going all in, because it believes it has no choice. If it succeeds, it will transform the nature of sea power. Britain’s navy has two big jobs which are essential for the security of its European allies and NATO. The first is keeping submarine-mounted nuclear missiles at sea at all times to deter attacks. The second is protecting communication lines and infrastructure in the North Atlantic and High North. These would be crucial in a war, especially with Russia, which is pouring resources into its northern fleet. Partly because of the first job’s vast cost (a quarter of Britain’s defence budget), the navy thinks it may fail in the second unless it shakes things up. The fundamental idea behind the hybrid navy is to split up the functions of the exquisite, enormously expensive warships it now relies on. These will instead be spread over many smaller, nimbler, cheaper ships performing specialised roles. The “Common Combat Vessels” that will act as command centres for these fleets should cost £200m ($260m) each, compared with about five times that for a high-end frigate and over £2.5bn for an air- defence destroyer. The new fleets promise to control the seas far more efficiently than existing ones. With more hulls they can cover a greater area than a smaller number of warships. Vessels will be much bigger and more sophisticated than the one- way attack ships used to great effect by Ukraine in the Black Sea. But they will be able to take more risk than traditional ones, being both cheaper and often uncrewed. Since the hybrid navy’s vessels should be simpler to build, more commercial shipyards will compete to do so. That will help control costs. It also carries echoes of NASA’s decision, in the 2000s, to welcome startups into its rocket programme. That policy helped spawn SpaceX, one of the world’s biggest firms, and handed America a gigantic lead in space warfare. However, getting the hybrid navy afloat will involve daunting risks. Naval drones tend to have less endurance and armament than larger, crewed warships. Also, Britain’s sea forces do not work in isolation. Securing the North Atlantic and High North will take close co-operation with America,

France and Norway. These allies will need to plug seamlessly into the hybrid navy’s new operating system. Mistakes are inevitable: no country has tried this before. Some may arise from the new doctrinal questions hybrid fleets pose. If Russia attacked a crewed warship, Britain would see it as an act of war. What about an uncrewed one? The navy must learn from the mistakes it will make, and fast. Moreover, without enough money the project will sink. Britain’s sailors have clocked that a radical new idea is a good way to grab cash from a limited pot. Launching the hybrid navy will still take perhaps another £32bn by 2035—about half of this year’s defence budget. The government has yet to say how it will fund such investments. Even so, huzzah for the Royal Navy. Too often, Britain balks at risk and so commits itself to stagnation. Its navy is doing the opposite. ■ This article was downloaded by zlibrary from https://www.economist.com/leaders/2026/11/19/can-britannia-rule-the-waves-again

Leaders · Leaders | Tinker, Tailor, Soldier, AI

America must get serious about Chinese chip spying The Jensen Huang affair exposes an AI underworld Nov 19th 2026 Somehow, try to move on from Jensen Huang. The leather-jacketed ex-boss of the chip designer Nvidia, who was turfed out after two (alleged) affairs with subordinates and one (admitted) dalliance with an (alleged) Chinese agent, is eyeing a comeback. Yet, captivating as Mr Huang’s manoeuvres are, the fracas at Nvidia has underscored a deeper, and more consequential tale: China’s hell-for-leather scramble to gain access to the technology behind American artificial intelligence, by whatever means necessary. Smuggling chips into China is only the start. Probably more important has been remote access to data centres, mainly in South-East Asia. It offered a sometimes-legal route for Chinese developers to train their models with top- end chips—until November 16th, when America’s Bureau of Industry and

Security, the branch of the Commerce Department that runs chip controls, set out a rule restricting the practice. Remote access and smuggling together roughly doubled the compute, or processing power, available to Chinese AI companies from domestic production and legal imports alone, estimates the Institute for Progress, a think-tank. But China wants more than just to get the best chips. It also wishes to build them itself. Behind the rapid growth of China’s domestic semiconductor industry have been campaigns of talent-poaching and industrial espionage, especially focused on researchers in Taiwan and South Korea. CXMT, a Chinese memory giant, was jump-started by former Samsung staff who, South Korean authorities found, left in 2016 with hand-copied notebooks full of production techniques. South Korea jailed a senior defector in April. Top of China’s shopping list and still, tantalisingly, out of reach is extreme- ultraviolet lithography (EUV) technology, which is needed to produce today’s whizziest chips at scale. Only one company in the world, ASML, can produce an EUV machine. The Dutch government, under American pressure, has stopped it from ever exporting one to China. The staggering breadth of China’s campaign means America must respond forcefully. Not with a crude protectionist purpose, or to hamper Chinese industry broadly, but to ensure an authoritarian regime does not take the lead on AI, let alone win the race to superintelligence. The technology has serious military applications, as the battlefields of Ukraine show. And the closer to the frontier China gets, the harder it will be for America to slow down AI research, making it more likely that the technology brings about disasters. So far, America’s efforts have been chaotic and confused. That is in part because the Trump administration is divided. The Commerce Department, led by Howard Lutnick, who is cosy with Nvidia, has been astonishingly lax. Rightly, the Huang affair has imperilled Mr Lutnick’s job. Scott Bessent, the treasury secretary, used his first television interview after the midterm elections to blame Mr Lutnick for slow-walking restrictions. The instability of policy has frustrated chip companies, which operate in a capital-intensive industry where products and facilities take years to develop.

What now? America is at last cracking down on Chinese firms’ access to overseas data centres, although allowing some access through trusted providers with a kill-switch would offer more leverage than a simple shutdown. More could be done, such as requiring makers of cutting-edge chips to embed location trackers. Broadly, a reasonable standard would be only to allow chips at least two generations old into China (an “n-2” rule), targeting the tech most useful for training frontier AI. America must also work better with allies to tackle Chinese industrial espionage. Hearteningly, Congress is pushing to legislate on these issues. America cannot halt China’s chipmaking progress. DeepSeek, a Chinese AI champion, says its upcoming model will be pre-trained on domestically produced chips. But merely slowing Chinese AI down is worthwhile and achievable. The industry reckons that China is still ten-plus years away from a commercially viable EUV machine, for instance. Those years may matter enormously. Recently the mood in Silicon Valley has darkened. A growing share of AI researchers see months, years at most, until AI reliably surpasses humans—including at AI research itself. That might open the door to “recursive self-improvement”, where each model autonomously builds its own successor. If they are right, the next few years will be tumultuous. Governments in America and elsewhere might have to scramble to contain runaway AI systems while the technology disrupts the economy and society. The task would be daunting. With China threatening to overtake America, it might be impossible. ■ This article was downloaded by zlibrary from https://www.economist.com/leaders/2026/11/18/america-must-get-serious-about- chinese-chip-spying

Leaders · Leaders | Co-operate to accumulate

America’s neglect of global science is backfiring The Trump administration’s reforms have overcorrected for past mistakes Nov 19th 2026 AMERICAN SCIENCE leads the world because of both its home-grown strength and its openness to collaborating with foreigners. The Trump administration prefers to rely only on the first half of the formula. As we exclusively report, the number of grants made by the National Institutes of Health (NIH) involving collaborators abroad has been 38% lower on average in 2025 and 2026 than in 2022-24. At the same time the administration is making it far harder for foreign scientists to study and work in America. The funding of overseas research needed more oversight. Yet in its approach to foreign know-how and foreign brains, the administration has overcorrected. The results will be bad for humanity, because no other country can advance the frontiers of science as much as an outward-looking America can. They will also be bad for America itself.