Invest AI

投資者對人工智慧 (AI) 的熱情推動美國股市在 2024 年初創下多個歷史新高。人工智慧相關股票,尤其是大型股的強勁表現,讓一些市場觀察人士質疑,這是否類似於2000年破裂的互聯網泡沫。

我們仍然將人工智慧視為近代史上最具革命性的技術進步之一,這是一個多年的機遇。它才剛剛開始出現在公司的底線中。

自去年以來發生了很多事情,因此我們正在總結我們目前對人工智慧股票的看法,以及該技術可能採取的下一步行動和全球人工智慧的前景。

我們強調,人工智慧必須被視為一種全球現象,我們不建議只投資美國公司。我們還認為,至關重要的是,想要購買人工智慧的投資者必須確保投資組合能夠接觸到從基礎設施到軟體和人工智慧應用的整個人工智慧價值鏈中的公司。我們看到了該技術的“推動者”和“採用者”的潛在機會。

又大又閃亮?是的。泡沫?不。

人工智慧股票正在上漲,因為投資者一直在對技術需求處於長期增長期開始的跡象做出反應。自 2023 年初以來,人工智慧相關股票的回報率比美國和全球指數高出 30%。

一些投資者將這些巨大的波動與1990年代後期的互聯網泡沫相提並論。在此期間,隨著投資者開始認識到互聯網的潛力,科技股的表現大幅跑贏大盤。2000年3月泡沫破滅,納斯達克綜合指數在兩年內下跌了近80%,抹去了泡沫時代的收益。直到 2015 年才完全恢復。

這兩個時代有多相似?讓我們考慮一下數據。

股票市場泡沫的特徵是股票在投機和投資者過度熱情的驅使下變得不合理地昂貴。投資者遲早會意識到,公司將無法實現投資者的增長預期,泡沫破滅,導致價格迅速下跌,因為投資者意識到資產被高估了。

評估股價的一種常見方法是將股價與公司的預期利潤進行比較,以每股收益表示。每股收益只是公司的利潤除以市場上的股票數量。該指標是股價的主要燃料。每股市盈率稱為遠期市盈率或市盈率。高遠期市盈率可能表明對未來盈利增長的樂觀和信心,但也可能表明過度熱情。

2000年1月,五大科技公司(Microsoft思科、英特爾、朗訊和IBM)的平均遠期市盈率為59倍,根據其相對規模進行調整。當今五大科技股(Microsoft、英偉達、亞馬遜、Meta 和 Alphabet)的遠期市盈率為 34 倍,僅為一半。

數據顯示,在2000年,分析師預計當時的科技領導者每股收益將增長30%,而今天的分析師預計增長42%。這是股價更堅實的基礎。

在此基礎上,華爾街認為,即使以市盈率衡量,人工智慧股票的交易價格要低得多,但今天的人工智慧領導者將實現比互聯網領導者預期的更好的盈利增長。

人工智慧股價遠未達到科技泡沫高點

將納斯達克 100 指數和 AI Leaders 與 Dot.com 的今天價格表現進行比較

該圖覆蓋了納斯達克 100 指數從 1995 年到 2002 年的價格表現,以及納斯達克 100 指數和當今人工智慧領導者自 2019 年以來的價格表現。在此基礎上,我們沒有看到今天的人工智慧領導者已經達到了不合理的價格水準。

資料來源:彭博財經、摩根大通私人銀行。數據截至 2024 年 3 月。

查看資訊

雖然 2000 年代和當今領導者之間的估值差異並不能保證人工智慧主題股票將繼續跑贏市場其他股票,但在我們看來,很明顯 2024 年不是 2000 年。

與2000年科技公司相比,當今人工智慧領導者的估值較低,增長預期更高

比較 2000 年和 2024 年領先科技股的盈利增長和市盈率

該表比較了 2000 年和 2024 年領先科技股的盈利增長和市盈率。

資料來源:FactSet。數據截至 2024 年 3 月。EPS:每股收益。PE:市盈率。

查看資訊

向 AI 2.0 過渡

到目前為止,投資人工智慧主題可能感覺很簡單,因為最大的科技股已經取得了非常強勁的回報。我們認為,從這裡開始,更廣泛、更平衡的東西會更好。

我們在兩個領域看到了潛在的機會,我們稱之為AI 1.0和AI 2.0。

AI 1.0 是支撐 AI 的基礎設施。隨著對複雜 AI 功能的需求不斷增長,對可擴展且強大的基礎設施的需求也在增長。該技術建立在數據中心的基礎上,由於大多數 AI 工作負載都位於雲上,因此 AI 正在推動雲的進一步增長。

包括亞馬遜、Microsoft、Alphabet 和 Meta 在內的領先雲計算公司都迅速推出了多年投資計劃,以支援他們在人工智慧時代所需的更大雲容量。

人工智慧的需求正在迅速消耗現有的數據中心容量,促使公司建造新的設施。這也帶來了潛在的投資機會。公用事業公司可能不得不增加燃煤或燃氣發電,不斷增長的需求無疑將激發基礎設施投資和努力開發更節能的網路、更好的冷卻系統以及整合可再生能源的新解決方案。

人工智慧基礎設施和大型語言模型 (LLM) (如 ChatGPT)的進步的另一個關鍵層是快速處理大量數據所需的計算能力。LLM 是計算機程式,它使用在大型數據池上訓練的架構來學習和生成類似人類的語言。這些計算由稱為圖形處理單元 (GPU) 的半導體處理。十年來的重大 GPU 進步意味著當今更快、更高效的性能。GPU 的進步使 2020 年之前的大多數計算電子設備都過時了。

行業領先的 GPU 製造商 Nvidia 最近估計,對 GPU 的總需求可能達到 2 萬億美元 2。這包括來自數據中心的 1 萬億美元,以及與人工智慧相關的工作的 1 萬億美元,例如訓練新的 LLM、機器學習和科學類比。

英偉達數據中心季度收入(數十億美元)

此圖表顯示了 2020 年第一季度至 2024 年第四季度 Nvidia 數據中心部門的季度收入數據。

資料來源:彭博財經、摩根大通私人銀行。數據截至 2024 年 3 月。

查看資訊

AI 1.0 環境中有許多受益者,隨著對新計算基礎設施和人工智慧服務的需求增長、成本下降以及消費者更多地使用該技術,這一主題可能會繼續發揮作用。

人工智慧之旅還處於早期階段,過去的表現並不能保證未來的回報,但我們認為,隨著市場的發展,當今領先的人工智慧基礎設施公司,如數據中心和雲供應商以及半導體製造商,應該會繼續增長。

然而,人工智慧中大多數未被承認的價值都在軟體和應用程序等領域。我們稱之為 AI 2.0 主題,它專注於「採用者」。客戶服務、醫療保健、金融和物流等行業已準備好通過人工智慧進行重大轉型。

例如,先買後付公司 Klarna 最近開始使用由 OpenAI 提供支援的 AI 助手。在第一個月,它進行了 230 萬次對話,占 Klarna 客戶服務聊天次數的三分之二,相當於 700 名全職座席3。

我們認為,在AI 1.0和AI 2.0之間保持平衡的投資組合敞口可能是潛在利用這一領域前景的有效方式。

走向全球

美國科技巨頭在人工智慧時代受到了很多關注,可以追溯到 2022 年 11 月發布 ChatGPT,這引起了人們對人工智慧領域的新關注,並推動了該技術的採用。投資者可能錯過了其他地方人工智慧領導者的潛力。

中國在人工智慧領導權競賽中與美國勢均力敵。去年 12 月,中國科技公司百度表示,其生成式 AI 聊天機器人用戶已超過 1 億,可與 ChatGPT4 的 1.8 億使用者相媲美。

在印度,人工智慧的採用也有相當大的潛力,印度是一個數據豐富的國家,行動裝置的使用範圍很廣。假以時日,這可能會轉化為重大的投資機會。韓國、日本和新加坡仍然是創新中心。

歐洲各地的領先公司也可能成為人工智慧的主要受益者,因為該技術可以幫助他們變得更有效率和更有利可圖。其中包括德國強大的工業部門,法國的航空航天、汽車和化工公司,以及荷蘭的高端科技製造公司。後者也是全球物流的主要樞紐5。

政府開始瞭解訪問和控制其數據對國家安全的影響6。他們還在戰略上定位自己,以利用人工智慧的潛力。競爭可能會加劇,類似於美國限制向中國出售某些先進人工智慧晶元的政策的新規則可能會變得更加普遍7。

由於所有這些原因,未能從全球視角看待人工智慧的投資者可能會錯過真正的突破性創新。

人工智慧的發展才剛剛開始,我們預計未來幾年整個生態系統的機會將繼續以國家、公司和初創公司的形式出現。未來幾年,我們相信人工智慧將迅速改變我們的思維、工作和解決問題的方式,為突破性的創新和變革開闢道路。

1Indxx 人工智慧和大數據指數持有85隻製造人工智慧所需硬體或在其服務中使用人工智慧的股票,其表現比標準普爾500指數和MSCI世界指數高出30%以上。過去的表現並不能保證未來的結果。不能直接投資指數。

2NVIDIA.com。摩根士丹利科技、媒體和電信會議(連結)。3月4, 2024.

3Klarna AI 助手在第一個月(2024 年 2 月 27 日)處理了三分之二的客戶服務聊天(連結)。

4Evelyn Cheng,“百度表示其 ChatGPT 競爭對手 Ernie 機器人現在擁有超過 1 億使用者。 CNBC.com,2023 年 12 月 28 日(連結)。

5“繪製人工智慧的新興地理圖”,《哈佛商業評論》,2023 年 12 月 12 日(連結)。

6Keith Strier,“什麼是主權人工智慧?”,NVIDIA.com,2024 年 2 月 28 日(連結)。

7工業和安全域 2022 年 10 月 7 日(連結)。

Investors’ enthusiasm for artificial intelligence (AI) has propelled U.S. equity markets to multiple all-time highs in early 2024. The strong outperformance of AI-related stocks, especially mega-caps, has some market-watchers asking if this could be similar to the dot-com bubble that burst in 2000.

We still see AI, one of the most revolutionary technological advancements in recent history, as a multi-year opportunity. It is only just starting to show up in corporate bottom lines.

A lot has happened since last year, so we’re rounding up our current thinking on AI stocks, as well as the likely next steps for the technology and the outlook for AI worldwide.

We emphasize that AI must be viewed as a global phenomenon, and we don’t advise investing in U.S. companies exclusively. We also think it’s paramount that investors who want to buy into AI make sure portfolios have exposure to companies across the AI value chain ranging from infrastructure to software and AI applications. We see potential opportunities in “enablers” and “adopters” of the technology.

Big and shiny? Yes. Bubble? No.

AI stocks are on a roll as investors have been reacting to signs that demand for the technology is at the start of a long period of growth. Since the beginning of 2023, AI-connected stocks have delivered 30% better returns than both U.S. and global indexes.1

Some investors have compared these outsized moves to the dot-com bubble of the late 1990s. In that period, tech stocks outperformed dramatically as investors began to recognize the potential of the internet. The bubble burst in March 2000, and the Nasdaq Composite declined almost 80% over two years, wiping out the gains of the bubble era. It didn’t make a full recovery until 2015.

How similar are the two eras? Let’s consider the data.

An equity market bubble is characterized by stocks that become unreasonably expensive driven by speculation and excess investor enthusiasm. Sooner or later, investors come to the realization that companies will be unable to deliver on investor growth expectations and the bubble bursts, leading to a rapid decline in prices as investors realize the asset is overvalued.

One common way to evaluate stock prices involves comparing the share price to the company’s expected profits as expressed in earnings-per-share. Earnings-per-share is simply a company’s profit divided by the number of shares on the market. That metric is the primary fuel for stock prices. The ratio of price to earnings-per-share is known as a forward price-to-earnings, or P/E, ratio. A high forward P/E can indicate optimism and confidence in future earnings growth, but it can also signal excessive enthusiasm.

In January 2000, the five largest tech companies (Microsoft, Cisco, Intel, Lucent and IBM) traded at an average forward P/E ratio of 59x, adjusted by their relative sizes. The five biggest tech stocks today (Microsoft, Nvidia, Amazon, Meta and Alphabet) have a forward P/E ratio of 34x—barely half as much.

Data shows that in 2000, analysts expected 30% earnings-per-share growth from the tech leaders of the day, while today’s analysts expect 42% growth. That’s a more solid foundation for stock prices.

On this basis, Wall Street thinks today’s AI leaders will deliver better earnings growth than it expected from dot-com leaders even as the AI stocks trade at much lower prices as measured by the P/E ratio.

AI stock prices are far from tech bubble highs

Comparing today’s price performance for the Nasdaq 100 and the AI Leaders to the Dot.com

This graph overlays the price performance of the Nasdaq 100 Index from 1995 to 2002 and the price performance of the Nasdaq 100 and today’s AI leaders since 2019. On this basis, we don’t see today’s AI leaders having reached unreasonable price levels.

Source: Bloomberg Finance L.P, J.P. Morgan Private Bank. Data as of March 2024.

View info

While the valuation difference between the 2000s and today’s leaders is no guarantee that AI themed stocks will continue to outperform the rest of the market, in our view, it’s clear that 2024 is not 2000.

Lower valuation and higher growth expectations for today’s AI Leaders vs. 2000 Tech

Comparing earnings growth and P/E ratios for leading tech stocks in 2000 and 2024

This table compares earnings growth and P/E ratios for leading tech stocks in 2000 and 2024.

Source: FactSet. Data as of March 2024. EPS: earnings-per-share. PE: price-to-earnings.

View info

Transitioning to AI 2.0

Investing in the AI theme may have felt simple so far, as the biggest tech stocks have delivered very strong returns. We think something broader and more balanced will work better from here.

We see potential opportunities in two areas that we refer to as AI 1.0 and AI 2.0.

AI 1.0 is the infrastructure that underpins AI. As the demand for sophisticated AI capabilities grows, so does demand for a scalable and powerful infrastructure. The technology is built on the foundation of data centers, and because most AI workloads live on the cloud, AI is fuelling further cloud growth.

Leading cloud computing companies, including Amazon, Microsoft, Alphabet and Meta, have all rapidly rolled out multi-year investment plans to support the greater cloud capacity they will need in the AI era.

Demand from AI is rapidly consuming existing data center capacity, pushing companies to build new facilities. That, too, presents potential investment opportunities. Utilities may have to add coal- or gas-fired power, and rising demand will undoubtedly spark infrastructure investments and efforts to develop a more energy efficient network, better cooling systems, and new solutions to integrate renewable energy.

Another crucial layer in the AI infrastructure and the advancement of large language models (LLMs) such as ChatGPT is the computational power required to rapidly process the vast amounts of data. LLMs are computer programs that learn and generate human-like language using an architecture trained on large pools of data. These computations are processed by semiconductors known as graphics processing units (GPUs). A decade of significant GPU advancements means much quicker and more efficient performance today. The GPU advancement has rendered obsolete most computational electronics from before 2020 obsolete.

Nvidia, maker of industry-leading GPUs, recently estimated that total demand for GPUs might reach $2 trillion2. This includes $1 trillion from data centers, and $1 trillion from work connected to AI such as training new LLMs, machine learning, and scientific simulations.

Nvidia quarterly revenue from data centers (billions)

This chart shows the quarterly revenue numbers for Nvidia Data Center division from Q1 2020 to Q4 2024.

Source: Bloomberg Finance L.P, J.P. Morgan Private Bank. Data as of March 2024.

View info

There are many beneficiaries in the AI 1.0 environment, a theme that could continue to work as demand for new computing infrastructure and AI-enabled services grows, costs drop, and consumers make greater use of the technology.

The AI journey is in its early stages, and past performance is no guarantee of future returns, but we think today’s leading AI infrastructure companies, such as the data center and cloud providers and the semiconductor manufacturers, should continue to grow as the market develops.

However, most of the unrecognized value in AI is in areas such as software and applications. We call this the AI 2.0 theme, which focuses on “adopters”. Industries such as customer service, healthcare, finance, and logistics are poised for significant transformation through AI.

For example, the buy now pay later company Klarna recently began using an AI assistant powered by OpenAI. In the first month, it had 2.3 million conversations, or two-thirds of Klarna’s customer service chats, doing the equivalent work of 700 full-time agents3.

We believe maintaining a balanced portfolio exposure between AI 1.0 and AI 2.0 could be an effective way to potentially capitalize on the promise of this space.

Go global

U.S. tech giants have received much of the attention in the AI era dating back to the release of ChatGPT in November 2022, which drew new attention to the AI field and supercharged adoption of the technology. Investors may be missing the potential of AI leaders elsewhere.

China closely rivals the United States in the AI leadership race. In December, Chinese tech company Baidu said its generative AI chatbot had surpassed 100 million users, rivalling the 180 million users of ChatGPT4.

There is also considerable potential for AI adoption in India, a data-rich nation with widespread mobile device use. In time, this could translate into significant investment opportunities. South Korea, Japan and Singapore remain innovative hubs.

There are leading companies across Europe that could become major beneficiaries of AI as well, as the technology may help them become more efficient and profitable. These include Germany’s robust industrial sectors, France’s aerospace, automotive and chemicals firms, and the Netherlands’ high-end tech manufacturing firms. The latter is also a primary hub in global logistics.5

Governments are beginning to understand the national security implications that surround access and control of their data6. They are also strategically positioning themselves to harness the potential of AI. Competition is likely to intensify, and new rules similar to the U.S. policy limiting the sale of some advanced AI chips to China may become more common.7

For all these reasons, investors who fail to approach AI through a global lens could miss out on true ground-breaking innovation.

The evolution of AI is only just beginning, and we expect that opportunities throughout the ecosystem will continue to emerge over the next few years in the form of countries, companies, and startups. In the coming years, we believe AI will rapidly change the way we think, work, and solve problems, opening the path to ground-breaking innovation and change.

1The Indxx Artificial Intelligence and Big Data Index, which holds 85 stocks that make hardware needed for AI or that use AI in their services, outperformed both the S&P 500 and the MSCI World by more than 30%. Past performance is no guarantee of future results. It is not possible to invest directly in an index.

2NVIDIA.com. Morgan Stanley Technology, Media & Telecom Conference (link). March 4, 2024.

3Klarna AI assistant handles two-thirds of customer service chats in its first month, February 27, 2024 (link).

4Evelyn Cheng, “Baidu says its ChatGPT rival Ernie bot now has more than 100 million users.” CNBC.com, December 28, 2023 (link).

5“Charting the Emerging Geography of AI,” Harvard Business Review, December 12, 2023 (link).

6Keith Strier, “What is Sovereign AI?” NVIDIA.com, February 28, 2024 (link).

7Bureau of Industry and Security October 7, 2022 (link).

The above is a reproduction of JP Morgan Private bank analysis. Full access here.

Leave a Reply

Your email address will not be published. Required fields are marked *