•
J.P.Morgan：2025年全球經濟（聚焦5個行業）
未來資本聚焦的五大方向：
1. 人工智慧：繁榮還是泡沫？
2. 醫療行業的顛覆性變革
3. 自動化與機器人技術
4. 建構能源基礎設施
5. 重新定義國防安全
AI: Boom or bust?
Yes, U.S. big tech companies have opened the spigot for AI spending, but these are still early days. We think capital investment in AI could take off in the coming years, driven by rapid improvements in AI models and corporate adoption. Consider: AI could potentially impact all services activity in the economy.
Why are we so optimistic? First, because AI models are improving at a rapid rate. In 2021, large language models (LLMs, a type of AI) could answer less than 10% of competition-level math questions accurately. That share increased to 90% in 2024. The models are also becoming less expensive: The price per token for both OpenAI’s higher-performing GPT-4o mini model and Anthropic’s Claude 3.5 Haiku model are 90%–98% less expensive than their predecessors.
Second, overall corporate capital investment has been relatively muted, running at a 2.5% annual pace. By contrast, at the end of the dot com boom at the turn of the millennium, corporate capex was running at a 10% annual pace (on a five-year rolling basis). In other words, there is plenty of room for corporations across sectors to increase their AI spending as the use cases become more apparent—and persuasive.
Third, we see the potential for AI to “turn labor into software,” as Sequoia Capital has put it. As models improve their ability to reason instead of merely generating pre-trained responses, they will help create opportunities to disrupt the services sector. AI lawyers, AI software engineers and—dare we say it?—AI investment strategists could become commonplace. Public markets and their private partners have established a strong presence in the digital infrastructure of AI. The industrial and utilities companies that provide the physical components and energy needed for AI will also likely continue to benefit. Finally, the companies that most effectively improve their cost structures and increase productivity by incorporating AI tools into their workstreams should outperform.
In private markets, pure-play AI valuations have inflated, but we still see investment opportunities in the startups that can automate tasks and provide cost savings to businesses. Value may also be created in the potential applications that can help harness the technology for consumers. Over 20 major application layer companies (e.g., Salesforce, Meta, Uber) were founded during the cloud and mobile transitions. AI could create a similar ecosystem.
Adoption of AI could falter, of course. Regulation could stifle innovation. Energy sourcing could prove onerous. And the models could run out of the data they use to train. But our analysis—looking past the hype and drawing on the lessons of history—tells us that AI offers significant investment opportunity. We see the potential for a clear bull case for the global economy and equity markets next year and beyond.
一、人工智慧：繁榮還是泡沫？
美國的大型科技公司已經打開了AI投資的大門，但目前仍是早期階段。我們相信，隨著AI模型的快速進步和企業逐漸接受，未來幾年AI資本投資將迎來爆髮式增長。試想一下，AI可能會深刻影響經濟中的每一個服務領域。
為什麼我們對AI如此樂觀？
1、AI模型的性能在不斷突破。還記得2021年時，大型語言模型（LLMs）能精準回答的競賽級數學題不到10%，但到了2024年，這一比例已經飆升到90%。更重要的是，使用這些模型的成本也在迅速下降。以OpenAI的GPT-4o Mini和Anthropic的Claude 3.5 Haiku為例，每個Token的價格比上一代產品便宜了90%-98%。
其次，企業的整體資本支出目前增長緩慢，年均增長僅為2.5%。相比之下，在網際網路泡沫高峰期，企業資本支出年增長率曾高達10%。這意味著，隨著AI的實際應用越來越清晰，各行業的企業在AI領域還有很大的投資潛力。
2、AI可能帶來一次“勞動變革”。紅杉資本一份調查顯示：AI有潛力將“勞動變成軟體”。隨著AI模型從簡單的預訓練響應轉向更強的推理能力，服務行業將迎來新的顛覆機會。未來，AI律師、AI軟體工程師，甚至AI投資策略師都有可能成為我們日常工作的一部分。
AI的發展不僅依賴數字基礎設施，那些為AI提供物理裝置和能源的工業公司和公用事業公司同樣將受益。而那些能夠利用AI工具最佳化成本結構、提升生產效率的企業，將更具競爭優勢。
在私人市場中，雖然純AI業務的估值已經偏高，但那些能夠實現任務自動化、幫助企業節省成本的初創公司仍然值得關注。同時AI的潛在應用也蘊藏機會，比如為消費者開發實際用途的技術。在過去，雲端運算和移動網際網路的轉型催生了像Salesforce、Meta、Uber這樣的巨頭。我們相信，AI將創造出類似的生態系統。
當然AI的普及也面臨一些風險，比如監管可能限制創新、能源問題可能變成負擔，甚至AI模型可能耗盡可用的訓練資料。但從歷史經驗和實際分析來看，AI帶來的投資機會是巨大的。我們相信，未來幾年，AI不僅能推動全球經濟增長，還將為股市注入強勁動力。
Healthcare Disruption
AI may quickly impact the healthcare sector. The chart helps to understand why. The industries that could be most impacted by AI have a high share of labor costs in jobs that could be helped or displaced by LLMs. For example, the healthcare technology industry has labor costs equivalent to around 35% of its sales.
At the same time, economists estimate that 65% of the tasks performed by healthcare technology employees are exposed to AI disruption. For example, companies such as Veeva Systems, Teladoc and GoodRx that provide software solutions across the healthcare value chain could both reduce their labor costs and increase their revenues by incorporating AI into their businesses.
In the pharmaceutical and biotech sectors, AI could also potentially improve the quality and quantity of drugs that progress from early-stage trials to market. Right now, only 7% of new drugs make it to market. Just a 5% increase in that success rate could mean 60 new drugs and USD 70 billion in incremental revenue over a 10-year period. In life science services, companies could use AI to design drug trials more optimally: from compound identification to participant selection.
We also believe GLP-1 drugs will continue to drive revenue growth (glucagon-like peptide drugs control blood sugar and suppress appetite). According to our estimates, we think the total addressable market could grow from 16 million people in the United States and the European Union in 2027 to over 40 million people. Beyond GLP-1s, we are focused on identifying companies in the healthcare sector that can use AI to modernize their business models to drive earnings growth. The clearest examples right now are in robotic surgery and imaging technologies used for diagnostics.
二、醫療行業的顛覆性變革
人工智慧（AI）正快速改變醫療行業，這背後的邏輯其實很簡單：一些勞動成本佔比較高的崗位，很多任務都可以被AI最佳化甚至取代。拿醫療技術領域來說，勞動成本佔銷售額的35%左右，而專家預測，這些崗位中有65%的工作任務都可能受到AI的顛覆性影響。
比如像Veeva Systems、Teladoc和GoodRx這樣的公司，它們為醫療行業提供軟體解決方案。如果引入AI，不僅能降低人工成本，還能提升收入。還有製藥和生物技術領域，AI的加入或許能大幅提高藥物研發的效率和成功率。目前，新藥從早期試驗到上市的成功率只有7%。如果這一比例能提升5%，未來10年可能新增60種藥物，帶來高達700億美元的額外收入。此外，生命科學服務領域也可以通過AI最佳化藥物試驗流程，從化合物篩選到試驗參與者，選擇都更科學、更高效。
我們還特別看好GLP-1類藥物（胰高血糖素樣肽藥物），它們不僅能控制血糖，還能抑制食慾。據我們估算，到2027年，這類藥物的潛在市場規模將在美國和歐盟從1600萬人擴大到超過4000萬人。
除了GLP-1藥物，那些能利用AI升級商業模式的醫療公司同樣值得關注。目前最典型的例子包括機器人手術和用於診斷的成像技術，這些方向正在快速驅動行業的盈利增長。
Automation & Robotics
Capital investment in AI will also fast-track the adoption of automation and robotics, impacting industrial and consumer sectors. U.S. industrial companies are set to allocate 25%–30% of their capital spending to automation over the next five years, up from 15% to 20% over the last five years.
This theme isn’t new. Single-purpose robots have existed for over half a century, and robotics returns for investors have been modest so far. But we think momentum is building for broader applications. As access to training data increases and the cost of hardware declines, general-purpose robots may be closer to achieving the ability to reason. Globally, companies have invested over USD 4 billion in funding more than 20 “humanoid” robots.
Eventually, robots (humanoid and otherwise) may become a part of our daily lives. Waymo is already providing more than 100,000 autonomous taxi rides per week.
Healthcare and defense are two other areas where robotics will become more prevalent. In October, Intuitive Surgical delivered 110 of its semi-autonomous, AI-enabled Da Vinci 5 robotic surgery systems, trouncing the 70 placements from the previous quarter. Earlier in the year, the U.S. Air Force announced a contract award with Anduril and General Atomics to develop Autonomous Collaborative Combat Aircraft. The Department of Defense expects to spend nearly USD 3 billion per year on the program by 2029.
We see investment opportunities in both public and private markets, in the semiconductor companies that provide the computing power, the software companies that harness that computing power, the industrial companies that capitalize on higher efficiency, and the consumer companies that can deliver a game-changing product.
AUTONOMOUS AIRCRAFT SPENDING IS SET TO SURGE
AUTOMATION COULD ACCOUNT FOR 25% OF INDUSTRIAL CAPEX OVER THE NEXT 5 YEARS
三、自動化與機器人技術的崛起
人工智慧（AI）的資本投入正在加速自動化和機器人技術的普及，影響到工業和消費領域。預計未來五年，美國工業企業將把25%-30%的資本支出用於自動化，相比過去五年的15%-20%有明顯提升。
其實，這個主題並不新鮮。單一用途的機器人已經存在了半個多世紀，但過去的投資回報並不算突出。不過，隨著訓練資料的獲取越來越便捷，硬體成本逐步下降，機器人應用正向更廣泛的領域邁進。特別是通用型機器人，距離實現“推理”能力似乎越來越近。全球範圍內，企業已投入超過40億美元，用於支援20多個“類人型”機器人項目。
未來無論是類人型機器人還是其他形式的機器人，都可能逐漸融入我們的日常生活。比如Waymo目前每周已經提供超過10萬次自動駕駛計程車服務。
在醫療和國防領域，機器人技術的應用也在快速增加。今年10月，Intuitive Surgical公司交付了110套配備人工智慧的Da Vinci 5機器人手術系統，這一數字遠超上一季度的70套。而在今年早些時候，美國空軍還與Anduril和通用原子公司簽署了一項開發自主協作戰鬥機（Autonomous Collaborative Combat Aircraft）的合同。美國國防部預計，到2029年，該項目的年度預算將接近30億美元。
我們認為，投資機會不僅僅存在於機器人本身，還包括其相關的上下游產業。比如提供計算能力的半導體公司、開發軟體以充分利用計算能力的科技企業、通過自動化提升效率的工業企業，以及能將機器人技術轉化為消費級顛覆性產品的公司，都是值得關注的領域。
Building Power Infrastructure
We think capital investment into the power sector is about to ignite for three key reasons: the reindustrialization of U.S. manufacturing, increased use of electrification in clean energy solutions, and surging demand from data centers. Overall, we expect power demand growth in the United States to increase by 5x to 7x over the next 3–5 years.
Data center growth is a global phenomenon. The number of U.S. data centers, accounting for 40% of the global market, is growing ~25% per year. In Q1 2024, the European, Latin American, and Asia-Pacific data center markets grew inventory by 20%, 15%, and 22% year-over-year, respectively. Increased data center power requires more water for cooling and chip fabrication, often in water-stressed areas. Global data centers are expected to grow their water usage by 6% annually. Large semiconductor fabrication facilities use the same amount of water as 300,000 households. We see opportunities for water infrastructure and efficiency solutions in parallel with growing power usage.
More power will likely come from nuclear energy. We note the equity market’s validation of Constellation Energy’s and Microsoft’s agreement to restart the Three Mile Island nuclear power plant to supply energy to the tech giant’s data centers. This should spur further reinvestment in nuclear energy. Indeed, the surge in the Nuclear Renaissance Index (+75% year to date) is based on market speculation that small modular reactors will be successfully deployed in the next few years. While renewable energy sources will continue to grow (the International Energy Agency believes that for every $1 invested in fossil fuels, $2 are invested in clean energy), latency, transmission, and storage costs mean that natural gas will remain a critical energy source.
Investors looking to capitalize on the growing demand for power can focus on broad infrastructure funds, power generation, and utility companies.
THE MARKET HAS VALIDATED NUCLEAR EXPANSION
DATA CENTERS COULD IGNITE A SURGE IN POWER DEMAND
ARMED CONFLICTS ARE AT AN 80-YEAR HIGH
四、建構電力基礎設施
我們認為，電力領域的資本投資即將迎來快速增長，背後有三個關鍵原因：
1、美國製造業的再工業化
2、清潔能源解決方案中電氣化的廣泛應用
3、資料中心對電力需求的激增。
總體來看，預計未來3-5年，美國的電力需求增長可能達到5倍至7倍。
資料中心的增長是全球現象。目前，美國的資料中心數量約佔全球市場的40%，每年以約25%的速度增長。2024年第一季度，歐洲、拉丁美洲和亞太地區的資料中心庫存同比分別增長了20%、15%和22%。資料中心對電力的高需求還伴隨著對水資源的更大需求，主要用於冷卻和晶片製造，尤其是在水資源緊張的地區。預計全球資料中心的用水量每年將增長6%。一座大型半導體製造廠的用水量相當於30萬個家庭的用水需求。這種趨勢為水利基礎設施和水資源效率解決方案帶來了投資機會。
核能可能成為提供更多電力的重要來源。市場已經認可了Constellation Energy與微軟的合作協議，他們計畫重啟三里島核電站，為微軟的資料中心供電。這一合作有望進一步推動核能的再投資。事實上，“核能復興指數”今年已上漲75%，這主要源於市場對小型模組化反應堆（SMR）在未來幾年內成功部署的樂觀預期。
雖然可再生能源將繼續快速發展（國際能源署資料顯示，每1美元的化石燃料投資就有2美元投入清潔能源），但由於延遲性、傳輸及儲存成本問題，天然氣仍將是重要的能源來源。
對於希望抓住電力需求增長機遇的投資者，可以關注廣泛的基礎設施基金、電力生產公司以及公用事業企業。
Redefining Security
As governments reassess their national security, they will likely deliver higher levels of capital investment. Security covers not just traditional military defense, but cybersecurity, supply of critical natural resources, energy production, transportation, and infrastructure. We think markets do not yet fully appreciate the investment prospects that this secular shift will create.
In the United States, the government seems likely to continue incentivizing domestic production of critical supplies. Shares in a North Carolina–based chipmaker surged 40% on the news that it secured USD 750 million in CHIPs Act funding to build two new semiconductor plants in the United States. Further, we highlight a U.S. government goal to diversify its reliance on concentrated aerospace and defense specialists (which today account for 90% of the U.S. weapons production budget), and expand to established commercial companies and startups with expertise in areas such as AI, machine learning, and 5G technology. The defense market could potentially generate revenues and market share for companies outside the traditional defense space. The Department of Defense budget has halved as a share of GDP from the height of the Cold War. Of the European Union members of NATO, 16 of the 23 are currently on track to surpass the targeted 2% of GDP threshold in 2024. Ten years ago, the average member was only at 1.2% of GDP. Global military spending seems to have room to grow, especially in areas such as network-enabled weapons, which could be human-machine partnerships or fully autonomous.
Europe’s security concerns reflect its reliance on external sources for critical goods and commodities. The European Union imports over 90% of digital products and services, depends on Asia for 75%–90% of wafer fabrication capacity, and relies on China for up to 70% of key raw materials such as nickel, copper, and cobalt. Meanwhile, the BRICS+ economies control 5x the natural gas reserves of the G7, have 3x the active-duty military personnel, and double the oil reserves and uranium production. Last year, Russia boosted its defense budget by 25% to hit a new record.
In all, we believe global spending on security will be comparable to the annual investment in cloud computing and e-commerce during the 2010s. Building out semiconductors, infrastructure, and reliable, affordable power are not only pivotal levers for national security, but also for global economic competition.
We are looking for opportunities in the industrial, utilities, materials, and energy sectors. All but the industrials sector trade at a discount to the broad market. Investors don’t seem to be giving these companies much credit for future earnings growth, which we believe will be nearly double that of the market over the next few years.
In private markets, investors can find interesting prospects in smaller companies focused on innovation in technology-enabled defense systems and cybersecurity.
五、重新定義國防安全
各國政府正在重新審視國家安全策略，未來可能會投入更多資金，不僅包括傳統的軍事防禦，還涉及網路安全、關鍵資源供應、能源生產、交通運輸以及基礎設施建設。而這背後隱藏的投資機會，市場還沒有完全發掘出來。
在美國，政府似乎傾向於通過政策激勵推動關鍵物資的本土化生產。比如，一家北卡羅來納州的晶片製造商最近因為獲得《晶片法案》7.5億美元的資金支援，用於建設兩家新的半導體工廠，其股價直接飆升了40%。同時，美國政府也在努力減少對少數傳統防務巨頭的依賴（目前這些公司佔國防預算的90%），轉而支援那些在人工智慧、機器學習和5G等技術領域具有專長的商業公司和初創企業。這意味著，國防領域的市場份額和收入將不再侷限於傳統防務企業。
歐洲的安全問題則更加複雜。歐盟嚴重依賴外部供應：超過90%的數字產品和服務依靠進口，晶圓生產能力的75%-90%來自亞洲，關鍵原材料如鎳、銅和鈷有多達70%需要從中國進口。同時，金磚國家（BRICS+）在天然氣儲量上的控制力是七國集團（G7）的5倍，現役軍人數量是G7的3倍，石油儲量和鈾產量更是G7的2倍。這種資源依賴使歐洲面臨巨大的戰略風險。
儘管如此，全球軍事支出仍有很大的增長空間，尤其是在高科技武器領域。比如人機協作武器系統和完全自主的武器系統正迅速發展。僅去年，俄羅斯就將國防預算提高了25%，創下歷史新高。
投資的切入點在那裡？
能源、工業、公用事業和材料行業正成為投資焦點。除了工業類股外，其他行業的估值都低於市場平均水平，我們認為這些行業未來幾年的盈利增長可能會達到市場整體增速的兩倍。尤其是在能源和基礎設施建設領域，半導體生產能力提升、可靠能源供應的佈局都至關重要。
在私募市場，小型創新企業尤其值得關注。那些專注於技術賦能的防禦系統和網路安全解決方案的公司，未來可能有不小的增長潛力。
重新定義安全不只是國家戰略的轉變，更是未來幾年裡不可忽視的投資方向。如果能抓住這波機遇，可能會帶來長期穩定的回報。 (行業調研報告)