Version 2

US semiconductor export controls have entered an executive-legislative standoff. The Trump administration softened Biden-era restrictions through 2025 and 2026, culminating in approved H200 chip sales to ten Chinese firms [3], while Nvidia CEO Jensen Huang publicly declared the controls 'already largely backfired' and left Nvidia with zero percent China market share [8]. Against this, the Senate unanimously passed the Stop Stealing Our Chips Act on May 21, 2026, establishing whistleblower bounties for export control evasion [13]. Trump's post-Beijing signals — warning Taiwan against formal independence and reportedly making no commitment to defend it [20][21] — have added an explicitly transactional dimension to the Taiwan strand of the rivalry.

The US government is now arguing with itself over chip controls: Congress is legislating for tighter enforcement while the White House relaxes restrictions, and Nvidia's zero-China-market-share outcome illustrates what happens when controls are designed imperfectly — revenue loss for American firms without meaningful denial of compute to China. Whether the executive-legislative contradiction can be resolved, and whether the whistleblower enforcement mechanism can actually close gray-market leakage, will determine if US chip controls remain a credible geopolitical tool or become primarily a domestic commercial liability.

The US chip export control regime, built incrementally since 2022, has undergone a significant reversal under the Trump administration. By early 2026, East Asia Forum characterized the controls as having 'cooled down' [1], and Reuters reported Trump actively reining in technology curbs on China [2]. The most concrete expression of the reversal came in May 2026, when the US government cleared H200 chip sales to ten Chinese companies [3][4], giving Nvidia federal approval to export its most capable China-eligible chip to specific Chinese buyers. Earlier reporting described the structure as a cap of approximately 75,000 units with a 25% tariff [5]. The BISI think tank characterized the overall shift as Trump reversing US AI chip export policy toward China [6], while the Council on Foreign Relations assessed the new policy framework as 'strategically incoherent and unenforceable' [7].

The commercial context for the reversal is stark. Nvidia CEO Jensen Huang stated publicly that the company now holds 'zero percent' market share in China and that US export policy 'has already largely backfired' [8]. He characterized prior restrictions as measures that 'effectively closed' China to Nvidia [9]. Nvidia had previously attempted to navigate the restrictions by modifying the H20 chip specifically to comply with export rules [10], but those modifications proved insufficient when the H20 was subsequently banned. Nvidia is now reportedly designing a new lower-cost chip purpose-built for the Chinese market [11]. AMD CEO Lisa Su offered a different posture, saying her company is maintaining 'close cooperation with China ecosystem' even as AI chip sales broadly stalled [12] — suggesting the two major US GPU makers are handling the regulatory environment in distinctly different ways.

While the executive branch loosened controls, Congress moved in the opposite direction. The Senate unanimously passed the Stop Stealing Our Chips Act (S.1473) on approximately May 21, 2026 [13][14][15], with Senators Warner and Rounds as principal sponsors [16]. The House version (H.R.6322) runs in a parallel legislative track [17][18]. The bill creates a whistleblower bounty structure paying 10–30% of violators' fines from an Export Compliance Accountability Fund [19], directly targeting gray-market transshipment through third countries — a reference to the documented pattern of US chips reaching Chinese buyers via Southeast Asian intermediaries. The unanimous Senate vote, rare for any legislation in the current political environment, signals genuine bipartisan commitment to enforcement even as the White House relaxes substantive restrictions — a contradiction that will require resolution before the bill can be signed and implemented coherently.

The Taiwan dimension of the chip story has grown substantially more fraught. Following Trump's Beijing summit with Xi, he reportedly warned Taiwan directly against formal independence [20] and made 'no commitment' to defend Taiwan [21]. Trump advisers are said to fear China may move on Taiwan within five years [22]. Trump has also pressed semiconductor companies to relocate manufacturing to the US, citing Taiwan geopolitical risk [23]. Taiwan's President Lai pushed back publicly, asserting that 'the future will NOT be decided by external forces' [24]. Structurally, Western fab capacity is accumulating: TSMC is accelerating its Arizona production timeline [25], and Intel's domestic Fab 52 is reportedly larger and better-equipped than TSMC's Arizona facilities [26]. These facts provide some structural grounding for Chamath Palihapitiya's earlier forecast that Taiwan's strategic centrality would erode within 18 months as Western fab capacity scales [27] — but the transition remains years away, and in the interim Trump's transactional stance toward Taiwan introduces new uncertainty into the security architecture underpinning global semiconductor supply chains.