Internal Venture Model

Inverting the veto

Brian Armstrong describes an organizational structure at Coinbase that inverts the normal path of an internal idea.¹ In most companies, he notes, a new idea needs sequential approval up a chain: team lead, director, VP, senior VP, CEO. Each is a potential veto, and one "no" at any level kills the idea regardless of merit. The result he identifies is risk-aversion by default, where the organization optimizes for what is acceptable to everyone rather than what might be excellent to anyone, and the most unusual or non-consensus ideas die in committee.

The counter-design requires only a single believer. Twice a year, any Coinbase employee can pitch a new product idea to a room that includes Armstrong, the CFO, the COO, product group leaders, and a few senior engineers. Any one participant can fund the idea from their own budget. A greenlit idea then operates as an internal startup with a small team of two or three people, and only if it hits key milestones does it receive the equivalent of a Series A, a larger internal resource allocation. The rest of the organization does not need to agree. Armstrong's summary: "It's kind of coming in and pitching at 10 venture capitalists. You only need one yes."

The proof point Armstrong offers against himself

The example Armstrong uses is one where his own vote was wrong. He voted no on USDC, the stablecoin. Someone else in the room believed in it and funded it from their budget, and by his account USDC became roughly $800M in Coinbase revenue in 2025. "I am actually embarrassed to admit I voted no on that idea. Luckily, somebody else funded it out of their budget. It tells you that sometimes good ideas can come from anywhere." He draws the implication directly: even the CEO will sometimes be wrong on the most consequential bets, and the system is designed so that CEO wrongness does not equal company wrongness.

Why Armstrong frames it as venture capital

The model, in Armstrong's telling, mirrors venture funding on purpose: a founder-style pitch, early capital only if the idea clears a basic bar, a small team to validate, larger resources only on milestone achievement, and a portfolio mindset in which most bets fail and one USDC pays for many failures. He contrasts it with conventional corporate R&D, where budgets are allocated top-down and controlled by the people with the most to lose from failure; the internal venture model allocates bottom-up to the people with the most to gain. He invokes cautionary history to motivate it, citing Steve Wozniak taking the personal computer to HP and being told no, then leaving to found Apple, and Sam Walton offering his concept internally before building Walmart himself. The fear he names is that a company's best engineers have brilliant ideas, and if the innovation infrastructure kills them, they will leave and build them elsewhere.

Scale as a precondition

Armstrong is explicit that the model requires enough scale to have multiple budget-holders with meaningful discretionary budgets. At seed or Series A, he notes, the internal venture model collapses to "the founder decides." The underlying belief, that the best ideas are often non-consensus and need only one believer, is what he argues survives at any size; at small companies it becomes the case for assembling an early team with genuine domain disagreement, so that someone can fund the idea the founder would vote against. The structure sits close to Ray Dalio's idea meritocracy, which pursues the same goal, letting the best idea win regardless of source, through a different mechanism of transparent collective scrutiny rather than distributed budget authority.