Product1 publisher2 min readPublished
A16z counts one Databricks stake as a fifth of its assets under management
A16z's post to limited partners puts losses at 15 cents for every dollar invested and its Databricks stake at 20% of total AUM, and it uses both figures to argue that institutional venture allocations are too small.
The Product Desk · Product desk
What happened
- A16z wrote in a post addressed to limited partners that losses on investments across the funds it manages come to 15 cents for every dollar invested.
- The same post says the firm's stake in Databricks alone represents 20% of its total assets under management.
- It puts typical venture allocations at 5-10% of an overall portfolio, with endowment exposure running between 25% and 40%, some of it through appreciation rather than money actively deployed.
- A16z tells institutional investors who kept venture small inside their alternatives allocation that this is a call to resize again.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- decision The author says they meet an LP with nearly zero exposure to SpaceX, Anthropic and OpenAI every week. Those committees are now deciding whether to buy that exposure at today's marks, after the appreciation the post describes.
- exposure Founders already holding a16z money sit outside the errors the firm says it treats as fatal, which are picking the wrong company in a hot category and passing on a SpaceX, an OpenAI or a Stripe.
- precedent A16z is making its case for more venture with buyout's numbers, putting average and top quartile returns near 15-year lows into the next allocation argument between managers.
Limited partners are the audience for this post, and the ask is money. A founder who closed a round last month is not who it was written for.
The two figures at the top of it are measured against different things. Fifteen cents of loss is quoted per dollar invested, so it is a cost measure [1]. Twenty percent is quoted against total assets under management, so it is a mark [2]. Set one beside the other and the question of how much of the firm's realised money came from one company stays open.
SpaceX at roughly $2T, Anthropic at $965B and OpenAI at $852B add to about $3.82T, which is the floor of the $3.8T to $5T range a16z gives for equity value built largely in private markets [6][7][8][9][17]. Getting to the ceiling requires the rumoured $2T Anthropic listing; substitute that and the total reaches about $4.85T [7][17]. A16z also calls SpaceX the largest venture-backed IPO by a factor of ten, which puts the previous record somewhere near $200B [6][16].
On the founder's side of the table, the post is specific about what the firm fears. Term sheets and follow-on commitments do not come up. The author relays Marc Andreessen's version: a company that goes under eventually stops being mentioned, while a company you passed on keeps appearing in the press and in your nightmares [4].
The comparison it reaches for is buyout. The post quotes Robert Smith, CEO and founder of Vista Equity Partners, who it says famously once said: "Software contracts are better than first-lien debt" [10]. Its answer to that is that the next wave of disruption can make recurring revenue less durable than expected [19].
The share of current AUM sitting in the single largest position is a figure founders can ask any fund they talk to this quarter to produce. A16z has published its own. A manager whose value is concentrated in one holding has a reason to defend that holding, and the same post tells allocators that about a third of technology companies valued at $150B or more are privately held [12].
What to watch
- Whether Anthropic actually lists near $2T, since the top of a16z's own $5T figure depends on it.
- Whether a16z ever publishes Databricks' share of invested cost alongside its share of current AUM.
- Whether any endowment or foundation responds by raising a published venture target above the 10% band.