In a recent podcast, a16z co-founder Ben Horowitz used his 16 years of experience running a top venture firm to address the core debate around AI: it is not a bubble. His reasoning is simple yet powerful — demand intensity has never been seen before.
Unprecedented Demand: Valuation Matches Real Growth
Horowitz said people worry about an AI bubble because valuations have risen so fast. But he stressed that customer adoption rates and revenue growth are equally unprecedented. “We have never seen demand like this,” he noted. He gave Nvidia as an example, arguing its multiples are not unreasonable when paired with its growth rate and profit scale. He called AI the largest technology market he has ever seen in his career, not the most hyped.
The Art of Managing Genius: Provide a Framework, Not Instructions
Horowitz shared the management philosophy at Andreessen Horowitz — when you manage world-class experts, traditional methods fail. He mentioned GPs like Chris Dixon and Martin Casado, who are top in their fields. Horowitz said he does not tell Martin what to do but helps him understand the investment decision process and how conversations influence outcomes. He believes the biggest mistake in investing is over-weighting a company's weaknesses; instead, focus on whether it is the world's best at something.
The Essence of Judgment: Knowledge + Intelligence + First-Hand Info
Horowitz offered a formula: decision-making ability = intelligence + knowledge, and knowledge often resides with those doing the actual work. He spends significant time in team meetings, talking directly to deal partners, and learning about accounting and IT teams. He wants employees to immediately flag problems, even if solving them takes only 14 seconds, to prevent escalation. He emphasized that leaders need “clarity” not “correctness” — a clear direction enables team action.
Vertical Strategy: Teams No Bigger Than a Basketball Squad
Horowitz recalled a conversation with the late Dave Swensen: investment teams should not be much larger than a basketball team (about five starters), or discussion becomes performance. To balance market expansion with small-team efficiency, Andreessen Horowitz adopted a vertical structure. Each vertical team stays small, with cross-meeting attendance and annual off-sites for communication. Horowitz said despite the firm's size, internal politics are rarer than in 10-person companies — a result of cultural design.
Limits of Foundation Models: Application Complexity Matters More
Three to four years ago, many believed large foundation models would become all-knowing brains. Reality proved otherwise. Horowitz cited Cursor, which integrates 13 different AI models, each modeling a different aspect of programming, and even released its own coding model. This shows application complexity is extremely high and not captured by foundation models. A team article argued there is no “god-level” video model; different use cases need different models. Foundation models are infrastructure; real value creation is in the application layer. That is why AI will produce more winners than previous tech cycles.
New Ownership Balance: 20% and M&A Revival
Horowitz noted that in recent investments, the firm has secured 20% or higher ownership. For exceptional companies and founders, even a diluted stake can be more valuable if the company grows fast — 20% of a $10 billion company beats 40% of a $1 billion one. He also predicted that AI's disruption will force incumbents to acquire “future DNA” via M&A, reopening the M&A market and offering startups new exit paths.
Giving People a Chance: Technology's Ultimate Mission
Horowitz stressed that the best thing society can do for an individual is give them a chance. He and Mark Andreesen believe giving people the opportunity to do something bigger than themselves and contribute drives human progress. America's importance stems from its ability to offer such chances. Their work is not just about making money, but helping the country win technologically, which matters for humanity.
Horowitz's core message: in the AI era, judgment is the truly scarce resource. When demand is real, technology is solid, and management is sound, the bubble accusation collapses.