US Government Strategic Equity (Intel stake + de-facto SWF + Thrift Savings Plan)
US federal government — Dept. of Commerce (Intel equity from CHIPS conversion); a proposed/de-facto US sovereign wealth fund per the 2025 executive order (advocated by NEC's Kevin Hassett); and the Federal Retirement Thrift Investment Board, which runs the Thrift Savings Plan (TSP) — the de-facto US state retirement book.
The closest thing to a US sovereign AI/semi book has three strands. (1) A direct 10% US-government equity stake in Intel (Aug 2025), converted from $8.9B of CHIPS/Secure-Enclave awards — a literal state holding in a leading-edge foundry, with a warrant for +5% if Intel loses foundry majority. (2) A proposed US sovereign wealth fund (2025 EO) that today is an ad-hoc bundle of stakes (Intel, a US-Steel golden share, TikTok, seized crypto). (3) The TSP's C Fund — an S&P 500 index whose top holdings are Nvidia, Broadcom, Apple, Alphabet and Microsoft — making federal employees' $1T+ the de-facto US state book most exposed to the AI trade.
- Not open — Not an open-ended vehicle — analysis only, not an available allocation.
- Private marks — Material value sits in private marks that have not been exit-tested.
Top holdings
~10% government stake — Direct US-gov equity from $8.9B CHIPS conversion (Aug 2025); passive, no board seat; warrant for +5%
top S&P 500 weight — Largest holding of the ~$1T TSP C Fund index
top-5 C Fund — AI-semi exposure inside federal retirement book
top-5 C Fund — S&P 500 mega-cap AI exposure held on behalf of ~7M federal participants
ad-hoc — Other strands of the proposed US SWF; not AI/semi but part of the 'US state book' framing
Recent moves
Aug 22, 2025: US took a 10% Intel equity stake by converting $8.9B of CHIPS + Secure-Enclave awards into common shares (passive, no governance, +5% warrant). 2025 executive order stood up a de-facto US sovereign wealth fund. June 2026: reporting that the TSP C Fund may gain SpaceX exposure as private names enter index vehicles (FedSmith).
Our take
Why it matters: this is the page's 'US government holdings' entry — the Intel stake is an unprecedented direct state ownership of a leading-edge chipmaker, and the TSP is the largest pool of US-state-adjacent capital riding the AI mega-caps (NVDA/AVGO/AAPL top the C Fund). Edge/limits: it is a constructed composite, not one legal fund — the Intel stake, the proposed SWF, and the TSP have different owners, mandates and disclosure. Honest caveat: there is no consolidated NAV or return; the Intel position is explicitly passive (no board seat), the 'US SWF' is still largely aspirational/ad-hoc, and the TSP is a passive index vehicle, not an active AI bet — its AI exposure is incidental to S&P 500 weighting.
Three things the headline gain obscures. First, the entry price was a distress artifact: $20.47 was near Intel's multi-decade low, and the state got it because Intel needed the money — bailout economics, not stock selection.
Second, and sharpest: the US government is simultaneously Intel's subsidizer, regulator, national-security customer and 10% shareholder. When a reported preliminary Apple foundry deal repriced the stock ~14% on 8 May 2026, the state was marking a position whose value it can partly steer. That is not alpha — it is a profitable conflict of interest, and it is unaudited. Third, the conversion had a cost nobody scores: $8.9B earmarked to build fab capacity became equity instead. Taxpayers own more Intel and financed less silicon. The warrant makes this explicit — an anti-divestiture device, industrial policy dressed as an instrument, that only pays if the policy fails.
Thesis
The US state has become one of the largest equity holders in its own strategic sector — without a charter, a benchmark, or an exit policy. The Intel stake alone went from $8.87B to roughly $44B in under two years (433.3M shares at $20.47, Aug 2025; INTC ~$101.90 on 21 Jul 2026), a ~5x that dwarfs any sovereign fund's record. Our read: that return is a policy artifact, not an investment process, and the absence of governance around it is the story — not the P&L.
Three unconsolidated strands, no single entity. (1) Commerce converted $5.7B CHIPS + $3.2B Secure Enclave awards into 433.3M Intel shares (~10%), plus a 5-year warrant on +5% at $20 struck only if Intel falls below 51% foundry ownership. (2) Ad-hoc agency positions — DoD's ~15% of MP Materials, DOE's 5% of Lithium Americas plus 5% of its GM Thacker Pass JV, a US Steel golden share (veto, not equity); a partial list. (3) The TSP's C Fund ($507.9B, May 2026), S&P 500 index.
Assessment
- Entry discipline was accidental but real: converting grants at $20.47 into equity captured upside a pure subsidy would have handed entirely to existing shareholders.
- The stake is structurally passive — no board seat, no disclosed voting arrangement — which limits the worst governance failure mode of state ownership.
- The +5% warrant at $20, triggered only if Intel loses foundry majority, ties the position to the actual policy goal (domestic leading-edge capacity), not the share price.
- TSP is genuinely well-built: 0.035% total C Fund expense (2025) across 7.3M accounts is cost discipline most defined-contribution plans never reach.
- No consolidated NAV, no published mark-to-market schedule, no investment committee, no benchmark. Norway's NBIM discloses holdings continuously; here no single entity even reports the book.
- Governance terms differ deal by deal because each was struck by a different agency (Commerce, DoD, DOE) under separate authority. That is a pile of bespoke contracts, not a portfolio.
- Exit is politically trapped. Selling a ~5x invites 'you sold the national champion'; holding concentrates ~$44B of public money in one volatile single name with no stated sell discipline.
- The state is long the same AI trade three ways — Intel equity, the C Fund's mega-cap weighting, and tech-derived tax receipts. That correlation is undiscussed and unhedged.
- Grant-to-equity conversion means capacity that was funded is now unfunded. The gain and the capacity shortfall are never reported against each other.
Record
Attribute carefully. The Intel position is up ~4.9x — 433.3M shares at a $20.47 basis ($8.87B) against ~$101.90 on 21 Jul 2026 gives ~$44.2B and ~$35B unrealized; at the 8 May 2026 intraday high near $130.46 it showed ~$56.5B and ~$47.6B of gain. Almost none of this is manager skill: (a) a distressed entry price obtained through the leverage the state holds as funder, (b) a company-specific catalyst — the reported Apple foundry deal — in an environment the same government shapes via tariffs, export controls and procurement, and (c) sector beta. The TSP side has no active return by construction: the C Fund tracks the S&P 500, so its AI exposure is index weighting, not a view. No combined return exists because no combined entity does, and Treasury publishes no P&L on the Intel position.
Risks & fit
- Single-name concentration: a 30-40% INTC drawdown erases roughly $13-18B of paper public wealth with no hedge and no stated risk limit.
- Overhang on exit — 433.3M shares is ~10% of the float; any credible sale signal pressures the stock the state is trying to monetize.
- Precedent risk: equity-for-subsidy becomes the default template, and future conversions may be struck at far worse entry prices without the distress discount.
- Conflict of interest — procurement, export policy and tariff decisions benefiting a company the state owns are hard to distinguish from industrial policy.
- TSP politicization: pressure to admit non-index or private holdings into a federal plan would compromise its passive, low-cost premise.
The 'policy luck, not process' read breaks if the government publishes a consolidated register of its stakes with a stated mandate, an investment committee, a benchmark and a written exit policy — then exits Intel on that policy rather than a political window. It is confirmed if the position is still unmarked a year from now, or sold in a burst timed to a fiscal or electoral need. The test on skill: if the next agency conversions are struck at non-distressed entry prices and still outperform, that is a repeatable process rather than a one-off bailout multiple.
Analytically relevant to anyone modeling semiconductor policy risk, sovereign-balance-sheet mechanics, or the overlap of industrial policy and equity markets. Note plainly: none of this is investable. The Intel stake is held by the US Treasury; the agency stakes are statutory; the TSP is open only to federal employees and uniformed service members. The only public-market expression is INTC itself, a separate question from whether the state's ownership structure is sound.
No management or performance fee on the Intel or agency stakes — there is no manager, which is also why there is no risk framework. TSP is the opposite: C Fund total expense 0.035% in 2025 (~$0.35 per $1,000), among the lowest in US defined-contribution.