Screening the Pipeline: How Investment Controls Became the US-China Economic Architecture

01 The Regime That Quietly Replaced the Tariff War

AP reported in September 2026 that US-China investment negotiations remain constrained by national-security concerns, with Washington and Beijing each trying to preserve the economic benefits of integration while insulating themselves from the strategic risks of interdependence. That sentence describes a condition that would have been unrecognizable a decade ago, when capital flows between the two countries were largely treated as commercially neutral. Today, the binding constraint on US-China economic relations is not the tariff schedule. It is the screening stack: a set of legal instruments that examines who may buy what, in which technologies, and with what rights attached to the money.

On the American side, the core instrument is the Committee on Foreign Investment in the United States. Wikipedia describes CFIUS as an inter-agency committee of the US government that reviews the national-security implications of foreign investments in the US economy. That description understates how much the committee's reach expanded after 2018. The Foreign Investment Risk Review Modernization Act, passed as part of the 2019 defense authorization, extended CFIUS jurisdiction beyond controlling stakes to non-controlling investments in critical technology, critical infrastructure, and sensitive personal data — the so-called TID categories. In parallel, the Export Control Reform Act of 2018 gave the Commerce Department standing authority over emerging and foundational technologies, and Executive Order 14105 of August 2023 created an outbound screening regime, finalized by the Treasury Department with effect from January 2025, that requires notification of — and in defined cases prohibits — US investment in Chinese semiconductors, quantum computing, and artificial intelligence.

The structural fact is the symmetry. Both governments now screen capital moving in both directions. That symmetry is what turns a set of discrete regulatory programs into an architecture — a durable, rule-bound interface between two economies that remain deeply integrated in trade and portfolio investment while being administratively separated in direct investment and technology. The chart below tracks the macro consequence: two-way direct investment, which exceeded sixty billion dollars in combined flows at its mid-2010s peak, has collapsed to a fraction of that level, while portfolio flows and trade continue at scale.

02 Portfolio Versus Direct: The Line That Carries the World

The most consequential legal distinction in the US-China economic relationship is the one most observers never think about: the line between portfolio investment and direct investment. Portfolio capital — Chinese holdings of US Treasury securities, equities held through funds and custodians, bonds — is passive by construction. It confers no board seats, no access to source code, no ability to direct research agendas. Direct investment — mergers, acquisitions, greenfield plants, venture stakes with governance rights — confers all of those things. The screening architecture is built almost entirely on the direct side of that line, and it is deliberately porous on the portfolio side.

This asymmetry is not an oversight. It is the political economy of the standoff. As the CNBC Explains video accompanying this article documents, Chinese entities — principally the State Administration of Foreign Exchange and state banks — have at times held well over a trillion dollars in US assets, with Treasury holdings peaking above 1.3 trillion dollars around 2013 before declining to roughly the 770 billion dollar range by 2024 (approximate figures from US Treasury TIC data). Those holdings are a mutual hostage arrangement: they give Beijing a stake in US financial stability and give Washington the comfort that liquid, passive capital does not carry know-how out of the country. No US administration has proposed screening Chinese Treasury purchases. The moment capital becomes controlling or technologically proximate, the calculus reverses.

The outbound regime replicates the same distinction in mirror image. The Treasury final rule implementing Executive Order 14105 applies to transactions that give a US investor access to material non-public technical information, board or observer rights, or decision-making authority over a Chinese venture in the covered technology sectors. It carves out passive positions — index-fund stakes, trading of publicly traded securities, and investments below the ten percent voting threshold without governance rights. The rule, in other words, does not try to stop American money from touching Chinese artificial intelligence companies. It tries to stop American investors from becoming conduits for the technology itself.

03 Does Screening Stop Transfer, or Reroute It?

The empirical question is whether any of this measurably reduces technology transfer. The honest answer is that the observable record shows screening reliably blocking specific transactions while doing much less to block the diffusion of capability. On the blocking side, the record is concrete. CFIUS referrals produced the 2018 presidential order against Broadcom's hostile pursuit of Qualcomm on national-security grounds; forced divestitures of the dating app Grindr and the hotel-software firm StayNTouch by Chinese owners; the unwinding of the China Maritime Navigation deal in Houston; and the multi-year campaign to force a divestiture or restructuring of TikTok's US operations, resolved through statute rather than committee action. These are observed facts: named transactions, named orders, dated outcomes.

On the diffusion side, the record is equally concrete in the opposite direction. Chinese semiconductor fabricators continued to advance process technology after the October 2022 export-control rounds, using stockpiled equipment, domestically developed tools, and third-country channels — a well-documented pattern of delay rather than denial. Export controls did not stop Huawei from returning to the smartphone market with domestically produced advanced chips in 2023, an event widely reported as a marker of control leakage. The mechanisms of rerouting are structural: technology diffuses through open-source publication, talent migration, third-country subsidiaries, licensing arrangements, and contract manufacturing in jurisdictions the controls do not reach. Capital is the easiest channel to close; knowledge is the hardest.

Three competing explanations deserve separation. The first is that screening works but slowly — it raises the cost and time constant of technology transfer without eliminating it, and even a slower adversary is a security gain. The second is that the observed collapse in two-way direct investment is mostly a risk-premium effect rather than a rules effect: geopolitical risk, tariff uncertainty, and pandemic-era disruption would have depressed Chinese investment in the US and American investment in China even without CFIUS. The third is that screening succeeds precisely by being a signal rather than a sieve — its main function is to tell the market that certain deals will not clear, so the deals are never attempted, and the withdrawals that would show up as blocked transactions simply never appear in the data. All three are partly true, which is why the counterfactual matters: without the screening stack, direct investment would almost certainly be higher, but no published estimate credibly isolates how much higher.

04 Beijing's Mirror: Reciprocity and the Exit Problem

The American narrative treats screening as a US initiative and Chinese resistance as the friction. That framing is incomplete. China has built its own screening architecture, and it did so on roughly the same timeline. The 2020 Foreign Investment Law consolidated a negative-list approach while creating a national-security review mechanism explicitly modeled, in function if not in form, on CFIUS-style review. The same year brought the Export Control Law, the Unreliable Entity List, and long-arm data-security statutes that restrict the transfer of data — including source code and technical documentation — outside Chinese jurisdiction. The 2021 Anti-Foreign Sanctions Law gave Beijing legal authority to counter-asset foreign measures, converting sanctions resistance from diplomatic protest into statutory machinery.

The practical consequence is that the exit has become as screened as the entry. Reporting across the Financial Times, Reuters, and Bloomberg through 2024-2026 has documented foreign private-equity funds facing prolonged approvals and, in some cases, capital-gains assessments in the tens of percent of asset value when selling Chinese holdings and repatriating proceeds. These are reported claims, contested by Beijing, but they point to a structural asymmetry in the current phase: Washington's instruments deter money from arriving; Beijing's instruments increasingly deter money from leaving. For a US fund with legacy China exposure, the binding constraint is no longer the buy decision made in 2015 but the sell decision being contemplated now.

Reciprocity also operates at the negotiating table. AP's reporting that investment negotiations remain constrained by national-security concerns suggests both governments treat screening concessions as a bargaining category — side-letters, carve-outs, assurances about specific investors or sectors — rather than as matters that will be dismantled. The 2023-2024 pattern in bilateral management (the re-establishment of working groups on economics and counternarcotics following the November 2023 San Francisco summit) supports the inference that both sides are managing the screening regime, not negotiating its abolition. Historical precedent is instructive here: the Committee on Foreign Investment was created by an executive order in 1975 partly to monitor Arab petrodollar inflows, and every subsequent expansion — Exon-Florio in 1988, FINSA in 2007, FIRRMA in 2018 — has been a ratchet that never turned back. Screening regimes have historically grown and never shrunk; there is no precedent for one being repealed.

05 The Causal Chain and Its Second-Order Effects

Strip the architecture to its causal core and the chain runs like this. The driver is a security assessment: that technology transfer through capital — board access, technical due diligence, management rights over US ventures — is a meaningful vector by which adversary capability advances. The mechanism is transaction-cost inflation: screening converts what was a commercial deal into a regulatory proceeding with uncertain duration, uncertain conditions, and a nonzero probability of forced unwinding after closing. The first-order effect is the observable collapse of two-way direct investment and a migration of deals toward forms and jurisdictions the rules do not reach.

The second-order effects are where the economics get uncomfortable for both governments. Thinner cross-border M&A in technology sectors means fewer exit routes for startups, which compresses venture valuations in exactly the deep-tech segments the controls exist to protect — semiconductors, quantum, AI infrastructure — because US venture funds have broadly retreated from syndicating with Chinese limited partners and Chinese strategic acquirers, removing a class of buyers from the exit stack. It also means efficiency losses: capital that cannot flow to its highest-return use stays domestic, and the literature on FDI consistently finds productivity spillovers flow in both directions. Both societies are paying a static efficiency tax to buy a dynamic security margin, and neither has quantified the trade.

The third-order effect is normative diffusion. The EU adopted a Union-wide FDI screening framework in 2019; Japan strengthened its review in 2020 and added a core-sector list that mirrors FIRRMA's TID categories; the United Kingdom's National Security and Investment Act of 2021 created a mandatory notification regime. Middle powers that once treated investment openness as an ideological commitment now operate security review as routine. China's outbound capital, facing CFIUS-type scrutiny in every advanced economy, has rotated toward Southeast Asia, the Gulf, and Latin America — the Belt-and-Road constituency — while Chinese portfolio money continues to hold US securities. The global system is converging on a two-tier structure: screened direct investment, unscreened portfolio investment, and persistent trade.

06 Three Scenarios: Stabilization, Persistence, Escalation

Scenario one, stabilization. Washington and Beijing negotiate practical de-confliction of the screening regimes — white lists for approved investors, pre-clearance channels for deals below sensitivity thresholds, mutual restraint in entity designations — while preserving the statutory framework. Two-way direct investment recovers modestly from a very low base; the working groups established after the 2023 San Francisco summit expand into investment facilitation. Triggers to watch: a bilateral investment-adjacent side agreement, a named Chinese investor cleared through a new fast-track, or the lifting of specific Unreliable Entity List designations. Probability is the lowest of the three: the political incentives on both sides reward visible toughness, and the ratchet history of screening regimes has no reversal precedent.

Scenario two, persistence. This is the current condition and the modal path. The screening stack stays as built, direct investment remains depressed, portfolio flows and trade continue at scale, and both governments negotiate at the margin — individual deals, individual sectors — rather than at the architecture. Technology competition proceeds through export controls and industrial policy (the CHIPS and Science Act on the US side, Made in China 2025 successor programs on the Chinese side), with investment screening as the fence around the industrial-policy contest. Indicators: annual CFIUS notice counts holding in the several-hundreds range, Treasury outbound notification volumes growing steadily without new covered sectors, and no change in Chinese Treasury holdings behavior.

Scenario three, escalation. The screening perimeter extends into the portfolio tier — restrictions on Chinese holdings of US securities beyond designated military-linked firms, or Chinese retaliation against US portfolio positions — or outbound controls expand to additional sectors such as biotechnology, energy storage, or foundational AI models. A forced-divestiture cycle in either direction, or a Taiwan-contingency sanctions package, would accelerate this. The historical analog for the portfolio tier is the freezing of Russian central-bank reserves in 2022, which demonstrated that portfolio investment is, in extremis, seizable. Triggers to watch: any legislative proposal to screen Treasury purchases, an expansion of the outbound covered-sector list, or the first US order forcing divestiture of a Chinese portfolio stake.

07 Signal Versus Noise: The Counterfactual and the Indicators

There is a confound at the heart of every claim about screening efficacy. Two-way direct investment would have fallen from its 2016 peak regardless — tariffs, COVID-19 supply shocks, Hong Kong's national-security law, and generalized political risk all raised the risk premium on cross-border deals in the absence of any committee action. The screening stack overlays a decoupling trend that has independent causes. Anyone citing the FDI collapse as proof the regime works is reading a confounded time series; anyone citing continued Chinese chip progress as proof it fails is making the mirror-image error, since the counterfactual pace of progress without controls is unobservable. The defensible conclusion is narrower: screening demonstrably blocks named transactions, plausibly slows diffusion, and has not prevented the underlying military-technical competition from intensifying.

The signal in this domain is not the annual flow numbers but the institutional fact: both governments have converted investment review from an ad hoc instrument into standing, rule-bound machinery with annual reporting, defined procedures, and bureaucratic constituencies. CFIUS has operated since 1975 and has grown through six expansions without contraction. Institutions that persist across administrations of both parties and across leadership changes in Beijing are the load-bearing signal. The noise is the quarterly deal count, which swings on macro variables that have nothing to do with the security regime.

The indicator set for the next twelve months: first, CFIUS notice volumes and the share involving Chinese acquirers, published in the committee's annual report; second, Treasury outbound notification filings and any enforcement action under the 2025 rule; third, Chinese holdings of US Treasuries as a portfolio-tier stress gauge; fourth, US venture syndication with Chinese limited partners in AI and semiconductor funds; fifth, exit-tax and repatriation disputes affecting US funds divesting from China; sixth, any proposal — in either capital — to extend screening into the portfolio tier. Movement in the first two tells you the perimeter is holding; movement in the last tells you the scenario is shifting from persistence toward escalation.

08 Bottom Line: What We Know, and What We Cannot Yet

What we know: both governments operate standing investment-screening regimes covering both directions of capital flow; the US regime expanded to outbound coverage of semiconductors, quantum, and AI effective January 2025; two-way direct investment has collapsed relative to its mid-2010s peak while portfolio investment and trade persist at scale; specific transactions have been blocked or unwound on named, dated orders; AP reported in September 2026 that investment negotiations remain constrained by national-security concerns on both sides.

What we think we know: the screening stack slows rather than stops technology diffusion, raises the transaction cost of strategic-sector investment enough to deter unattempted deals, and is best understood as the institutional fence around an industrial-policy competition rather than as the competition itself. Chinese chip progress despite controls suggests delay, not denial — but the counterfactual pace is unobservable, so this remains inference, not measurement.

What we do not know: the elasticity of technology transfer to governance rights — that is, how much know-how actually flows through the specific channels (board seats, due diligence, management roles) that the rules target; whether Chinese portfolio capital would behave differently under stress in ways that matter; and whether either government has internally quantified the efficiency cost of the regime. No published study isolates the causal contribution of screening from the broader decoupling trend, and one should treat any confident claim to that effect with suspicion.

Watch next: the first enforcement action under the outbound rule, the 2026 CFIUS annual report's Chinese-notice share, any move to extend screening to portfolio holdings, and whether the bilateral working groups graduate from managing the screening regime toward negotiating its perimeter. The architecture is now old enough to have a track record — and the track record so far is that it reroutes capital with precision while slowing technology only at the margin, and that both governments find that trade worth making.