Synthesis

Fund maths for seed investors

How seed funds think about portfolio construction, cheque size, ownership and pro rata rights when valuations inflate, and how LPs and new vehicles fit in.

Type
Synthesis
Confidence
Medium
Sources
10 videos
First published
Updated

Six episodes touch fund mechanics, and nearly everything is Sam’s or Brit’s view from inside seed funds. Sam says OpenAI seed investors sit at about 25x, which is not a fund returner for a seed fund, and Dave says a roughly $1 million Anthropic core cheque in 2023 would not have returned a seed fund either, though about $80 million is ‘not nothing’ (neither figure is verified; Sam says he has not run the numbers on one case). Brit says portfolio construction is breaking under inflated seed valuations, so she thinks of percentage of the fund per bet rather than percentage of ownership, and LPs and GPs argue over marks. In one October episode, Sam says pro rata rights are critical to Slow’s model and must be made explicit, citing an early case when a large VC (an LP) tried to waive Slow’s pro rata. Brit says big Series A funds renegotiate and the only protection is a good founder relationship. She also describes Megalith, a vehicle that buys pro rata rights seed funds cannot or will not use, and a 64 percent LP co-invest figure with no source. Sam calls private markets ‘treaty markets’ with no standards. An unattributed host warns against building a fund around a post-AGI world.

Key points

  • Seed outcomes in the biggest AI names are said not to return a seed fund: OpenAI seed about 25x (Sam), a 2023 Anthropic core cheque about $80M on roughly $1M (Dave). — 1 episode. (Sam · The Seed Market Is Broken: Meg (Jan 2026) ▶ 08:44)
  • Brit measures percentage of the fund per bet rather than percentage of equity, because portfolio models break when cheques and valuations jump. — 1 episode. (Brit · Allbirds Pivoted to AI Data Ce (Apr 2026) ▶ 39:42)
  • Pro rata rights are core to Slow’s seed model; Sam says get them in writing and tell founders you will enforce them, because later-stage funds try to waive them. — 1 episode. (Sam · Silicon Valley Is Great at Tec (Oct 2026) ▶ 42:02)
  • New LP-side structures are reported by Brit: Megalith buys pro rata rights for about 1.25 percent of carry, and 64 percent of LPs want co-invest rights (source not given). — 1 episode. (Brit · Silicon Valley Is Great at Tec (Oct 2026) ▶ 40:26)
  • LP appetite is described as weak (denominator effect, few LPs at a summit) and GP-LP relations as treaty-like and unstandardised. — 2 episodes. (one of the hosts · Apple, OpenAI & Why Nobody Mak (Feb 2026) ▶ 09:11)
  • Fund strategies built on a single macro bet (post-AGI) are called risky by an unattributed host, with Brit arguing physical AI may be the revolution. — 1 episode. (one of the hosts · AI Frontier Lab Is Essentially (Jul 2026) ▶ 48:11)

Frameworks

  • Brit’s fund math under inflated seed valuations — Brit says portfolio construction models break when cheque sizes and valuations jump, so she measures the share of the fund per bet and discusses marks with LPs. (Brit · Allbirds Pivoted to AI Data Ce (Apr 2026) ▶ 38:35)
    • Forecast cheque sizes and counts for the fund
    • Decide whether to pay up for higher-valuation seeds
    • Measure percentage of the fund per bet
    • Align the conversation with LPs on marks

Tactics

  • Buy a small slice of a weird idea you will not work on — Sam recently bought 10 percent of two companies for a quarter million dollars each because they were interesting, weird ideas he helped shape and does not need to own outright. (Sam · The Seed Market Is Broken: Meg (Jan 2026) ▶ 06:35)

Stories & examples

  • A large VC tries to waive Slow’s pro rata: Early in Slow’s history a large VC and LP tried to say Slow did not need its pro rata on a deal. Lesson: Make pro rata expectations explicit and enforceable. (Sam · Silicon Valley Is Great at Tec (Oct 2026) ▶ 41:46)

Numbers

  • OpenAI seed investors’ multiple: 25 x return (host’s framing, approx.) (Sam · The Seed Market Is Broken: Meg (Jan 2026) ▶ 08:44)
  • Proceeds from a core Anthropic seed check taken in 2023, per Dave: 80000000 currency unstated (from a roughly 1 million check) (host’s framing, approx., currency unstated) (Dave · The Seed Market Is Broken: Meg (Jan 2026) ▶ 10:11)
  • Slow’s cheque size: 1 to 3 million dollars (host’s framing) (Sam · The Seed Market Is Broken: Meg (Jan 2026) ▶ 07:47)
  • Slow’s target ownership: 10 % ownership (host’s framing) (Sam · The Seed Market Is Broken: Meg (Jan 2026) ▶ 07:49)
  • Share of LPs wanting co-invest rights (Brit, source unstated): 64 percent (host’s framing, approx.) (Brit · Silicon Valley Is Great at Tec (Oct 2026) ▶ 44:22)
  • Megalith’s share of the carry: 1.25 percent (host’s framing, approx.) (Brit · Silicon Valley Is Great at Tec (Oct 2026) ▶ 43:32)

Caveats & warnings

  • Sam says he could be wrong on the OpenAI seed math and invites correction. (Sam · The Seed Market Is Broken: Meg (Jan 2026) ▶ 08:37)
  • Sam says he has not actually run the numbers on his former colleague’s Anthropic equity. (Sam · The Seed Market Is Broken: Meg (Jan 2026) ▶ 11:00)
  • Host has a stake: Brit says her fund invests in Coco Robotics and is about to raise another fund. (Brit · Allbirds Pivoted to AI Data Ce (Apr 2026) ▶ 05:27)
  • One speaker (Sam, Dave or Brit) says some LPs, often from finance, are personally affronted when GPs make money. (one of the hosts · Silicon Valley Is Great at Tec (Oct 2026) ▶ 44:57)

Also covered here: Capital intensity and efficiency

Six episodes, but the evidence is thin and comes from three people. Sam says a seed investor does not want to be in ‘super capital-consumptive’ things and should stay in the fairway of small cheques for about 10 percent. In June he adds that after the SpaceX IPO he expects a ‘Mark effect’ of more ambitious, capital-hungry startups alongside a group going the other way, with the middle dropping out. Stuart Landesberg, a founder, describes planning a hardware business to be capital efficient early (prove the concept on modest capital) and capital consumptive late, which he says fits the venture model. Suds, a guest investor, avoids the app layer and backs infrastructure, energy and data centres that extend the frontier labs’ revenue. Peter Deng says over 20 percent of Felicis capital in recent funds went to resilience, energy, manufacturing and defence tech, which shows capital-heavy sectors attracting some funds. Mike Brown’s capital allocation point (acquisitions first, organic growth with one to two year paybacks) is a holding-company version.

Key points

  • As a seed investor, avoid capital-consumptive companies and stay with small cheques for about 10 percent (Sam). — 1 episode. (Sam · The Seed Market Is Broken: Meg (Jan 2026) ▶ 08:57)
  • Plan to be capital efficient early and capital intensive late, to match how venture money arrives (Stuart Landesberg). — 1 episode. (Stuart Landesberg · Stuart Landesberg (Oct 2025) ▶ 28:16)
  • Money is flowing to capital-heavy sectors (resilience, energy, manufacturing, defence), and Sam expects a split between capital-hungry and capital-light startups with ‘the middle’ dropping out. — 2 episodes. (Peter Deng · Peter Deng & Rachel Holt (Aug 2026) ▶ 49:58) (Sam · SpaceX Hit $2.5 Trillion. Now (Jun 2026) ▶ 16:22)
  • Suds prefers infrastructure and energy that extend the labs’ revenue over the app layer. — 1 episode. (Suds · Suds (Sep 2026) ▶ 31:59)
  • A holding company’s analogue is capital allocation: Teamshares reinvests where payback is one to two years. — 1 episode. (Mike Brown · Mike Brown (Jun 2026) ▶ 41:49)

Where guests disagree

Should a seed investor avoid capital-hungry companies? ⚠️

  • Sam says seed funds should avoid super capital-consumptive things. (Sam · The Seed Market Is Broken: Meg (Jan 2026) ▶ 08:57)
  • Peter Deng reports Felicis put over 20% of recent fund capital into energy, manufacturing, resilience and defence. (Peter Deng · Peter Deng & Rachel Holt (Aug 2026) ▶ 49:58)

Tactics

  • Design the business plan to prove concept on modest capital and consume capital late, to match the venture model. — Proof of concept on a relatively modest amount of capital; scale-out later. (Stuart Landesberg · Stuart Landesberg (Oct 2025) ▶ 28:16)

Numbers

  • Felicis capital to resilience, energy, manufacturing, defence: 20 percent (over) (self-reported, approx.) (Peter Deng · Peter Deng & Rachel Holt (Aug 2026) ▶ 49:58)

Caveats & warnings

  • Peter says over 20% of Felicis capital in recent funds went to global resilience, energy, manufacturing and defence tech. (Peter Deng · Peter Deng & Rachel Holt (Aug 2026) ▶ 49:58)
  • Build the business so it still reaches its long-term goal whether or not the current AI and robotics trends are overhyped. (Stuart Landesberg · Stuart Landesberg (Oct 2025) ▶ 31:22)

How much to trust this

  • Evidence: 11 episodes, 5 independent speakers (the four hosts and unidentified host speech count as one voice, the show; guests counted individually).

Blueprints that use this

Sources

  1. (01)

    Is Google About to Kill Canva? │Claude Goes Rogue, Tesla vs. Waymo

    YouTube

    31:59 into “Is Google About to Kill Canva? │Claude Goes Rogue, Tesla vs. Waymo”

  2. (02)

    Teamshares IPO: $60M EBITDA, 92 Companies, Zero Exits

    YouTube

    38:13 into “Teamshares IPO: $60M EBITDA, 92 Companies, Zero Exits”

  3. (03)

    The Seed Market Is Broken: Mega-Rounds, Narrative Games, and Walkouts

    YouTube

    6:35 into “The Seed Market Is Broken: Mega-Rounds, Narrative Games, and Walkouts”

  4. (04)

    Allbirds Pivoted to AI Data Centers | SpaceX $1.5T IPO, Anthropic $800B, $175M Seed Rounds

    YouTube

    5:27 into “Allbirds Pivoted to AI Data Centers | SpaceX $1.5T IPO, Anthropic $800B, $175M Seed Rounds”

  5. (05)

    Apple, OpenAI & Why Nobody Makes Money in AI | The SaaS Apocalypse

    YouTube

    7:49 into “Apple, OpenAI & Why Nobody Makes Money in AI | The SaaS Apocalypse”

  6. (06)

    The Risk Of Securitizing GPUs Like Long-Lived Infrastructure & More - A MOL Special Team Edition

    YouTube

    49:58 into “The Risk Of Securitizing GPUs Like Long-Lived Infrastructure & More - A MOL Special Team Edition”

  7. (07)

    SpaceX Hit $2.5 Trillion. Now Watch What Elon Buys Next.

    YouTube

    16:22 into “SpaceX Hit $2.5 Trillion. Now Watch What Elon Buys Next.”

  8. (08)

    Silicon Valley Is Great at Tech and Terrible at Business

    YouTube

    40:26 into “Silicon Valley Is Great at Tech and Terrible at Business”

  9. (09)

    AI Frontier Lab Is Essentially Doomed | Open AI Offered U.S. Equity, Xbox Is In Trouble, Post-AGI

    YouTube

    48:11 into “AI Frontier Lab Is Essentially Doomed | Open AI Offered U.S. Equity, Xbox Is In Trouble, Post-AGI”

  10. (10)

    Stuart Landesberg's Plan to Save the American West from Wildfires

    YouTube

    28:16 into “Stuart Landesberg's Plan to Save the American West from Wildfires”

(By email)

Compounding Insights