Managing startup investments is essentially managing risk, and Marc Andreessen has an interesting theory on how to model it.
Marc Andreessen — co-founder of Andreessen Horowitz, one of Silicon Valley’s most influential venture firms — uses something called the Onion Theory of Risk to evaluate startups and decide when and how much to invest. It offers a surprisingly systematic lens through which to understand how startups raise capital, hit milestones, and ultimately survive.
“On day one, a startup has every conceivable kind of risk. You can basically just make a list of the risks,” Andreessen explains. “You’ve got founding team risk — are the founders going to be able to work together? Do you have the right founders? You’re going to have product risk — can you build a product? You’ll have technical risk, which is maybe you need a machine learning breakthrough or something to make it work. Are you going to be able to do that? You’ll have launch risk — will the launch go well? You’ll have market acceptance risk. You’ll have revenue risk.”
The list keeps growing. “A big risk you get into in a lot of businesses that have a sales force is: can you actually sell the product for enough money to actually pay for the cost of sale? So you have cost of sale risk. If you’re a consumer product, you’ll have viral growth risk.”
Once you have the full list, the framework kicks in. “The way that I always think about running a startup is also the way I think about raising money, which is it’s a process of peeling away layers of risk as you go. You raise seed money in order to peel away the first two or three risks — the founding team risk, the product risk, maybe the initial launch risk. You raise the A round to peel away the next level of product risk. Maybe you peel away some recruiting risk because you get your full engineering team built. Maybe you peel away some customer risk because you get your first five beta customers.”
The implication for fundraising is direct: each round should correspond to a specific set of risks eliminated, not just a number on a term sheet. “The way to think about it is you’re peeling away risk as you go. You’re peeling away risk by achieving milestones. And then as you achieve milestones, you’re both making progress in your business and you’re justifying raising more capital.”
When pitching a B round to a firm like a16z, Andreessen says the winning approach is simple narrative discipline: “You say, ‘Okay, I raised the seed round — I achieved these milestones, I eliminated these risks. I raised the A round, I achieved these milestones, I eliminated these risks. Now I’m going to raise a B round — here are my milestones, here are my risks. And then by the time I go to raise a C round, here’s the state that I’ll be in.’ And then you calibrate the amount of money that you raise to the risks that you’re pulling out of the business.”
Andreessen acknowledges the framework might sound basic, but he makes the case for it precisely because of how rarely it’s followed: “I go through all this because it’s a systematic way to think about how the money gets raised and deployed as compared to so much of what’s happening, especially these days, which is just: ‘Oh my God, let me go raise as much money as I can. Let me go build the fancy offices. Let me go hire as many people as I can and just kind of hope for the best.'”
The cautionary note at the end of Andreessen’s remarks is the part most worth dwelling on. The “raise as much as possible and hope for the best” instinct isn’t a straw man — it has defined entire eras of startup funding. WeWork became the most cited example of what happens when capital outpaces risk elimination: the company raised billions while foundational risks around unit economics, governance, and the basic viability of its business model remained entirely unpeeled. SoftBank’s Masayoshi Son, who ploughed $18.5 billion into the company, later called the investment a straightforward mistake in judgment.
The Onion Theory reframes what milestones are actually for. They aren’t just growth metrics to impress the next investor — they are proof that specific risks no longer exist. A seed round that doesn’t resolve founding team dynamics or basic product viability hasn’t done its job, regardless of how much runway it buys. This is also why startups that raise oversized rounds early can paradoxically end up in worse shape: the capital masks unresolved risk rather than eliminating it, and the cash burn continues against a backdrop of compounding unknowns.
In the current AI investment cycle, where capital is again flowing abundantly into companies with enormous technical and market uncertainty, Andreessen’s framework is a useful corrective. The risks haven’t gone away — they’ve just been temporarily obscured by enthusiasm. Founders and investors who can name them precisely, and structure their capital raises around eliminating them layer by layer, are the ones most likely to still be standing when the cycle turns.