A startup is a bundle, and Windsurf broke the bundle

We had a good thing going!

I am delighted that Windsurf has turned out well. It sounds like everyone is getting a good outcome of some kind, and the principal actors involved get to claim good faith. So that’s reassuring.

But I am not delighted about what it foreshadows, which I wasn’t able to articulate well in the moment but I’ll explain in this post. I think this event marks the breaking of the startup bundle: perhaps an inevitability in the new economics of the AI talent wars, but a bad thing for startups and for the innovation economy overall.

The hardest thing a startup has to do is solve a coordination problem between different people, who have different goals and make different contributions. You may not have ever thought of it this way, but startups employ a timeless economic coordination mechanism - the bundle (like your cable bundle) - as a way of aligning founders, VCs and employees to invest together and win together. If this bundle falls apart, as the cable bundle did, we all lose in much the same way.

Bundling benefits both buyers and sellers

Time to review our bundle basics, which means going back to Chris Dixon’s classic post from 2012. Read it if you haven’t, it’s worth your time; then come back here.

To recap: the basic intuition around bundles starts by acknowledging buyers’ heterogenous preferences: different people are after different things. Some buyers love history and less so sports: they’d happily pay $10 monthly for the history channel, but only $3 for the sports channel. Other buyers are the opposite. The seller’s dilemma is how to price these channels: at $3, 10, or somewhere in between? None are ideal.

The smart move is to combine the channels into a single bundle offering at $13. Buyers overall get more product, at a price they happily pay, with more consumer surplus on top of it. Sellers overall get to sell more product, and minimize deadweight loss due to non-consumption. Everybody is happy.

I wrote about this a few years back with respect to another familiar purchase: cars. To me, there is an obvious reason why higher-end SUVs have exploded in popularity over the last decade, at the expense of other more varied kinds of cars. It’s because a performance SUV is a bundle of a bunch of things you want in a car (e.g. Minivan plus Fun to Drive) and bundling it into a single, expensive product draws out more overall consumption from buyers, letting carmakers invest more back into that premium product.

The key to bundle economics is that the benefits of bundling are proportional to how varied the buyers’ preferences are. The steeper the demand curve across all of the individual bundle components (e.g. individual TV stations), the more powerful the combined economics of a single subscription, because there is more deadweight loss that can be squeezed out by the combined bundle.

A startup is a bundle

So, bear with me; we’re going to describe a startup in these same terms. Specifically, we’re looking at the transaction of pricing and selling equity. Optimally priced equity is the thing that makes the startup go.

Startups have a product called “upside”. That’s all they are, at first. And the founder has to create a market for that upside. They are the sellers; investors (which include angels and VCs, but also include employees of all kinds) are the buyers.

All of these different customers have distinct concepts of what they think upside is. They have different utility functions of “I’m interested in paying X to get Y” where ‘Paying X’ means giving your money or time in exchange for equity, and ‘Getting Y’ means working towards, and being rewarded by, liquidity scenarios.

So a VC on their own might prefer to pay a lower valuation (and put in less cash), in order to hire less people, fund a smaller company but where they own more, and aim for acquirable cash flows. An employee, conversely, might rather get more equity percentage in the company for their work, raise less venture capital, and explicitly hope for the company getting acquired for its engineering talent. Each person might be individually more interested in a more bespoke version of the company and its upside potential, but that’s more targeted to what they want.

In other words, each individual buying the equity has their own preference for what they’d want to pay, and the kind of upside potential path they’d want to pursue. On the other hand, the founder’s aim is they want to maximize the total amount of company potential they can create. They want to make one big beautiful company.

And so, what founders do is offer all of the kinds of investors and employees, more or less, the same bundle offering. (Yes, VCs get protection in the preferred stock, but the upside scenario converts to common.) This is particularly important for employees, where no matter if you’re junior or an ambitious VP, you all get the same stock. The stock is a bundle. Everyone buys the combined product, paying a higher price for the bundle. And like the cable bundle, the founder can now put more resources into the combined product, and maximize the overall offering (the potential to win.)

There is one oversimplification going on here, which you may have spotted, which is the assumption that a startup is like a consumer product in that, the higher the price and the more money you raise, the more the startup’s chance of success. (Because all of the money collected from people buying in goes towards realizing the upside.) Of course it’s not as simple as that, but for the purpose of this argument I think you should accept it as true - because the ability to coordinate others and raise money is the thing that founders must be able to do here, just like “sell TV subscriptions” is the thing that the cable company must do.

What happened with Windsurf?

So, here’s a rough sequence of what happened. (And, to reiterate, I have no inside info here, I may be wrong or incomplete about details; this about the general gist and narrative.)

OpenAI proposes to acquire Windsurf (an AI coding IDE that it makes reasonable sense for them to acquire.) For various reasons, possibly because Microsoft put pressure to kill it, the deal with OpenAI falls apart and instead, the founders and a few senior employees walk to Google for a massive pay package. In exchange, Google gives Windsurf a pile of money for a hand-wavey non-exclusive license to the technology (so, basically, gave them a pile of money). The pile of money gets distributed between investors and tenured employees, and then subsequently the remainder of the company got acquired by Cognition.

Early on, there was a reflexive cynical read here (which I’d had) that the founders were pulling a fast one: parlaying an acquisition term sheet into themselves being acquired for a massive pay package, not for the company’s benefit but for their own. (And supposedly defensible under ‘this my free will employment, you can’t enforce a non-compete, and you can’t sue me personally.”) This take didn’t tangibly materialize, and thank goodness, because that would’ve set an absolutely terrible precedent of behaviour and I’m glad that it clearly was not set. (Although some, like Delian above, retain the cynical take and that’s justified.)

There is another read, which came into vogue a few hours later as details emerged, which went “these kind of weird negotiated disbursements of a startup’s component value is a weird and shitty consequence of regulatory M&A scrutiny.” If regulators won’t let Google buy other startups for antitrust reasons, then this is a way around that. Also, fine - if you want a team’s talent and you can either get it now or after months of antitrust review, obviously you’d rather have it now.

After a few days’ reflection, I have a different way to articulate the issue. It’s that the war for AI talent, and the new sky-high stakes involved in securing hot researchers, is putting pressure on the startup bundle. It tempts everybody involved to unbundle the different components of upside - too good on their own not to pursue.

The AI talent wars unbundle the startup

So what we have here, I think, is the “unbundled disbursement of a startup’s upside.” That is a weird and new thing to contemplate. Yes, everybody got paid here and that’s great. But let’s look at the precedent it sets, which is that the valuable employees of a company can be acquired separately from the business of the company. And if each can be negotiated separately, in an upside scenario, then each can be negotiated for at the outset, when the equity is first issued.

We used to all accept that a startup was one cohesive thing - aside from in some downside scenarios. (You can have unhappy liquidity events where the startup gets sold for spare parts in some way.) But in happy scenarios, the arrangement we all accepted was that successful liquidity came as a singleton event. One company (or the public market) bought you, it bought common stock, you all win together.

Now we have this curious thing which is an upside liquidation event in parts. The founder and some employees got their payment in one kind of way, the rest of the company and the investors got their payment in another kind of way. And maybe everyone is happy. Sure! But this is like one group of people enjoying just the history channel, and the other group of people enjoying just the sports channel. They start wondering, “Maybe I could specifically pay for the sports channel?”

The analogy starts to become more relevant if term sheets start changing to accommodate or anticipate these kinds of unbundled disbursements, and employees start asking (or looking) for bespoke kinds of upside. This is the unbundling of the startup as a singleton product, being replaced by a new kind of upside that’s tailored around the kind of work and the kind of reward that an individual contributor might want. This is like subscribing to the niche streaming service. It’s exactly what you want, and you’re paying a price you think is fair, and getting a product that’s what you asked for.

No individual person can be blamed for wanting this, and in particular, you can’t really be upset with the superstar AI talent for whom the unbundled upside event is so obviously better. You can’t blame an individual for breaking off the bundle if they’re acting rationally and in good faith. You can’t blame superstar AI talent for whatever they can negotiate in their pay packages.

But if startup funding goes the way of the cable bundle - replaced by a more individually crafted mishmash of upside possibilities - I don’t think that’s a good thing. Because the job for founders gets much harder. You lose the economic benefit of the bundle, which was avoiding all of the deadweight loss due to non-consumption that you get when everyone is after their own individual thing. The irony is that the very thing that makes bundles work - the heterogeneity of buyers’ preferences - is itself the thing that breaks the bundle.

---

For email updates you can subscribe here on Substack, or find an archived copy on alexdanco.com.