Investors should embrace artificial intelligence (AI) as a long‑term investment theme, even if some popular AI stocks may appear a little “frothy” in the short term, says Rahul Ghosh, portfolio specialist, global equity at T Rowe Price.
He adds that the firm’s many global technology stocks have seen accelerating organic earnings growth and operating margin expansions over the past 12 months. These trends are expected to continue through the remainder of the year
Through industry conversations, contacts and events, the analyst observes that use cases for AI and generative AI are expanding across sectors such as pharmaceuticals, financial services, media, advertising and information technology.
“I think that is giving us the conviction that this is something that has legs, driven by a combination of supply and demand changes, unlocking technology that can be used and adopted with revenue and monetisation opportunities,” he says. Therefore, investing in AI will be a theme for the future.
Currently, the world is in the infrastructure build-out phase of AI. Companies are purchasing servers, graphics processing units (GPUs) and central processing units (CPUs) to develop their large language models (LLMs) and compete in the market.
To provide solutions for everyday individuals, companies such as Meta and Google will keep investing in hardware to stay the most competitive, but there is a physical limit to how much companies can produce, grow and sell.
See also: Thai exchange to use AI to improve oversight of listed firms
Typically, volume cycles will have to ease as the rest of the market catches up; that is how cycles evolve, Ghosh explains. Only once the hardware that has been produced has absorbed all its capacity can the innovations start taking off again.
But while AI is and may continue to be cyclical, the team at T Rowe Price believe that the market has likely underestimated its S-curve trajectory. For one, the demand for AI has been underestimated as Moore’s Law continues to play out. As the speed and ability of semiconductors doubles while power consumption lowers, the opportunity for GPUs to overtake CPUs in data centres to give companies more service space to run analytics and AI capabilities has grown.
Manufacturers of GPUs and CPUs found themselves upgrading their estimate of the total AI addressable market — from US$50 billion ($67.5 billion) to US$150 billion, to US$50 billion to US$400 billion from 3Q2023 to 4Q2023 last year, Ghosh says.
“The other thing I’d say is that we have not seen a lot of companies, outside traditional technology companies, talk about the benefits of AI and what they can do yet,” he adds. The analyst highlights industries that have yet to fully utilise AI tools to their benefit, such as medicine, renewable energy production and cybersecurity.
“The history of innovation tells us that innovation is almost always underestimated in the long term, and when you think about the ability of AI and generative AI to affect almost all parts of the economy, you will see that that’s where the growth will come from,” says Ghosh.
The analyst also cautions investors to be careful, as valuations of AI-related stocks have surged. He believes the industry might be in a hype cycle, though not quite in an “AI bubble” where expanded valuations lack earnings growth.
Ghosh: I’m not saying [AI stocks] are cheap per se, but I always encourage any potential investor to think about what the earning returns have done around the stock price. Photo: T Rowe Price
Are AI stocks too expensive?
Should investors continue investing in this tier of AI-related stocks? To Ghosh, most companies are not trading on ridiculously expensive multiples. In terms of their enterprise value-to-sales (EV/Sales) multiple, the financial valuation metric that compares a company’s enterprise value to its annual sales, these companies are trading four to seven times.
To stay ahead of Singapore and the region’s corporate and economic trends, click here for Latest Section
Likewise, their EV/Ebit levels are trading at a range of low double-digit ebit to two to three times multiples, and their multiples reflect that their returns are significantly improving, says Ghosh.
Using Nvidia as an example, Ghosh says that the company’s 2023 consensus estimate was a return on equity (ROE) of 25% to 30%, which will see a more than double to a high 80% by 2024 or 2025, a trend that will be observed in other peers such as Advanced Micro Devices (AMD) as well.
Other broader technology and semiconductor companies are trading in the high 20s to low 30s range, which, as Ghosh recalls, is “nothing compared” to the dot-com bubble boom when stocks traded at a P/E of 200 or more than six times today’s levels.
“Again, I’m not saying they’re cheap per se, but I always encourage any potential investor to always think about what the earning returns did around the stock price,” Ghosh adds.
Perhaps most important to Ghosh is that these companies are profitable — Nvidia has a high 50% to 60% ebit margin and semiconductor companies have 28% to 35% margins. These are significantly profitable and very free cash flow generative, and it is something that people should not forget about, says the analyst.
Due to supply shortages, T Rowe Price predicts substantial growth in the chip ecosystem in the next six to 18 months. Ghosh adds that demand for semiconductor packaging equipment will rise as new AI applications require advanced assembly methods.
“Some of those companies only started seeing their orders grow y-o-y for the first time one or two quarters ago, so even though the AI hype has been there, they haven’t seen that growth in numbers for a long time,” he adds.
Pyramid of AI ecosystem companies, according to T Rowe Price. Photo: T Rowe Price
The chip ecosystem offers significant growth opportunities in the development phase, considered the “base layer” of the pyramid. This is where companies including Nvidia, ASML, AMD, Lam Research and TSMC are positioned. Investing up the pyramid becomes slightly trickier for investors, as a significant portion of the development is private, Ghosh explains. “So if you think about it, even OpenAI was not an investment available for the average investor.”
Investors who may choose to have a super-focused strategy in investing in a segment of chips or applications may find their portfolios could be up 200% to 300% “if everything went right”, but if things went south, they could incur a dramatic loss, adds Ghosh. He cautions that such volatility and risk are things investors should try to insulate their portfolio from as much as possible.
Beyond Nvidia
Ghosh and his team encourage investors to diversify their portfolios across various hardware, software, infrastructure players and enablers opportunities. This could include investing in the chip ecosystem and large technology companies like Meta and Google, which are developing their LLMs. Additionally, investors can explore different angles, such as data centres, GPUs, CPUs and consumer electronics.
He advises investors to look beyond a single geographical area. While the leading players in most AI sub-sectors are typically concentrated in the US due to its historically large market, Ghosh suggests exploring other markets with significant experience in different AI sectors, such as semiconductors.
“But there are very good semiconductor players in other areas of the world,” says Ghosh. “We have in our fund, we own some European names. You know, there are very interesting ideas in Japan and other areas in countries that have a history of technology and semiconductors.”
Investors should also consider the broader segments around AI, which will be crucial for its future growth. One key area is the significant power required for AI operations. For example, Bloomberg recently reported that new data centres planned for Silicon Valley could add 3.5 gigawatts of electricity demand, exceeding the output of three nuclear power plants.
Ghosh suggests that companies providing renewable power solutions for data centres, essential for AI growth, are part of the broader AI investment chain. This allows investors to diversify beyond a strategy entirely dependent on Nvidia, Amazon or Microsoft.
Yet, he says that investors need to ask themselves who they think will capture the lion’s share of the value created in this AI boom. “And if that’s companies in the US, then there’s no shame in being overweight there. If that’s companies outside, there’s no shame in being overweight there.”
