Chip controls add risk

The United States is trying to sell more artificial intelligence chips abroad at the same time it is making the permission process slower. Inside the Commerce Department, the Bureau of Industry and Security is now taking months to review some licenses after nearly 20% turnover among rulemaking and licensing staff cut processed licenses by about 25%, according to Bloomberg’s reporting published April 10, 2026. (bloomberg.com) That office is the traffic light for sensitive chip exports. The Bureau of Industry and Security says companies must use its Simplified Network Application Process Redesign system to apply for licenses and its System for Tracking Export License system to monitor approvals under the Export Administration Regulations. (bis.gov) The slowdown matters because Washington changed the rules in January. On January 13, 2026, the Bureau of Industry and Security said it would review exports of Nvidia H200 chips, Advanced Micro Devices MI325X chips, and similar products to China case by case if buyers met new security conditions. (bis.gov) That made the licensing desk more important, not less. A system built for controlling a narrow stream of shipments now has to judge deals involving Nvidia processors for China and the Middle East while also helping run tariff investigations on steel, autos, and other sectors. (bloomberg.com) At the same moment, Congress is pushing in the opposite direction on chipmaking tools. House bill H.R. 8170, introduced on April 2, 2026 by Representative Michael Baumgartner, would restrict exports of certain semiconductor manufacturing equipment and components, and it was referred to the House Foreign Affairs Committee the same day. (congress.gov) The target is deep-ultraviolet lithography equipment, which is the machine that projects circuit patterns onto silicon the way a tiny, brutally precise stencil projects ink onto paper. Asia Times reported on April 10 that bipartisan lawmakers want to block Chinese access to deep-ultraviolet immersion tools that Chinese chipmakers have still been able to buy. (asiatimes.com) Those machines sit one step earlier in the supply chain than Nvidia’s finished chips. If a country cannot easily buy the processors and also cannot easily buy the tools needed to make more advanced processors at home, the whole compute chain starts to look less like a market and more like a customs checkpoint. (congress.gov) This is not a brand-new fight. A Congressional Research Service report published on September 19, 2025 says the United States has been tightening semiconductor export controls on China since 2018 to slow China’s access to advanced chips, related computing, and artificial intelligence applications while trying to preserve United States leadership. (congress.gov) What changed in 2026 is that uncertainty is now coming from both policy direction and administrative capacity. Bloomberg reported that top officials tightened control over individual licenses, seasoned staff left, and attention shifted after late February toward the war in Iran, delaying other technology-export priorities. (bloomberg.com) When hardware access gets harder to predict, software starts doing more of the work. A cloud company that can squeeze more output from the same Nvidia cluster through better scheduling, tighter model optimization, and closer integration between chips, networking, and software can keep selling compute even when the next shipment sits in a licensing queue. (rand.org, bloomberg.com) That is why export controls are no longer just a story about who gets a box of chips. In April 2026, they are also a story about who has the cleaner software stack, the faster approval path, and the fewer points in the supply chain where one government office can make a multibillion-dollar delay. (bloomberg.com, congress.gov)