Ben Horowitz walks through the systems choices behind a16z: why they centralized control, treated the firm as a network effect, and how AI just rewrote the rules of what makes a company defensible.
a16z was built as a system, not just a checkbook. They centralized control so the firm could reorganize into new categories, spent fee money on building a relationship network instead of salaries, and bet that software would create ~200 breakout companies a year instead of the industry's assumed 15. The deeper lesson: with AI you can finally throw money at a problem and catch a competitor, so code and UI stop being moats and defensibility shifts to network effects, supply-chain relationships, and culture.
Two dated VC ideas
When they started in 2009, Horowitz saw two stale assumptions. First, VC firms had a great product for investors but a bad one for founders, since they mostly just handed over money. Second, everyone believed only about 15 companies a year would ever reach $100M in revenue, so the whole industry just fought to invest in those 15. a16z bet that software eating the world would push that number closer to 200.
Centralize control, share economics
Traditional partnerships share both economics and control, which makes reorganizing almost impossible because everyone gets a vote and nobody willingly gives up power. a16z split those: partners share economics but control is centralized. That let them reorganize and expand into whole new categories like American Dynamism, crypto, and bio. They also kept splitting into small groups, since a real truth-seeking conversation tops out around seven people with good chemistry, and 30 people is a presentation, not a conversation.
The firm as a network effect
Horowitz treated the firm itself as a network. Network value scales like n-squared, so five nodes is worth 25 and six is worth 36, and at internet scale it becomes nearly impossible to rival. a16z tried to build relationships with every engineer, executive, and technology-buying corporation in the Valley, so a founder could tap that network and get powerful instantly. The hard part is always bootstrapping, like selling the first telephone when there is nobody to call.
Bootstrapping with hacks
To bootstrap the network from behind, the partners paid themselves nothing and spent the fee money on building relationships. Their edge was a hack: having sold a prior company to Hewlett-Packard, they knew HP's enterprise briefing center staff, so they called weekly to get the numbers of visiting corporations and invited those companies to see a16z's startups. Overnight they knew more big buyers than firms 50 years old. Rivals nicknamed them 'A-ho' and dismissed the moves as 'just marketing,' which meant they never copied what was working.
AI breaks the old moats
For Horowitz's whole career the rule was you could not throw money at a problem, because nine women can't have a baby in a month and communication overhead kills you. AI changed that: with enough GPUs and data you can now catch a competitor. So code is no longer a moat and neither is UI. Demand is effectively unlimited because the products just work, which is why companies can go from single-digit to tens of billions in run rate fast. The question becomes what actually differentiates you now.
Culture and defensibility
Culture, quoting the samurai, is a set of actions, not a set of beliefs, so it means concrete behaviors like whether you answer a question in an hour or a week, not platitudes about integrity. A clear standard lets you correct people without infighting. On defensibility, Horowitz pointed to Navan, a travel-software company whose moat is thousands of global airline and hotel supply relationships plus the channel to the travel manager, exactly the boring, unglamorous work an AI lab would never bother to replicate.
- a16z centralized control and shared only economics, which is what let the firm reorganize into new categories like crypto and bio.
- A truth-seeking investment conversation tops out around seven people; beyond that it becomes a presentation, so keep splitting into small groups.
- Network value scales like n-squared, but the real difficulty is always bootstrapping the network when it has few nodes.
- They bootstrapped a16z's network by paying themselves nothing and spending fee money on relationships, plus an HP briefing-center hack to reach big corporate buyers.
- With AI you can finally throw money at a problem and catch a two-year lead, so code and user interface are no longer durable moats.
- Defensibility now comes from network effects, hard-to-copy supply-chain relationships (like Navan's global travel deals), and culture, not from the software itself.
- The best ideas come from trying to solve a real problem you have, which usually surfaces a bigger problem, rather than trying to swallow the earth from day one.
- Culture is a set of actions, not beliefs; a company needs one leader to break ties, because a dictatorship beats a democracy in a competitive fight.
In their words
“You can't share control. We'll share economics but we'll centralize control.”
“A culture is not a set of beliefs, it's a set of actions.”
“With AI you can throw money at the problem, because if you have enough GPUs and enough data you can basically solve most problems right now.”
Terms to know
- Network effect
- The property where each new user makes the whole network more valuable, with value scaling roughly like the square of the number of nodes.
- Bootstrapping a network
- The hard early phase of getting a network started when there are too few nodes to be useful, like selling the first telephone.
- LP (Limited Partner)
- An investor, such as a university endowment, that puts money into a venture fund rather than running it.
- Moat
- A durable barrier that stops competitors from copying a company; Horowitz argues code and UI no longer qualify in the AI era.
- SaaS apocalypse
- The Wall Street narrative that AI will one-shot cheap software companies, which Horowitz treats as an overblown, tradable overreaction.
- Voting vs weighing machine
- Buffett's idea that markets price on narrative in the short run and on actual financial results in the long run.
Ben Horowitz at Stanford CS 153: Frontier Systems
New to this? Come build with us.
Reading is good. Building with people is better. Our drop-ins are free and open to total beginners.