Nancy Pelosi — personal/spousal disclosures (Paul Pelosi book)
Rep. Nancy Pelosi (D-CA); trades executed by spouse Paul Pelosi, a SF venture/real-estate investor. Disclosed under the STOCK Act PTR regime.
The single most-tracked congressional book and a near-archetype of the AI/compute/power trade: concentrated, long-dated call options on mega-cap AI semis and the power names that feed datacenters, often exercised deep in-the-money. The 2026 repositioning rolled big-tech equity into fresh calls on AI compute (NVDA), hyperscalers (GOOGL, AMZN), AI-security (PANW), AI-health (Tempus), plus a clean datacenter-power bet via Vistra (VST) calls exercised into stock.
- Disclosure record — A public official's statutory filings, not a fund. No NAV, no manager, no terms.
- Reconstructed — Figures are third-party estimates rebuilt from filings, not reported results.
Top holdings
long (calls + shares) — Core AI-compute holding; Dec-2025 sold 20k shares ($1M-$5M), Jan-2026 exercised 50 calls (strike $80) into 5k shares ($250K-$500K).
~$500K-$1M (50 calls→5k shares) — Datacenter-power / nuclear-adjacent IPP; Jan-2026 exercised 50 VST calls (strike $50) into 5,000 shares — the clearest AI-power-demand bet.
AI networking/custom-silicon; Jun-2025 exercised 200 calls into 20k shares at ~71% discount to market.
Hyperscaler / AI model + TPU exposure; new call options in the Jan-2026 disclosure.
AWS datacenter capex proxy; new calls in Jan-2026 disclosure.
AI-security layer; new calls in Jan-2026 disclosure.
AI-driven precision medicine; new calls in Jan-2026 disclosure — a higher-beta AI bet.
Recent moves
Jan-2026 disclosure (filed ~Jan 23, 2026, covering late-Dec-2025→mid-Jan-2026): repositioned ~$69M across big tech — sold 20k NVDA shares, exercised NVDA + VST calls into stock, bought 25k AllianceBernstein (AB) shares ($1M-$5M), and opened new calls on GOOGL, AMZN, PANW, TEM (Capitol Trades, Jan 27 2026).
Our take
Why it matters: Pelosi's book is the public face of the AI-infrastructure trade — long-dated ITM calls on compute (NVDA, AVGO), hyperscalers (GOOGL, AMZN), AI-security (PANW), and crucially datacenter power (VST), exactly the value chain this page tracks. Edge/limits: the options structure means the equity-only view understates leverage and timing; PTRs are wide $-ranges, lagged ~30-45 days, and are her spouse's trades, not hers. Honest caveat: trackers' eye-popping return figures are estimates, not audited, and this is transparency tracking — not a buy signal or endorsement.
Most disclosed names map onto one theme with three layers: compute (NVDA, plus the May-2026 INTC calls), the platforms consuming it (GOOGL, AMZN), and the electricity feeding it (VST).
The Jan-2026 filing was a repositioning, not an exit — AAPL and NVDA shares sold alongside 50-contract exercises in NVDA ($80 strike), GOOGL ($150), AMZN ($150), VST ($50) and TEM ($20) on 16 Jan 2026, plus 25,000 AllianceBernstein units. The theme is not exclusive: the same May-2026 filing carrying INTC also carries 200 UBER calls, $50 strike, 19-Mar-2027 expiry, $500,001–$1,000,000 — a name outside the AI-capex chain. What a reader can conclude: direction, instrument, expiry and an approximate scale band per disclosed trade. What a reader cannot: position size, portfolio weight, cost basis, what was held and not traded, or any return. PTRs report transactions, not holdings — no denominator appears anywhere.
- 2026-07July 2026 split the three layers apart: NVDA +0.3% and AVGO +3.1% while INTC fell 35.4% on the month and VST sits -11.2% YTD and -27.5% over a year. Silicon and power are separate risks, not one theme.
Thesis
Not a fund — a public official's STOCK Act disclosure record. Rep. Nancy Pelosi (D-CA) files Periodic Transaction Reports covering a household book traded by her spouse, Paul Pelosi, a San Francisco venture and real-estate investor. The filings show a concentrated, options-heavy book expressed largely through long-dated call options on mega-cap technology, AI compute and the power infrastructure supporting it, plus adjacent names outside that chain. There is no NAV, no audited return, and no investable vehicle.
As reported: single-name call options with 12–18 months to expiry, frequently exercised into stock rather than sold; occasional outright equity blocks and fund units; low trade frequency in bursts rather than continuous turnover. Recent disclosed cadence: a large multi-name filing signed 23 Jan 2026, then roughly five months with no disclosed trades, then a two-name INTC/UBER filing covering 29 May 2026 trades.
Assessment
- Instrument choice is internally consistent: long-dated calls exercised into stock rather than sold, expressing a multi-year rather than short-dated view.
- Most names form a legible value chain — silicon, hyperscaler, datacenter power — though UBER sits outside it.
- Disclosure is granular by congressional standards: per-trade dates, strikes, expiries, contract counts and value bands.
- Trade cadence is low and clustered — the Jan-2026 filing was followed by roughly five months with no disclosed transactions.
- PTRs report value RANGES ($1,000,001–$5,000,000 on the May-2026 INTC calls), never an amount. Any point estimate is a reconstruction.
- Statutory lag is normal filing mechanics: the 29 May 2026 trades were signed 23 June and filed later that month, inside the 45-day statutory window.
- Trades are executed by her spouse and disclosed as household assets; attribution to the official is a filing requirement, not a statement of who traded.
- No denominator: PTRs disclose transactions, not the full book, so weights, concentration and cash are unknowable from the filing.
- Tracker reconstructions rest on assumed fill prices and sizes within each band, and cannot price the options component accurately.
Record
No official performance figure exists — PTRs contain no NAV. The widely circulated estimates originate with a single source, Unusual Whales' annual Congress Trading Report: a Pelosi portfolio return of +70.9% for 2024 (29 Dec 2023 to 30 Dec 2024) against the S&P 500's +24.9%, and 65% for 2023 against +24.8%. Motley Fool, Capitol Trades, Quiver and others republish that figure rather than each producing it independently. The method must assume an entry price within every disclosed range and a portfolio weight the filing never provides; the options component in particular cannot be reconstructed accurately from range data. The benchmark basis is the S&P 500, not risk- or leverage-adjusted against a book containing long-dated calls. Treat these as one reconstruction of disclosed trades, not a measured return.
- 2026-08The estimates are not all downstream of one source. PolyTICK builds its own series and lands far apart: 2024 +45.6% against Unusual Whales' +70.9%, 2025 +54.1% against +20.1%. Dispersion is the deeper problem.
Risks & fit
- Theme concentration: most disclosed names sit on one AI-capex chain, so they share a single demand driver.
- Options structure means the disclosed exposure is time-bounded — a Jan-2027 or Mar-2027 expiry either converts or lapses.
- Datacenter-power exposure (VST) carries regulatory and interconnection-queue risk distinct from the semis it is paired with.
- Single-name idiosyncratic risk is undiversified by construction: a handful of tickers, no index sleeve disclosed.
- Disclosure itself ends: Pelosi has announced she is not seeking re-election, so the STOCK Act filing obligation lapses after her term.
A future PTR showing sustained selling across the compute and power names, or further positions outside that chain, would break the single-theme reading — the May-2026 UBER calls already qualify the theme rather than confirm it. An annual OGE Form 278e showing large holdings outside the traded names would show the PTR-derived picture is a small, unrepresentative slice of the household book, falsifying the concentration reading directly. And if the Jan-2027 and Mar-2027 expiries pass without exercise or roll, the multi-year-view reading weakens.
Useful as theme and market-structure context: a date-stamped public record of how one AI-capex thesis was expressed across compute, hyperscaler and power layers, with instrument and expiry visible. Also a case study in what disclosure regimes do and do not reveal. It is NOT a signal to mirror — a lag of weeks up to the 45-day statutory window, value ranges, a missing denominator and options structure make replication impossible in principle — and nothing here is investment advice.
Not applicable — a personal/household disclosure record, not a managed fund. No management fee, no performance fee, no outside capital. Not investable.