China would face a “hellscape” in a war over Taiwan New tech may give America an edge Sep 15th 2026|TAIPEI AND HONOLULU|7 min read   IN AN OFFICE block on the outskirts of Taipei, Chan Hsiang-chih opens a workshop door and offers a glimpse at the future of warfare in the Taiwan Strait. Inside sit two torpedo-shaped undersea drones, prototypes for Taiwan’s navy. “That’s where the bomb goes,” he says, nodding to the rear of one. It can identify targets and drop its payload or launch a kamikaze strike, alone or in a swarm. Dr Chan started his business, AwareOcean, in 2015 to supply offshore windfarms. Now he builds the kind of low-cost autonomous weapons that Taiwan and America need for their new plan to thwart a Chinese invasion.   Dubbed “Hellscape”, the plan draws on lessons from Ukraine and Iran in using large numbers of low-tech drones, combined with artificial intelligence, to frustrate a far more powerful adversary. It envisions flooding the Taiwan Strait with relatively cheap aerial, sea-surface and underwater drones just as Chinese forces launch an attack. Assuming that China jams communication links with these weapons, they will strike Chinese ships using inertial guidance, image recognition or on-board AI that prioritises targets, evades counter-measures and co-ordinates attacks with other drones. That will not stop an invasion. But it would raise the costs and buy time and space for American warships and planes to reach Taiwan.   Whether America would actually intervene is an open question, especially since Donald Trump’s return to the White House. Visiting Beijing in May, he questioned the logic of fighting so far from America and said he would use arms sales to Taiwan as a bargaining-chip with Mr Xi. Sure enough, in the run-up to Mr Xi’s visit to Washington on September 24th, Mr Trump has held off approving a $14bn arms package for Taiwan that was cleared by Congress in January. Taiwanese officials worry that more concessions to China may follow to secure deals on trade and technology. And they fear that deliveries of earlier American arms packages could be delayed by shortages due to the war against Iran.   Yet American and Taiwanese military commanders are quietly forging ahead with Hellscape, a plan originally conceived when Joe Biden was president. They think China is unlikely to attack before the end of the decade, when America will have a new leader. And they believe their odds of success have improved in the light of the wars in Ukraine and the Middle East, even accounting for China’s growing capabilities. American strategists have long taught that an attacking force needs three times the firepower of a defender to secure victory. They now think the attacker needs an advantage of more like 20 or even 30 to one.   “Uncrewed systems are to today’s conflicts what the machine gun was in World War I and the tank in World War II: a game-changing technology that redefines the battlefield,” Raymond Greene, America’s de facto ambassador in Taipei, told a conference there on September 9th. Taiwan could easily exploit the lessons from Ukraine and Iran because of its world-leading semiconductor industry, strong manufacturing base and booming economy. It also has access to battle-tested American aerial drones (of which Taiwan bought 2,000 in August). And it is collaborating with America on sea drones and AI-enabled swarming technology, he said. The problem is not the lack of a good plan. It’s that to make it work, America needs the political will and all the drone and missile capacity it can get. Proponents of the plan are looking beyond Mr Trump, banking on bipartisan congressional support for Taiwan (which remains strong) and the island’s importance to America’s AI industry. But some fear the influence within the Republican Party of a “restrainer” camp that favours abandoning Taiwan. Others worry that Mr Trump has deepened public opposition to overseas wars and done lasting damage to relations with Asian allies, whose support they would hope to have to make Hellscape work.   America’s efforts to build more low-cost drones have borne some fruit. It reverse-engineered an Iranian aerial-strike drone in 18 months. And the Pentagon plans to buy 200,000 cheap UAVs by 2027. But progress has been slowed by testing backlogs and shortages of non-Chinese components. Hellscape needs far fewer sea drones: a single successful strike can sink or cripple a warship, as Ukraine showed in the Black Sea. Some American models designed for a war over Taiwan have been used against Iran. And American troops tested a Ukrainian kamikaze boat in the Philippines in April. Still, these are still mostly experimental systems, produced in negligible numbers, says Professor Michael Horowitz of the University of Pennsylvania, who led the Pentagon’s drone drive under the Biden administration.   He and others also warn that such drones cannot replace bigger, older and more expensive weapons. The plan is for drones to exhaust China’s missile interceptors and damage its invasion flotilla sufficiently to make it easier for American aircraft-carriers, submarines and other platforms to hit their targets. The worry is that the Iran war has diverted America’s forces from the Pacific and depleted its weapons stockpiles, especially of precision missiles and missile interceptors. Restocking America’s magazines will take years. Some worry that Mr Xi might try his luck before then, possibly as soon as 2028, when Taiwan holds its next presidential election—about 10 months before Mr Trump leaves office.   “The other thing to keep in mind”, says one former Pentagon official, “is that it would not be consistent with American policy…to put large quantities of American weapons in Taiwan that we would then come in and use.” Changing that policy would invite an unholy row with China. That is why, says the former official, “it is all the more important to sufficiently train and equip Taiwanese forces.”   Taiwan presents other challenges. Chief is its capacity to produce enough drones. Soon after Russia’s full-scale invasion of Ukraine, Taiwan launched a programme to develop a drone industry free of Chinese components. It aims to buy 208,000 locally produced aerial-attack drones and 1,320 suicide boats by 2032. Output and exports have grown in the past two years. But the plan has been delayed by squabbling over funding, which was only approved in August. The main obstacle was the Kuomintang, Taiwan’s biggest opposition party, which favours closer ties with China. Should it return to power in 2028, there could be further delays. “If a real crisis hits, the timeline would probably be compressed significantly, just like in Ukraine,” says Chen Ping-hei, a professor at Taiwan National University who works on the drone programme. But as of now “the manpower and funding is clearly insufficient.” He thinks Taiwan needs 4m UAVs and 10,000 marine drones.   There are also sceptics within Taiwan’s armed forces. They worry that investment in drones could come at the expense of bigger-ticket items, such as Patriot missiles for a new air-defence system. Some also want fighters and warships to help counter Chinese actions that fall short of an invasion, such as a blockade. (Chinese customs vessels recently began regular patrols east of Taiwan.) Others place more emphasis on striking the Chinese mainland early and preserving Taiwanese forces for a counter-attack.   China’s response raises troubling questions too. It initially mocked Hellscape, with one official military outlet branding it a “psychotic plan” and calling America a paper tiger unable to produce the necessary drone swarms. The People’s Liberation Army (PLA) still enjoys an overwhelming advantage in drone numbers and production capacity. It has been testing its own unmanned systems in drills around Taiwan, while studying developments in Ukraine and the Middle East. It has also sent officers to Russia to train with Russian counterparts in drone and counter-drone operations and to gain access to relevant data from the front lines. Recently, however, there have been signs of concern within the PLA. Mr Xi’s purges have made it harder to meet his order for it to be ready to execute an invasion, if instructed, by 2027. Some Chinese military experts now suggest that Hellscape requires the PLA to update its invasion plans, and to invest more in shipborne weapons that could identify and destroy drones.   Hellscapers concede that the obstacles are daunting. But to deter conflict, its proponents say, Taiwan and America do not need to demonstrate that they could defeat China outright. They just need to convince Mr Xi that he would struggle to achieve the quick and relatively cost-free victory that he hopes for. ■   在台北郊區的一棟辦公大樓裡，詹翔智打開工作坊的大門，讓人窺見台灣海峽未來戰場的一角。房間裡擺著兩艘雪茄狀的水下無人機原型，正準備交付給台灣海峽的海軍。「炸彈就是裝在這裡，」他點頭指向其中一艘無人機的後方說道。這款無人機可以自動辨識目標、投放彈藥，或者以單機、群飛的方式發動自殺式襲擊。詹博士在 2015 年創立了「智海科技」（AwareOcean），原本是為了服務離岸風電產業；如今，他轉而製造這類低成本的自律型武器，而這正是台灣與美國為了挫敗中國入侵計畫，所急需的關鍵配備。 這項名為「地獄景觀」（Hellscape）的戰略計畫，汲取了烏克蘭與伊朗的實戰經驗，主張利用海量低成本無人機結合人工智慧，來反制實力懸殊的強大對手。它的核心構想是：一旦中國發動進攻，美台將立即在台灣海峽鋪天蓋地投放大量相對便宜的空中、水面及水下無人機。考量到中國屆時必然會進行電磁干擾、切斷訊號，這些無人機將依靠慣性導航、影像辨識，或是機載 AI 來自動排定目標優先順序、避開反制手段，並與其他無人機協同攻擊。雖然這無法完全阻止入侵，但能大幅提高敵方的犯台代價，並為美軍艦艇與戰機爭取趕赴台灣的時間與空間。 然而，美國屆時是否真的會出兵干預，目前依然是個未知數，特別是在川普重返白宮之後。川普今年五月造訪北京時，曾公開質疑為何要在遠離美國本土的地方作戰，甚至表示他會把對台軍售當作與習近平談判的籌碼。果不其然，在習近平預計於 9 月 24 日造訪華府前夕，川普扣下了美國國會早在今年一月就批准的 140 億美元對台軍售案。台灣官員非常擔心，美方為了爭取在貿易與科技領域達成協議，未來可能會對中國做出更多讓步。同時，他們也焦慮先前訂購的美國武器，會因為中東爆發對伊朗的戰爭、導致產能吃緊而延遲交付。 儘管如此，美台兩國的軍事指揮官仍默默推進著「地獄景觀」計畫——這項戰略最早是在拜登任內孕育而生的。軍方評估，中國在 2030 年以前不太可能發動攻擊，屆時美國已經迎來新一任總統；此外，借鑑烏克蘭與中東戰場的經驗，即便中國軍力不斷擴張，美台的勝算其實已經顯著提升。過去美國戰略家習慣認為，進攻方需要擁有防守方 3 倍的火力才能獲勝；但如今他們認為，進攻方恐怕需要取得 20 倍甚至 30 倍的絕對優勢才行。 美國駐台實質大使（處長）谷立言（Raymond Greene）在 9 月 9 日於台北舉辦的一場論壇上表示：「無人駕駛系統對當今衝突的重要性，就如同第一次世界大戰的機關槍、第二次世界大戰的坦克一樣，是一項徹底重塑戰場規則的革命性科技。」他指出，台灣擁有全球領先的半導體產業、雄厚的製造業基礎以及蓬勃發展的經濟，完全有能力吸收並運用烏克蘭與伊朗的教訓。此外，台灣不僅能取得經過實戰檢驗的美國空中無人機（台灣已於今年八月採購了 2,000 架），同時也在水面無人艇與 AI 蜂群技術上與美國展開合作。 問題不在於沒有好的計畫，而是要讓這項戰略奏效，美國必須展現足夠的政治意志，並備妥龐大的無人機與飛彈產能。該計畫的支持者將目光投向川普之外，寄望於美國國會長期以來跨黨派對台灣的強烈支持，以及台灣在全球 AI 產業鏈中的關鍵地位。然而，部分人士擔心共和黨內部主張孤立主義的「克制派」（Restrainers）勢力抬頭，這派聲音傾向放棄台灣；也有人憂心，川普加深了美國民眾對海外戰爭的反感，並對亞洲盟友關係造成了長期傷害，而這些盟友的支援，正是「地獄景觀」能否成功的關鍵之一。 美國在打造低成本無人機方面的努力已經取得了一些成果。美軍在 18 個月內逆向工程複製了伊朗的攻擊無人機，五角大廈更計畫在 2027 年前採購 20 萬架平價無人機。然而，由於測試流程塞車以及非中國製零組件短缺，整體進度受到了拖延。相對來說，「地獄景觀」所需的水面無人艇數量較少，因為正如烏克蘭在黑海所展現的那樣，單次成功的突襲就足以擊沉或重創一艘大型軍艦。部分針對台海衝突設計的美軍無人艇模型，目前已在打擊伊朗的行動中投入使用；美軍甚至於今年四月在菲律賓實測了一款烏克蘭的自殺式無人艇。不過，曾於拜登政府時期主持五角大廈無人機推動計畫的賓州大學教授麥可·霍羅威茨（Michael Horowitz）指出，這些目前多半仍屬於實驗性系統，量產規模小得可憐。 霍羅威茨教授與其他專家同時警告，這類無人機無法完全取代體積更大、更傳統且昂貴的重型武器。「地獄景觀」的設定，是用無人機消耗中國的防空飛彈、重創其渡海艦隊，進而為美軍航空母艦、潛艦等大型平台創造更容易打擊目標的條件。令人擔憂的是，對伊朗的戰爭分散了美軍在太平洋的兵力部署，並嚴重消耗了美國的武器庫存，特別是精準導引飛彈與防空攔截彈。補充美國的彈藥庫需要耗費數年時間，有些人開始擔心習近平可能會在此之前鋌而走險，甚至最早可能在 2028 年台灣舉行下一屆總統大選時動手——那時距離川普卸任還有大約 10 個月。 一位前五角大廈官員提醒：「另一件必須記住的事是，按照美國現行政策，我們不可能在台灣預先囤積大量美軍武器，等衝突爆發後再派美軍過去使用。」若要改變這項政策，勢必會引發與中國的大規模外交風暴。這位前官員補充道，這正是為什麼「為台灣軍隊提供足夠的訓練與裝備，會顯得如此至關重要」。 台灣自身也面臨著其他嚴峻挑戰，其中最核心的問題就是無人機的產能不足。在俄羅斯全面入侵烏克蘭後不久，台灣便啟動了一項去中國化供應鏈的無人機產業發展計畫，目標是在 2032 年前採購 20.8 萬架國產空中攻擊無人機以及 1,320 艘自殺無人艇。儘管過去兩年產量與出口量有所成長，但該計畫卻因預算爭議而一再延宕，直到今年八月預算才獲得通過。主要的阻力來自台灣最大的反對黨國民黨，該黨主張與中國建立更緊密的關係；若國民黨在 2028 年重新執政，計畫可能會面臨更多變數。參與無人機計畫的台灣大學教授陳炳輝表示：「如果真正的危機爆發，時間表大概會像烏克蘭一樣被大幅壓縮。但就目前而言，人力與資金顯然嚴重不足。」他評估，台灣實際需要高達 400 萬架無人機與 1 萬艘水面無人艇。 台灣軍方內部同樣存在質疑的聲音。部分軍官擔心，將大量資源砸在無人機上，會削弱對大型昂貴裝備的投資，例如用於新防空系統的愛國者飛彈。也有人主張應該優先購買戰鬥機與大型艦艇，以應對中國未達全面入侵標準的「灰色地帶」行動（例如中國海警船近期開始在台灣以東海域進行常態化巡邏）。另外一些人則強調，應該具備源頭打擊中國本土的能力，並在初期保存台灣戰力以備後續反攻。 中國方面的反應，同樣拋出了一連串令人不安的懸念。中方最初對「地獄景觀」嗤之以鼻，官方軍事媒體甚至痛批這是「精神病式的狂想」，並嘲諷美國是無力生產足夠無人機蜂群的「紙老虎」。中國人民解放軍在無人機數量與製造產能上，至今仍占有壓倒性優勢。解放軍不僅在台海周邊演習中持續測試自家的無人系統，也密切研究烏克蘭與中東的最新戰況；他們甚至派遣官員前往俄羅斯，與俄軍一同進行無人機及反無人機演練，藉此取得來自前線的第一手實戰數據。 然而，解放軍內部最近也開始出現焦慮的跡象。習近平發動的大規模軍方肅清行動，讓解放軍難以確保能在 2027 年前達到「具備備戰犯台能力」的要求。部分中國軍事專家現在開始建議，為了應對「地獄景觀」，解放軍必須更新其登陸作戰計畫，並在艦艇上加裝更多能夠精準辨識並摧毀無人機的武器裝備。 「地獄景觀」的推動者坦言，前面的障礙確實非常巨大。但他們強調，為了達成威懾、阻止戰爭爆發，台灣與美國並不需要證明自己能徹底擊敗中國；他們只需要讓習近平深信，他絕不可能輕易拿到他所渴望的那場「快速且低成本的勝利」。■ --Hosting provided by SoundOn