Find value in the AI wars by purchasing the ammo-makers

Big Tech is vying for control of artificial intelligence

It's difficult to predict the winner. Regardless of the result, astute investors will support businesses that will make money.

According to some of Silicon Valley's leading CEOs, artificial intelligence (AI) represents the greatest opportunity in their company's history. It's also their biggest threat, but few will admit that. Portfolio manager Alec Cutler of Orbis Investment says, "For those big players, this is an existential fight for survival." "They're going to share a lunch. Like everyone else, the large tech firms at the center of the AI boom are facing significant business disruptions from this technology. If AI can begin to reliably and consistently respond to people's queries, Google Search may become obsolete. Due to its apparent inability to integrate AI into its products, Apple's standing as the leading consumer goods company in the world is in jeopardy. After its cloud division grew more slowly than Google's or Microsoft's, Amazon's stock fell precipitously. The Magnificent Seven are fighting a costly war among themselves to make sure they emerge from AI as one of the major winners rather than a loser, according to Rupert Thompson, chief economist at IBOSS Asset Management.

The fact that the biggest names in AI trade on enormous earnings multiples is an unavoidable fact, regardless of whether this is a winner-take-all situation or something more complex. Of course, many investors will be discouraged. "I'm more convinced than ever that there is as much risk as reward in the US's burning sun of AI," Cutler says. The upside is extremely limited, the downside is significantly less so, and the costs are enormous. However, he believes that any of their clients who win (or survive) the AI war will benefit the semiconductor companies, ammunition suppliers, and related suppliers.

Even though Nvidia recently became the first £4 trillion company in history, Cutler does not include it in that category. Both AMD and Broadcom have the potential to challenge its successful monopoly on AI chips, and all three chip companies are highly valued. Nvidia encourages its own customers to create alternatives the more it attempts to control prices. However, the same tiny and comparatively underappreciated group of companies are making all of these businesses possible.

Central to it all is the AI chipmaker.

Taiwan Semiconductor Manufacturing Company, or TSMC (NYSE: TSM), is at the top of this list. Despite being "chipmakers," Nvidia, Broadcom, and AMD do not produce their own chips. Although a third party handles the manufacturing, they still design them. Because TSMC is the world's leading manufacturer of the most advanced semiconductors, it is that third party in practically all cases involving AI. For decades, TSMC has established and maintained its near-monopoly. Due to the high barriers to entry and the incredibly complex and capital-intensive nature of the technology involved, TSMC has already defeated the majority of its rivals and enjoys an almost impenetrable moat in the market for high-end chips.

The latest chips from Nvidia, Broadcom, AMD, and Qualcomm are all produced by TSMC. In addition, it produces chips for Microsoft, Google, and Amazon, the "hyperscalers" engaged in the AI war, and it has been producing chips for Apple for more than ten years. Even if Nvidia's market share were to be taken by Broadcom, AMD, or its own clients, TSMC would still be involved. TSMC doesn't care if Nvidia's market capitalization declines, according to Cutler.

With a £4 trillion market valuation, Nvidia is the most valuable company in the world, and TSMC just joined the £1 trillion club. This makes it difficult to argue that it is an underappreciated gem and places it squarely in mega-cap territory. In contrast to Nvidia, which trades at over 40 times forward earnings, TSMC is trading at less than 25 times. TSMC appears to be significantly undervalued when you take into account the size of its moat and the fact that all of Nvidia's customers, regardless of its rivals, are actively attempting to join its scheme.

Because of geopolitical risk, most commentators attribute this. For TSMC, a decision by China to annex Taiwan could be disastrous. However, Nvidia and the rest of the US tech sector that depends on its chips would also face difficulties, as Cutler notes. He claims that "because Taiwan is in the name," the market undervalues TSMC. However, TSMC and Nvidia have equal leverage in Taiwan. He believes that TSMC is "the only company out there that deserves to sell on no-competition margins."

This value chain is data-centered.

TSMC is only the beginning. It has its own suppliers, and experts with extensive knowledge are essential to its dominance in the semiconductor foundry industry. Lithography, which prints patterns onto silicon wafers that are used to make computer chips, is a sophisticated technology. Although it still holds 90 percent of the market, the Dutch company ASML (Nasdaq: ASML) was once the most notable name here. However, in mid-July, it issued a warning that it could not guarantee growth in the upcoming year. The company's outlook is being negatively impacted by worries about export controls and tariffs. Additionally, historically, it has had a near-monopoly on supply to TSMC; however, in April, TSMC disclosed that its most recent manufacturing procedures do not depend on ASML's extreme ultraviolet lithography (EUVL) instrumentation. American depository receipts for ASML have decreased by 8% over the past month.

But because this is such a specialized market, only two significant rivals can match ASML's lithography capabilities: Nikon (Tokyo: 7731) and Canon (Tokyo: 7751). Up until the 2000s, both of these controlled the lithography market, but ASML surpassed them in the competition to produce the extremely expensive machinery needed by cutting-edge semiconductor manufacturing. Now that they appear to be catching up, Canon opened its first chipmaking facility in 21 years on July 30. Nikon also began taking orders for its DSP-100 Digital Lithography System, which is used for advanced packaging and panel-level semiconductor manufacturing, last month. Nikon is below 20, Canon is just over 11, and ASML is trading at about 25 times projected earnings.

Franklin Small-Mid Cap Growth Fund portfolio manager John Scandalios also selects KLA (Nasdaq: KLAC). Although KLA, which trades at a forward price-earnings ratio of 26, isn't exactly cheap, it does control over half of the market for metrology equipment and over 50% of the market for wafer inspection, two essential steps in the semiconductor manufacturing process. "It's crucial for many of these companies that the average person or investor doesn't talk much about KLA, and may not even know anything about wafer inspection and process control," says Scandalios.

Profitability.

Even in the present day, when large tech companies appear to dominate the global stock market, the semiconductor industry is extremely cyclical in and of itself. Makoto Tsuchiya, an economist at Oxford Economics, claims that "it follows its own cycle, often called the silicon cycle." It's really erratic. However, the infrastructure sectorwhich is more conventionally defensiveoffers investors access to the market. AI requires a lot of data centers, which use a lot of energy and equipment.

Through its investments in digital and energy infrastructure, Pantheon Infrastructure (LSE: PINT) is one investment trust that provides exposure to the infrastructure firms driving the AI boom. It separates digital infrastructure into data centers, fiber, and towers. Its investments in data centers have specifically targeted the AI opportunity. Among them are Vantage Data Centers and CyrusOne. Richard Sem, a partner at Pantheon Infrastructure, claims that "it's effectively providing bricks and mortar" for the infrastructure required as AI infrastructure builds.

Large amounts of energy are used by data centers: according to Goldman Sachs, the global data center power consumption is approximately 55GW, and it is expected to rise by 50% by 2027 and 166% by 2030. "This is really new, especially in the US, where we haven't seen rising power demand in decades," says Janice Ince of Pantheons. Specialist companies like Constellation Energy are largely responsible for the hype surrounding the AI energy play, but this invariably distorts valuations; Constellation is currently trading at about 35 times projected earnings. Its nuclear power plants take years to build, and they account for the majority of its capacity. The AI war's "Hyperscalers" require power now.

Ince states that "right now, it's time to power and not cost of power" that counts. "A decade or more is the time to go nuclear. Thus, even though Amazon has contracted for nuclear facilities, they won't have that authority until at least 2035. She notes that solar energy is the quickest way to generate electricity. "A gas turbine is going to cost at least twice as much as it did five years ago," Ince says. Additionally, it will take roughly five years to gain access to the larger gas turbines. Thus, we have observed that the new grid system is installing solar, storage, and possibly small modular gas in addition to co-locating with data centers. The quickest way forward has been that.

Calpine, one of Pantheon's investments, encompasses solar, geothermal, gas, and carbon-capture assets in addition to battery-energy storage system products. It and CyrusOne inked a cooperative contract in July to supply electricity for a new data center located in Texas. Purchasing AI energy suppliers may be more cost-effective than purchasing the core, similar to the hyperscalers. An agreement to deliver the primary civil works at the Sizewell C plant in Suffolk is among the energy-infrastructure contracts that Balfour Beatty (LSE: BBY) is snatching up.

According to Cutler, "the worst-case scenario for Constellation Energy, Dominion Energy, or Duke Energy is that they build a nuclear-power plant on stack for AI, and AI falls on its face for a while." "You just invested £20 billion to construct the plant, which will sit idle for a while as it takes five to ten years to build. However, Cutler states, "I'm happy with that" if Balfour Beatty is being compensated for building the plants. The current price of Balfour Beatty is about 13 times its expected earnings. The lucrative conflict at the core of AI is constantly being pushed farther and farther away from the value opportunities that exist there. Cutler explains, "We're always changing where we see the edge of the play, where the expectations are the lowest."

TSMC, ASML, Broadcom, Nvidia, and KLA are all owned by the Franklin Technology Fund. TSMC ranks among the top ten holdings in the Global Balanced Fund from Orbis Investments.