AI's Forgotten Beneficiaries

The staggering cost of ubiquitous chatbots is a windfall for some.

Investor passion for anything to do with artificial intelligence has been the S&P 500’s savior so far in 2023. But it’s also been very much a zero sum game: Buying power flooding big technology stocks like Apple Inc (NSDQ: AAPL) and NVIDIA (NSDQ: NVDA) has come right out of pretty much everything else in the stock market.

And the losers include two sectors that are automatic beneficiaries from any mass move to AI: Electric utilities and the US Big 3 Telecommunications service providers.

The two main catalysts for increased AI interest over the past couple months are widespread adoption this year of Open AI’s ChatGPT, and the hyper-bullish guidance issued by NVIDIA Corp (NSDQ: NVDA), which makes the essential hardware and software.

ChatGPT is the first widely used application powered by what’s called “generative AI.” If you’ve even just casually experimented with ChatGPT—which is still accessible for free—it’s readily apparent it does a lot more than a Google search. With just a few prompts, ChatGPT can produce volumes of information on literally anything, as well as visual art.

NVIDIA CEO Jensen Huang has gone so far as to say generative AI will “augment” human output in practically every endeavor. And his company’s guidance for a record $11 billion in Q2 revenue set off a buying frenzy for NVIDIA shares, squeezing the considerable short interest and pushing market capitalization briefly over $1 trillion.

NVIDIA’s move has understandably ignited investor interest in any stock that could remotely be connected AI. So far, however, investors have ignored energy and telecoms. But if AI adoption come anywhere close to measuring up to hype, they’ll have plenty of reason to be taking notice soon.

Some analyses show a single session with ChatGPT has an all-in cost as much as 1,000 times higher than a Google search. Others estimate the application used as much electricity in January as would be needed to power a mid-size city. Small wonder then that a recent Biden Administration report called the computational costs of rapid AI adoption a “national concern.”

Many fortunes have been made and lost betting both on and against which technologies will ultimately prove to be transformative. I see two main reasons why generative AI will proliferate.  

First, it’s appearing at the same time industries across the board are “digitizing,” or utilizing data collection and processing technology. And its capabilities augment companies’ ability to make smarter decisions. The second is potential cost savings. Generative AI enables businesses to do the same volume of work with far fewer white collar workers, just as other automation technology has replaced blue collar jobs.

At the end of the day, the current version of generative AI has some pretty massive limitations. The main one is it’s only effective to the extent it has accurate data.

As most people are aware, there’s a great deal of contradictory and even outright false “information” posted on the web. So as a result, ChatGPT produced work is notorious for inaccuracies. Translated to other endeavors, fast thinking, fast moving generative AI is capable of making horrendous decisions based on faulty data that could prove impossible to correct.

To be sure, ChatGPT and its imitators are working to limit its shortcomings. But at this point, it’s a clear limiting factor. And potential cost savings that undermine revenue in a competitive environment aren’t worth the risk.

That still leaves plenty of places where AI can be rapidly adopted without risk to revenue. The key to how fast it will be: Data center operators.

Data centers already use massive amounts of electricity, powering robust sales growth at companies like Dominion Energy (NYSE: D). And moving from CPU (centralized processing units) to AI capable processing will require orders of magnitude more electricity they’ll have to purchase, one way or the other.

There’s also the challenge of coming up with the investment capital to make the upgrades: Even global data center leaders like Equinix (NSDQ: EQIX) currently spend more on CAPEX and dividends than they generate in operating cash flow.

That means they’ll likely deploy only in tandem with customers’ orders. But even so, implications for increased power use are massive.

MDU Resources (NYSE: MDU), for example, will boost its generation by 28 percent in North Dakota to provide electricity for Applied Digital’s (NSDQ: APLD) new data center. That’s a huge investment with a guaranteed return.

Utilities also stand to benefit from AI’s ability to help manage backup power and microgrids, as well as improve grid utilization and asset management. And regulators will be more incented than ever to promote utility investment, ensuring companies grow earnings and dividends.

As for telecom, investors are now mono-focused on as yet unconfirmed rumors Amazon.com (NSDQ: AMZN) will offer discount wireless service through Prime. And they’re ignoring the fact that Big 3 wireless carriers AT&T Inc (NYSE: T), T-Mobile US (NSDQ: TMUS) or Verizon Communications (NYSE: VZ) 5G and broadband networks are absolutely essential for handling massive connectivity needs of AI.

It’s unfortunately pretty clear the ban of China’s Huawei aborted a real launch of 5G in the US. But the success of Chinese telecoms with uptake and applications is a good sign US carriers will ultimately benefit.

There are also immense potential cost savings for telecoms adopting generative AI. The UK’s BT (London: BT, OTC: BTTGY), for example, expects to cut 55,000 jobs by the end of 2030.

Outside the Big 3, US telecom is best avoided, as even larger companies are now in what amounts to a death struggle of falling revenue and crushing debt. But US electric utilities and big telecoms are getting no credit at all for their place in the AI firmament. And that means opportunity.