What Makes a Great Entrepreneur? | Ben Horowitz on Vintage x Vintage

Ben Horowitz argued that great entrepreneurs have an exceptional strength capable of outweighing serious weaknesses, and that every great entrepreneur has something seriously wrong with them. Investors should seek distinctive ability rather than a candidate who appears flawless on every conventional measure. He would choose an entrepreneur over an idea, although the idea can reveal a founder’s originality and capability. Assessing founders is difficult because investors initially know little about them and may not see how they act under existential pressure—the “WIFIO” moments when the outcome is uncertain and the stakes are high.

Horowitz’s own move into leadership came without a conventional CEO track record. After Netscape was sold, Mark proposed starting another company, and Horowitz agreed to become its CEO despite never having held the role. He had joined Silicon Graphics as an engineer, moved to Netscape, where he met Mark, and received about $1 million from Netscape’s IPO. At 25, married with three children, he was focused on keeping the lights on. He later helped create Andreessen Horowitz because he believed founders needed practical help building companies, while venture firms were largely organized around selecting entrepreneurs and serving limited partners. The firm aimed to use fee income to build company-support resources and valuable founder connections.

Horowitz said familiar résumés and interview signals can hide skills that hiring managers do not know how to recognize. Andreessen Horowitz had people with different backgrounds review job profiles to identify the full range of capabilities each role required; the aim was to widen the hiring bar, not lower it. He cited Chris Lyons, hired for relationship skills rather than investment-banking credentials, who has since been promoted eight times. The firm also saw a gap in Silicon Valley networks for Black entrepreneurs and entertainment figures. NewME, an accelerator for Black entrepreneurs, connected it with founders and talent, and Horowitz said the firm benefited from the relationship, including through its investment in Wonderschool.

Horowitz sees AI as a reinvention of the computer and potentially a bigger shift than the internet, with development compressed relative to the computer industry’s roughly 80-year evolution. Cheaper, easier-to-build software may make creativity and deep knowledge of a problem more valuable: people who understand a need can use AI to turn that knowledge into products. AI tools can also let founders move from an idea to a company without first solving the hiring barriers that once made that difficult. In his framing, the opportunity is to identify a problem, use AI to solve it, and build something people will pay for. He distinguished the current boom from the internet bubble by arguing that AI works, is advancing quickly, and is being adopted unusually fast. The earlier bubble arose in part because technological progress outpaced consumer adoption: there were only 50–60 million internet users, about half on dial-up. Bubbles can form when technology, adoption, and investor sentiment move at different speeds.

Horowitz argued that realizing AI’s potential requires rebuilding infrastructure from models down through underlying systems. AI workloads are harder to define than database queries, and the systems to manage them properly do not yet exist. He compared the task with cloud computing, which required changes to storage, networking, operating systems, virtualization, and containers—not simply moving existing software to the cloud. He questioned whether power-hungry, low-yield NVIDIA chips originally designed for video games are the right long-term answer, and whether models will keep relying on vast quantities of internet data for training. He described chip demand as effectively unlimited while manufacturers struggle to produce enough, but cautioned that power or cooling constraints could create a temporary GPU glut. Displacement of incumbent companies, he said, is part of technological change and does not by itself establish that a bubble exists.

The company 11 Labs illustrates how product use can help build a defensible position: it developed a strong developer product and stack, collected real-world voice data through use, and tuned its model to valuable applications. Its Spanish voices preserve intonation and laughter, helping it win deals with Netflix and Televisa and generate further feedback data. Horowitz also noted that AI may let customers rewrite some software themselves, contributing to low SaaS multiples. After building a CRM and fundraising platform, he said the CRM took one week and questioned whether software costing $400 a month was worth four days of self-building; he cautioned that recreating complex enterprise systems, such as American Airlines’ SAP setup, is harder. He hopes future VCs will say he helped shift investing toward helping people build companies, and measures success partly by whether people who worked with him became better people.