, power is “a feature of our offering.”

But he distinguishes that from the internal culture the firm wants to project. He refers to a16z’s early cultural idea of doing “first-class business in only a first-class way.” Applied to entrepreneurs, that means respect for what they are building. It does not mean always being their best friend, he says. It means showing up on time, responding promptly, being honest, and attending to the small behaviors that prevent the relationship from becoming “we’re the powerful and you’re the weak.”

That distinction is central to his justification of scale. a16z wants enough institutional weight to move through Washington, large customers, and complex markets on behalf of founders. But Horowitz argues that if the firm behaves as though its power makes founders subordinate, its position at the top would be short-lived.

The structure of venture capital, in Horowitz’s view, is changing for related reasons. Historically, he says, venture capital was organized around scarcity: in any given year, a study suggested that roughly 15 companies would be started that would ever reach $100 million in annual revenue. The traditional venture model was built to get into as many of those 15 as possible and avoid the rest.

That scarcity justified small partnerships with shared economics and shared control. If there were only a handful of companies that mattered, there was little reason to build a firm with hundreds of people.

Horowitz says software changed the premise. Marc Andreessen’s “software is eating the world” thesis meant that every industry and company would become better with a great software team at its core. Now, he says, that shift is evolving into AI. The result is that “every interesting company that gets started is a technology company,” creating many more investable companies and requiring venture firms to scale their organizations.

The problem, in his view, is governance. Older firms with shared economics and shared control struggle to reorganize. Reorganizations redistribute power; when power is redistributed, people get angry; if everyone has a vote, they can block or distort the change. “You can’t scale without reorging,” he says. a16z, by contrast, has been able to scale because it has centralized control.

The resulting venture landscape, as Horowitz describes it, is barbell-shaped. On one side are large firms capable of covering all major technological areas. On the other are specialized firms focused on specific domains such as AI infrastructure, bio, crypto, or games. The firms in the middle, he believes, are being squeezed.

Ulevitch agrees and adds that outsiders sometimes ask how he can work at a 600-plus-person firm. His answer is that a16z operates as many small teams using shared platform services, which lets teams move with urgency while drawing on the “might” of a large organization. He also says many general partners at a16z have been CEOs, which makes hierarchy and “a benevolent dictatorship” more appealing rather than threatening. The critique of traditional venture partnerships is that “death by committee” makes adaptation difficult precisely when the market demands it.

Ben Horowitz argues that the media environment has changed less because of a clean split between “old media” and “new media” than because every channel now plays by new rules. In the old environment, he says, media strategy was defensive. There were few channels and strict formats. A newspaper might include a few quotes and characterize the rest. A cable-news appearance might mean a short, hostile exchange under bright lights. Because distribution was scarce, a mistake could become permanent.

In the current environment, there are effectively unlimited channels and formats. That changes the objective. The key is no longer simply avoiding mistakes; it is being interesting enough to get attention at all. If a speaker says nothing interesting, Horowitz argues, the message will not travel because there are too many other interesting things competing for attention.

He uses Alex Karp and Donald Trump as examples of figures who win attention by being entertaining. Karp, whom David Ulevitch says had appeared earlier that day and is effective at going directly to customers, is described by Horowitz as captivating precisely because he is hard to predict. Horowitz jokes that he was not sure Karp knew what he was going to say next.

But Horowitz does not equate being interesting with being undisciplined. Karp, he says, is consistently pro-America. That message is predictable even if the delivery is not. If someone makes a mistake in the new media environment, Horowitz argues, the remedy is to appear on “10 podcasts tomorrow” and flood the zone. The Wall Street Journal, podcasts, and other channels are all part of the same overwhelmed media system.

Founders and government officials, in this telling, should not treat communications as a purely defensive exercise. Narrative and trust still matter, but in a saturated environment, trust has to travel through directness, repetition, and a message that is both consistent and compelling enough to be noticed.

When David Ulevitch asks what worries him despite the firm’s optimism, Ben Horowitz does not name China’s model capabilities, venture competition, or procurement barriers. He says his biggest worry is the perception of technology in America.

He cites a poll showing that more than 70% of people in China are optimistic about AI, while less than 30% in America are optimistic. The reason, he argues, is that Americans focus heavily on the dangers of AI and not enough on what it can make possible. He contrasts that with Japan, where someone told him the startup ecosystem is restarting because people are “so fired up about AI.” Japan’s enthusiasm for robots becomes part of the example: a culture more inclined to welcome the technology may move faster to build with it.

Horowitz does not deny the risks. He names “AI overlords,” mass surveillance, and other dangers as real concerns. But he says the positive uses are “extremely positive” and the negative ones can be managed, as societies have managed the downsides of earlier technologies. His analogy is fire: it can burn down a village, but it also heats homes and cooks food.

The positive case he wants emphasized is sweeping. Horowitz says AI could help end traffic deaths, cure cancer, and end poverty “as we know it.” His concern is that a society that talks only about danger may lose the will to build, while competitors and allies that see possibility may move ahead.

The final on-screen disclosure states that the views expressed are those of individual a16z personnel, that the content is informational only and should not be relied on as legal, business, investment, or tax advice, and that a16z and its affiliates may maintain investments in companies discussed.