3 market experts share the 8 companies that will profit the most from AI's insatiable energy appetite

A man uses his laptop with AI capabilities

dowell/Getty Images

• Energy supply is emerging as a major bottleneck to the AI revolution.
• Companies that can maximize AI energy efficiency will emerge as the winners, experts say.
• 3 market experts share where they see the most opportunity within the AI energy trade.

AI development might be running into a problem soon: the electric grid doesn't have enough juice.

Data centers and the cloud computing infrastructure housed inside them are voracious power consumers. But after years of stagnant electricity consumption and neglected infrastructure, experts aren't sure if the US power grid is ready to handle this sudden surge in power demand. Earlier this year, Meta CEO Mark Zuckerberg declared energy the foremost bottleneck to AI progress.

And as companies race to develop newer and smarter models, energy consumption will only go up. According to Epoch AI, a nonprofit research institute specializing in AI development, the computational power demands for AI are doubling every nine months. Moody's estimates that between 2023 and 2028, the electricity usage in data centers will grow 43% yearly. That's not including the energy-intensive process of constructing even more data centers, which Big Tech is scrambling to do.

For those investing in the AI trade, the energy requirements could pose a potential hurdle to AI's adoption and monetization. Business Insider spoke with three market experts to get their thoughts on what companies are best positioned to succeed as AI's appetite for energy strains existing infrastructure.

Jennifer Foster, co-chief investment officer at Chilton Investment Company

Energy constraints could pose a serious threat to AI's growth, according to Foster. She points out that states like Virginia have begun introducing legislation to limit new data center buildouts due to concerns about the electric grid.

Foster believes there are beneficiaries among both the infrastructure hardware and software sides of the AI value chain.

On the hardware side, Foster sees optimization opportunities in AI architecture that spend less energy.

Both Broadcom (AVGO) and Marvell Technology (MRVL) specialize in producing ASIC chips, which are customizable to the unique demands of a specific application or task. Traditional GPUs are extremely useful for general AI models, but their broad capabilities aren't always fully utilized for AI, especially AI applications that don't involve a graphics component, according to Foster. That results in a lot of wasted energy. As a result of their more specialized workloads, ASIC chips provide optimized performance and lower power consumption.

On the software side, Foster sees IBM (IBM) as a standout company. IBM is developing smaller LLMs that are more efficient and tailored to a specific use case. Unlike ChatGPT, which is trained on data from all over the internet, IBM's Granite model LLM is trained on a more select dataset to assist in automating certain routine coding tasks. Smaller LLMs are helpful for enterprises to develop specific AI use cases and are also less energy intensive.

Graeme Baker, portfolio manager at Ninety One

According to Baker, who oversees Ninety One's Global Environment Strategy, the future of AI is heavily linked to sustainable energy sources, as Big Tech companies need to juggle increased energy demand from data centers and their net zero goals. Baker is betting on the utilities and cooling solutions providers that service the energy needs of data centers.

He believes Schneider Electric (SBGSY) will be a major beneficiary of increased energy demand from AI workloads. The company has a significant proportion of data center clients that use its energy management equipment. Schneider Electric is also investing heavily in its renewable electricity offerings.

Trane Technologies (TT), a provider of HVAC and other indoor climate solutions, is another AI beneficiary, according to Baker. Data centers generate massive amounts of heat from processing AI workloads, and so data center cooling solutions will become increasingly important as energy demands increase, Baker said.

Jakob Wilhelmus, director of thematic research at PGIM

Wilhelmus believes the existing electric grid needs to be upgraded to support the energy demands of AI computing.

New infrastructure needs to be built out, and the existing infrastructure needs to be bolstered, according to Wilhelmus.

For this reason, he likes Eaton (ETN), a power management company that provides energy storage solutions and manufactures renewable energy components. The company offers grid modernization solutions and builds out transmission lines to increase grid capacity. In Wilhelmus' view, the energy storage services that Eaton provides will be in high demand to support data centers' needs for a steady flow of uninterrupted power.

Wilhelmus is also bullish on the future of copper producers and mining companies such as Ero Copper (ERO) and Freeport-McMoRan (FCX). Copper is a key component of electrical conductivity, with copper wiring being used in power generation, transmission, distribution, and many other instances. With copper demand set to more than double in the next 25 years thanks to increased electrification needs, Wilhelmus expects these companies to take off.

Read the original article on Business Insider

Contributer : Business Insider https://ift.tt/HAmFbIl

Something that you might miss to see?