This year token offerings raised approximately six billion dollars — capital approaching the size of the entire venture-backed IPO market, a comparison that would have been unthinkable as recently as 2015. We did not participate and we have foregone meaningful would-be returns by not doing so. This letter is about why we made that choice and whether we still endorse it.

On the ICO Wave We Did Not Surf

The token offerings of 2017 raised capital, in many cases, against business plans that would not have qualified for a traditional Series A round. The capital was raised from non-institutional buyers under regulatory frameworks that had not yet been clarified, on terms that gave the buyers no governance rights and no path to redemption. The structure was, in our judgment, more closely related to securities issuance than to currency creation, and we believed at the time — and continue to believe — that the eventual regulatory response will be more aggressive than current participants expect.

We may be wrong about this. We are willing to be wrong about it. The conversations we have had with peer firms throughout the year have been instructive; several of them are participating, citing, in roughly equal measure, exposure to the underlying technology, the strength of the founders, and the velocity of returns available. We do not contest the first two arguments; we have meaningful exposure to the underlying technology through our 2009-2014 commitments to traditional companies in the space, and we are aware that several of the founders raising via tokens are operators we admire. We contest the third argument. The velocity of returns, in our reading, reflects a temporary capital imbalance that will correct, and the correction will damage the holders who entered late more than the founders who issued early.

We have estimated, internally, what our 2017 vintage would have looked like if we had participated proportionally to our peer firms' allocations. The number is meaningful. We are not pretending it is not. The framework that justified declining is the same framework that has produced our better outcomes over twenty years; suspending it for one year would mean that we did not have it.

On the Underlying Technology We Do Believe In

We separate the regulatory question from the technical one. The technical claim of decentralized cryptographic consensus — that meaningful coordination among untrusted parties can be performed without a central intermediary — is, we believe, real and durable. The applications of that claim to financial infrastructure, identity, and certain categories of provenance are likely to produce significant companies over the next decade.

None of those companies have yet been built. Most of the capital raised this year has been allocated to companies that will not build them. The pattern is familiar from the 1999 cycle: the underlying technological shift was real, the companies built to exploit it were largely not. The companies that will compound from cryptographic consensus, in our prediction, are the ones being founded right now by teams that are not raising via tokens and are not focused on speculative returns. They will be smaller, less visible, and more boring than the companies dominating the year's headlines. We are looking for them.

On the Founders Who Asked Us to Reconsider

Three founders we respect approached us this year about token-based fundraising structures for their companies. We declined, in all three cases, to participate. The disagreements were respectful and substantive. We do not yet know whether we will be vindicated. The framework we used — that capital structures with no recourse and no governance produce worse incentives over a five-year horizon than capital structures with both — is the framework that has served us in every prior cycle.

The harder of the three conversations was with a founder who had been our portfolio company for six years and was considering a token raise as a way of accelerating product distribution. We told her, plainly, that we would not participate, and that our refusal should not be read as a vote of confidence in alternative paths. We told her, also, that our position was not negotiable and that, if she proceeded, we would expect the company's relationship with us to evolve in ways that neither side would prefer. She did not proceed. We do not know whether she made the right operating decision; we do know that she made it on the basis of clear information from us, which is the most we believe we owe to founders we partner with.

On the AI Investments That Look Inevitable

Behind the noise of the token market, the substantive technical progress of 2017 has been in artificial intelligence. We deployed materially more capital into AI-native companies this year than in any year prior. The valuations are high, but the technical work is real, and the gap between what is possible in the laboratory and what is shipped in production is closing in a way that suggests the next several years will produce the most consequential applied-AI companies of our generation.

The companies we have backed in this category share a property: they have built their products around models that the company itself has trained, on data the company has acquired, against problems the company has chosen because the problem-shape rewards iteration on the model rather than iteration on the user interface. This is, we believe, the durable structure of AI-native businesses. Companies that integrate models trained by others against problems chosen for general appeal will, in our prediction, be commoditized within five years. Companies that have integrated models against problems whose shape rewards proprietary iteration will not.

We are pricing this distinction into our underwriting. It is, in 2017, the single most useful filter we have for evaluating AI investments.

A Closing Note

Foregoing returns we could have earned by lowering our standards is the cost of having standards. We endorse the cost. We will note, for the record, that the 2017 vintage will likely be the vintage about which our 2017 decisions are most second-guessed in our 2018 conversations. The 2018 letter will address what we found.

The Partners
Winzheng Family Investment Fund · December 2017