Shury's Share: 20VC: Marc Andreessen on the future of venture capital
- 6 days ago
- 2 min read
From Shury: Marc Andreessen is co-founder of Andreessen Horowitz (a16z), a venture capital firm now managing around US$90bn with investments in many of the most consequential start ups of the past two decades. He also co created the Mosaic internet browser, went on to co found Netscape, and has built or backed a string of internet era companies. He is one of the most influential investors and commentators in the technology industry.
The conversation with Harry Stebbings on 20VC ranges across founders, venture capital, AI, and the future of Silicon Valley. I found it a genuinely useful framework for the decisions I'm making every day, from how I think about allocating capital, to what I'm observing in my own work in the data centre industry as AI demand surges, to the longer term question of how all of this reshapes the career options that will be available to me and then my young children.
A common thread runs through the discussion: Andreessen's belief in the power of agency. It shows up in what he looks for in a founder, what makes a successful VC, and how he thinks about AI and labour. His most contrarian argument is that the labour displacement theory around AI is wrong, that AI will be a net job creator, and that the value of AI will accrue to its users rather than its providers (the application layer rather than the model layer).
The economic mechanism underneath this is Jevons' paradox: when a resource becomes radically cheaper per unit, total demand for it expands rather than contracts. Applied to AI, if cognitive work becomes cheaper, demand for cognitive work grows, it doesn't stay flat (as the zero sum theory of labour displacement assumes) and it certainly does not shrink. Predictions of mass unemployment after every major technological wave have been consistently wrong for the same reason: they assume static demand. SpaceX is a useful current example. As the cost per kilogram to orbit has fallen sharply, applications that simply were not economically possible before have been unlocked: Starlink, lunar missions, and a much wider range of commercial space activity. Each layer of technological unlock creates new demand and applications that were not economically viable before.
The discussion left me more optimistic about the future with AI on the reminder that the future is, to a large extent, in my own hands. AI is a more powerful tool than anything that has come before, but a tool is only ever a tool to the degree to which it is willed and directed. As tools become easier to use and more abundant, the bar for agency rises rather than falls, it becomes less about doing the doing and more about knowing what to deploy the tool on, what questions to ask, and what direction matters. Execution gets cheaper and judgement gets more valuable. The bar is genuinely higher in a world of AI agents, but that is also the opportunity.
Comments