Craft Ventures (David Sacks)
David Sacks (co-founder & general partner; White House AI & Crypto Czar Dec 2024–Mar 2026)
Early/growth B2B SaaS and AI-application VC. AI exposure runs through application-layer and AI-infra bets (Glean, Replit, ElevenLabs, CrewAI, Supabase, Horizon3.ai) plus compute/space/defense via SpaceX and counter-drone robotics (Allen Control Systems). Sacks pairs a deregulatory AI/crypto policy stance with a 449-AI-company portfolio flagged for conflicts during his government tenure.
- Not open — Not an open-ended vehicle — analysis only, not an available allocation.
- No public record — No published return series — performance cannot be verified in either direction.
- Self-reported — Headline returns are manager- or firm-reported and unaudited.
Top holdings
not disclosed — Space/compute. In Craft's current portfolio; Sacks also holds a personal angel position from a 2002 $27M-valuation check (disclosed, not market-weighted — pre-dates Craft's 2017 founding).
AI enterprise search / work assistant. Craft co-led the 2024 Series C — flagship AI-application bet.
not disclosed — Defense robotics / autonomous counter-drone weapons. Craft participated in the $200M Series B at a $2.2B valuation (2025–26).
AI coding agents / browser dev environment — AI-application + dev-infra exposure.
Open-source backend-as-a-service / AI-app infra. Disclosed as a Sacks angel investment (2024), not market-weighted.
AI voice generation / generative-AI infra.
~7.8% ownership of BitGo — Crypto custody, not AI/compute. Cited in COI scrutiny — Craft reportedly owns ~7.8%; BitGo filed to IPO after the GENIUS
Recent moves
Sacks served as White House AI & Crypto Czar (special government employee) Dec 2024–Mar 2026 under a financial-conflict-of-interest waiver. Per his ethics disclosure he divested all liquid crypto (BTC/ETH/SOL, Coinbase, Robinhood, Bitwise index) and >$200M in digital-asset positions before taking the role; he says he divested 'hundreds of millions' at a cost to net worth. Reporting (NYT, late 2025) found he/Craft retained ~449 AI-related and ~20 crypto investments. Stepped back to Craft GP role March 2026.
Our take
Multi-sector growth/early VC with deepest AI exposure at the application + AI-infra layer (Glean, Replit, ElevenLabs, Supabase) and meaningful compute/space/defense via SpaceX and Allen Control Systems — not a semiconductor-heavy book. Sacks pairs a publicly deregulatory, anti-'regulatory-capture' AI stance (his Anthropic critique) with a pro-crypto policy posture, which directly intersected his portfolio during his czar tenure. CONFLICT-OF-INTEREST CAVEAT: as 2024–26 AI/Crypto Czar he shaped policy (e.g. the GENIUS Act) touching sectors where Craft holds ~449 AI and ~20 crypto positions, including ~7.8% of BitGo, which IPO'd after that act; subject to a congressional ethics inquiry — treat all stake/AUM/divestment figures as self-reported or press-estimated, not market-weighted.
The point marketing won't volunteer: Craft is a private fund with zero public performance — no NAV, IRR, DPI or TVPI.
The headline record (26 unicorns / 15 IPOs; 13 holdings marked >$1B) is self-reported and unauditable, and none is cash returned. Value is almost certainly power-law-dominated by SpaceX — which IPO'd Jun 2026 (SPCX) — the exit happened but the mark went the wrong way: SPCX closed $108.37 on 31 Jul 2026, ~20% below its $135 IPO price, and the first lockup tranche unlocks 6 Aug 2026. The 2023 vintage sits early in the J-curve, so attribution to skill vs a few marks vs 2020–21 vintage beta isn't knowable from public data. The real differentiator is governance: as czar, Sacks operated under a broad OGE waiver an ethics scholar (Kathleen Clark) called a 'sham'; the ~7.8% BitGo stake alongside the GENIUS Act stablecoin framework is the cleanest documented policy-portfolio overlap, and a congressional inquiry exists. Sacks says OGE cleared him and denies conflicts.
Thesis
Craft Ventures is an early/growth B2B-SaaS + applied-AI VC (~$3.3–3.5B self-reported AUM across four early + two growth funds; latest 2023 vintage $1.32B — Craft IV $712M, Growth II $608M). AI exposure sits at the application/infra layer (Glean, Replit, ElevenLabs, Supabase) plus SpaceX (IPO'd Jun 2026, Nasdaq: SPCX) and counter-drone defense (Allen Control Systems), not the chip/compute layer. Its distinctive — and contested — feature is founder David Sacks's Dec-2024–Mar-2026 White House AI/Crypto Czar tenure overlapping a reported ~449-AI-company book.
Concentrated operator-brand VC in B2B SaaS and application-layer AI, power-law dependent on a few marquee marks (SpaceX above all). The two-fund structure (early Ventures + later Growth) lets it follow winners into growth rounds rather than dilute. AI is played through apps and infra, deliberately sidestepping the capex-heavy, commoditizing GPU/compute layer.
Assessment
- PayPal-mafia network + operator brand generates genuine early access — the historical book (Slack, Airbnb, Affirm, Reddit, SpaceX) reflects real proprietary dealflow.
- Two-fund early+growth structure lets it double down on winners (disciplined power-law follow-on) rather than spray capital.
- Application/infra-layer AI tilt sidesteps the capital-intensive, commoditizing GPU/compute layer where returns are hardest to defend.
- SpaceX anchor plus counter-drone defense (Allen Control Systems) gives space/defense exposure few SaaS-focused funds carry.
- No public performance at all — no IRR/DPI/TVPI; the '26 unicorns / 15 IPOs' record is self-reported marketing, not audited fund returns.
- Top marks are largely unrealized: SpaceX is now public (SPCX) but lockup-restricted to ~Dec 2026, so it's a mark-to-market position, not cash returned; realized DPI on recent vintages is unknown.
- Acute keyperson risk: brand, dealflow and now political controversy all attach to Sacks personally.
- Governance/COI overhang: broad OGE waiver criticized as 'sham' by an ethics scholar; congressional inquiry; documented BitGo / GENIUS Act overlap (Sacks denies conflict).
- 2023 vintage still early in the J-curve; ~250–330 holdings per aggregators, but outcomes concentrate in a handful of names.
- 2026-08Update, Aug 2026: the SpaceX lockup begins releasing 6 Aug 2026 — ~911.5M shares, ~$116-123B, about 20% of insider stock — with further tranches through Aug and Sep; 8 Dec 2026 is the final release.
- 2026-07Update, Jul 2026: that anchor cut the other way — SPCX fell 27% in July to below its $135 IPO price after Starship Flight 13's 16 Jul launch abort erased ~$100B in a day, with the S&P 500 at -0.1%.
Record
No public return series exists — a private VC with no NAV or ticker. Craft self-reports 26 unicorns, 15 IPOs and 31 acquisitions across its history (Slack, Airbnb, Affirm, Reddit, Lyft, Twilio) and 13 current portfolio companies marked above $1B — all self-reported and unaudited, none a fund-level IRR or realized DPI. The dominant value driver is almost certainly SpaceX, which IPO'd Jun 2026 on Nasdaq (SPCX) at ~$1.8T and traded near ~$2T soon after; that converts a former private secondary into a public mark, but Craft's stake is lockup-restricted, with the FIRST tranche unlocking 6 Aug 2026 (~$116-123B) and the final release 8 Dec 2026, so it is still a mark-to-market position rather than cash returned. The latest 2023 vintage (Craft IV + Growth II) sits early in the J-curve, so realized DPI is likely low and TVPI mark-dependent.
- 2026-07Update, 31 Jul 2026: SPCX closed $108.00, ~20% under the $135 IPO price and ~52% below the 16 Jun high of $225.64; market cap ~$1.48-1.53T against ~$2.1T at the 12 Jun day-one close. July: -27%.
Risks & fit
- Application-layer AI bets face margin compression as foundation models absorb app features — Glean, ElevenLabs and Replit are exposed to model-vendor encroachment.
- A broad AI-valuation compression would re-rate the paper marks the TVPI depends on.
- Political/reputational exposure: an administration change or an adverse ethics finding could impair brand and dealflow.
- Single-name dependence: SpaceX's post-lockup selling and price dominate fund outcomes; a drawdown from ~$2T re-rates the stake.
- Illiquidity — recent vintages and the SpaceX lockup mean realized DPI may lag public/paper marks for years.
If Craft published audited fund-level IRR/DPI showing recent vintages returning cash ahead of a venture benchmark, the 'unknowable, mark-dependent, self-reported' critique weakens materially. SpaceX has IPO'd well above prior ~$350–800B secondary marks, self-triggering part of the exit clause — if Craft distributes SpaceX proceeds post-lockup at/above the carrying mark and returns don't collapse to one name, the power-law-fragility concern is falsified. A completed OGE/congressional review finding no actionable conflict would neutralize the overhang as an ethics (not performance) matter.
An analytical read for someone studying operator-brand VC and the governance/keyperson dimension of a politically entangled fund. Craft is a private VC — its funds are open only to qualified LPs, not retail; this is a critique of strategy and disclosure quality, not a solicitation, and implies no ability or suggestion to invest.
Not disclosed. Assume standard venture economics (~2% management fee, 20% carry; growth vehicles sometimes 1.5–2%). Craft does not publish terms — treat as unverified.