What If the $3trn AI Investment Boom Goes Wrong?對人工智能投資熱的冷思考

Even if the technology achieves its potential， plenty of people will lose their shirts.

即使人工智能實現其潛能，還是會有很多人傾家蕩產。

It already ranks among the biggest investment booms in modern history. This year America’s large tech firms will spend nearly $400bn on the infrastructure needed to run artificial-intelligence （AI） models. OpenAI and Anthropic， the world’s leading model-makers， are raising billions every few months; their combined valuation is approaching half a trillion dollars. Analysts reckon that by the end of 2028 the sums spent worldwide on data centres will exceed $3trn.

在現代史上，這已然躋身規模最大的投資熱潮之列。今年，美國大型科技公司將在運行人工智能（AI）模型所需的基礎設施上投入近4000億美元。全球領先的AI模型開發商OpenAI和Anthropic每隔數月即融資數十億美元，兩家公司總估值接近5000億美元。分析師預測，到2028年底，全球建設數據中心的總花費將超過3萬億美元。

The scale of these bets is so vast that it is worth asking what will happen at payback time. Even if the technology succeeds， plenty of people will lose their shirts. And if it doesn’t， the economic and financial pain will be swift and severe.

這些賭博式的投資規模巨大，真該想想投資的回報如何。即便這項技術成功，仍將有許多投資者血本無歸。而倘若該技術失敗，經濟與金融將會遭受迅速且劇烈的打擊。

Investors always flock to promising technologies， but the AI rush is more extreme than many past booms. Boosters say that artificial general intelligence （AGI）—models that are better than the average human at most cognitive tasks—could be only a few years away. The first firm to achieve it could reap unimaginable returns. Investors and innovators know they may not be backing the right model. But if they spend slowly and cautiously， they may as well not bother to spend at all.

投資者總是蜂擁追逐前景廣闊的技術，但當前這場AI熱潮比以往許多熱潮都更為極端。鼓吹者宣稱，通用人工智能（AGI）——在大多數認知任務上超越普通人類水平的模型——可能僅需數年便能實現。首家獲此技術的公司可能得到難以想象的收益。投資者與創新者深知自己的押注未必正確，但若投資步伐遲緩或過于謹慎，則無異于主動放棄。

Consequently， a relentless investment race is under way， with big tech firms splurging on the computing power needed to build the biggest models. A growing cast of extra players， from property developers to electricity generators， has been drawn in. Oracle is the latest to join the party. Its value surged on September 10th after it published an ambitious forecast for its AI related cloud business， briefly turning its boss， Larry Ellison， into the world’s richest man.

因此，一場激烈的投資競賽如火如荼展開了，為構建最大的模型，大型科技公司在算力上投入重金。從房地產開發商到發電企業，越來越多的行業卷入了這場浪潮。甲骨文公司是最新一例：公司雄心勃勃地公布進軍AI云業務后，9月10日，其市值應聲暴漲，而其老板拉里·埃里森也一度成為世界首富。

No matter what happens， many investors will lose money. In the rosiest scenario AGI will arrive and usher in a new world of economic growth of perhaps 20% a year. Some shareholders would enjoy astronomical returns; many others would face big losses.

無論結局如何，很多投資者都會面臨虧損。最樂觀的情形是，AGI順利落地，從而開創年經濟增長率或達20%的新紀元。倘真如此，有些股東將獲得天文數字級的收益，而其他許多人則會蒙受巨額虧損。

More mundane scenarios should also be considered， however. The technology may evolve in ways that investors do not expect. When alternating current eventually prevailed in America in the 19th century， for instance， direct-current electricity firms were overshadowed and forced to consolidate. Today investors reckon that the probable AI winners are those that can run the biggest models. But， early adopters are turning to smaller language models， which could suggest that less computing capacity may be needed after all.

然而，我們亦應考慮那些波瀾不驚的情形。技術演進方式可能會出乎投資者的預料，比如，19世紀，交流電在美國最終大行其道時，直流電企業便黯然失色并被迫整合。如今，投資者認為，能夠運行最大模型的公司或將成為AI領域的贏家。但是，早期使用者正悄然轉向更精簡的語言模型，這可能意味著所需的算力也許根本沒那么大。

Or the road to widespread adoption could be slower and bumpier than investors expect， giving today’s AI laggards a fighting chance. Niggles in the technology， the difficulty of quickly supplying electric power， or managerial inertia could mean that take-up is more gradual than first hoped. As they revise down their expectations for AI revenues， many investors and creditors could become less willing to countenance huge investments. The flow of capital could slow; some startups， struggling under the weight of losses， could fold altogether.

或許，技術普及之路會比投資者預期的更為緩慢曲折，這將給當前AI領域的落后者奮力一搏的機會。技術層面的小故障、電力供應的遲緩或企業的管理不力，都可能使技術跟進的速度不及最初的預期。隨著很多投資者和債權人對AI收益預期的逐步下調，他們可能不再那么心甘情愿地投入巨資。資本流動可能放緩，而部分在虧損重壓下掙扎的初創企業或將徹底出局。

What would such an AI chill be like？ For a start， a lot of today’s spending could prove worthless. After its 19th-century railway mania， Britain was left with track， tunnels and bridges; much of this serves passengers today. Bits and bytes still whizz through the fibre-optic networks built in the dotcom years. The AI boom may leave a less enduring legacy. Although the shells of data centres and new power capacity could find other uses， more than half the capex splurge has been on servers and specialised chips that become obsolete in a few years.

AI寒潮何堪想象？首先，目前很多的巨額投入可能會打了水漂。19世紀的鐵路狂熱過后，英國尚遺留下鐵軌、隧道與橋梁等基礎設施，其中大部分至今仍在服務客運；互聯網熱潮時建設的光纖網絡，至今仍有數據馳騁其間。而AI熱潮留下的遺產或許遠沒有那么持久。盡管數據中心外部建筑與新擴建的電力設施或可另作他用，但超過一半的資本都投向了數年內即告淘汰的服務器和專用芯片。

The good news is that today’s financial system could probably absorb the blow. Some technological busts have been brutal; after Britain’s railway bubble burst in the 1860s banks were left with big losses， leading to a credit crunch. So far， however， much of the investment in data centres has been bankrolled from the deep well of big-tech profits.

所幸當今金融體系或許尚能承受這樣的沖擊。歷史上有過殘酷的技術泡沫破滅：19世紀60年代，英國鐵路泡沫破滅后，銀行體系承受巨額虧損，繼而引發信貸緊縮。但迄今為止，數據中心建設的大部分資金都源自科技巨頭雄厚的利潤儲備。

Although firms including Meta are turning to debt to help fund their latest investments， their lucrative businesses and robust balance-sheets mean they are well positioned to finance a technological boom. Among the keenest to provide that credit are private-market funds， which are typically funded by rich individuals and institutions rather than ordinary depositors. AI startups tend to be financed by well-capitalised venture and sovereign-wealth funds that could withstand losses.

盡管Meta等公司正通過舉債為最新投資計劃融資，但其利潤豐厚的核心業務與穩健的資產負債表意味著它們完全有能力為技術熱潮提供資金。最熱衷提供這類信貸的是私募基金，這些資金通常來源于富豪個人或機構而非普通儲戶。而AI初創企業往往由資金雄厚的風險投資機構及主權財富基金注資，這些機構有能力承受潛在損失。

However， trouble spots could still emerge. The more the investment boom spreads， the more financing structures could get riskier， and the more indebted firms could be drawn in. Power companies are desperate to increase their investments to supply AI with the electricity it needs; a heavily indebted utility could easily become overextended.

然而，風險隱患仍可能悄然出現。投資熱潮越是蔓延，融資結構就越有風險，陷入債務的企業也就越多。電力公司正孤注一擲地增加投資以滿足AI的電力需求，但對于負債累累的公用事業公司而言，極易被過度擴張拖垮。

America’s economy， too， would suffer a nasty shock. By one estimate， the AI boom has contributed 40% of its GDP growth over the past year—a staggering figure for a sector that accounts for just a few per cent of total output. If investment projects are scaled back， or ditched altogether， that will translate into economic pain as fewer data centres are built， and fewer workers are employed to build them.

美國經濟亦將遭受重挫。據估算，過去一年AI熱潮貢獻了其GDP增長的40%——對于僅占經濟總量幾個百分點的行業而言，這一數字堪稱驚人。若投資項目縮減或徹底終止，數據中心建設數量就會減少，相關就業崗位也會萎縮，這一切將直接轉化為經濟陣痛。

To make matters worse， falls in the stock market could cause asset owners to cut back on their spending. Because the valuations of AI-related companies have rocketed， portfolios today are dominated by a handful of tech firms. And households are more exposed to stocks than they were in 2000; if prices fall， their confidence and spending could take a knock. The poorest would be spared， because they tend to hold few stocks. But it is the rich who have fuelled consumption in America over the past year. Robbed of its sources of strength， the economy would weaken as tariffs and high interest rates take a toll.

更令人擔憂的是，股市下跌可能引發資產所有者緊縮開支。由于AI相關企業的估值已飆升，現今的投資組合均由少數科技巨頭所主導。而且，與2000年相比，美國家庭更易受股市影響，一旦股價下跌，民眾信心與消費能力將遭受雙重打擊。因持股較少，最貧困人群或能幸免。但是，過去一年推動美國消費的核心力量恰恰是富裕階層。若失去這關鍵動力，加之關稅與高利率持續施壓，美國經濟必將疲弱。

The bigger the boom gets， the bigger the knock-on consequences of an AI chill could be. If the technology ends up fulfilling the extravagant promises that have been made for it， a new chapter of history will open. However， the story of its frenzied pursuit will make the textbooks， too.

這場AI熱潮聲勢愈浩大，其降溫引發的連鎖沖擊便可能愈深遠。若技術最終能兌現人們對其寄予的遠大期許，歷史將開啟新篇。無論怎樣，這場狂熱追捧也都可以編入教科書，予人警示和啟迪。