The AI tide rolls on

Over the past nearly four years, we have characterized succeeding iterations of the AI era as “the year of training and graphics processing units (GPUs)”, “the year of inference”, “the year of the agent”, etc. Most recently, we highlighted the surge in demand for central processing units (CPU), memory chips, optics, and other key AI inputs. An apt moniker for the next phase of the AI rollout is “all of the above”, or more specifically, “supply bottlenecks in all of the above”.

AI deployment has proceeded like a wave, with the initial splash of ChatGPT’s 2022 release expanding in concentric circles, impacting islands of the tech sector that many had thought would not be a major part of the story. The degree to which these technical inputs are required is reflected in their suppliers’ order backlogs, which, in many cases, are measured in years.

It’s difficult to overstate how these supply constraints could dictate tech’s prospects over the coming years. We are witnessing demand for AI applications – and thus AI-enabling components – grow by an order of magnitude. Supply does not move so fast: Data centers must be permitted and constructed, critical inputs must be produced, and labor to build these facilities and electricity to power them must be procured.

This confluence of bottlenecks, in our view, should act as a governor that will keep supply growth running well below that of demand into 2028 – and for many components, beyond.

Cyclical is the new secular?

Even while we are still in what we call the AI enablement stage, with its emphasis on AI infrastructure, demand for access to frontier models is exploding. This has led to two noteworthy developments: First, it has allowed AI hyperscaler revenue to reach an inflection point, validating our thesis that these massive investments are already leading to monetization. Second, it has provided visibility into earnings streams for many component providers.

Semiconductor manufacturers, for example, have long been dogged by the specter of cyclicality, perhaps most notably in memory. Industry consolidation, along with swollen order books, has allowed investors to value the space without the trepidation that the industry may be setting itself up for yet another earnings-killing inventory overhang. Furthermore, this unique earnings visibility could merit investors rewarding these stocks with higher valuations.

A bout of sticker shock?

In nearly any other context, the scale of AI infrastructure investment – measured in hundreds of billions of dollars annually through the end of this decade – would be considered implausible. This has sparked debate among investors.

Count us in the camp of the expenditure being merited. As intimated, we are still in the AI enablement phase. Once companies understand how to amplify this novel technology’s potential and end users deploy it across the global economy at scale, we believe demand for AI applications – and just as importantly, their ability to be a force multiplier in productivity – will surpass all but the most bullish estimates. Consequently, we believe valuations of AI infrastructure companies remain reasonable.

As illustrated by a few stretches of elevated volatility this year, the market has had a possibly understandable level of sticker shock with respect to these investments. While selloffs can leave investors feeling seasick, we leverage company-specific analysis to determine whether the fundamentals underpinning a stock’s thesis are holding – and in most cases regarding AI, they are.

The role of the capital market

Another development that could send more waves across the AI ecosystem is the expanding capital market cycle. Excitement around recent public offerings on the back of a robust venture-funding environment illustrates that capital is flowing to innovative AI companies that could shake up existing industries and business models and create new ones. We’ve seen this type of creative destruction in previous cycles, when disruptive upstarts led to personal computer makers displacing workstations, software-as-service (SaaS) companies coming to dominate their industry, and e-commerce fundamentally changing how households shop.

It is in periods like this where allocating capital toward its most productive use can have the greatest impact: by rewarding anticipated winners while minimizing exposure to those on the wrong side of the AI divide.

Past as precedent – just faster

As we’ve seen throughout the nascent AI era, this sorting out seems to be unfolding faster this time around. Case in point: the market aggressively discounting the value of legacy companies that either appear at risk of AI disintermediation or whose management teams fail to grasp the historic opportunity presented by this technology, leaving them vulnerable to more forward-thinking peers.

Discussion of any capital market cycle invariably leads to questions about bubbles. Referencing previous cycles in which there was notable froth, we believe the AI natives coming to market and hyperscalers seeking financing –  often through debt markets – have made a compelling argument that these investments are justified.

Bad ideas, however, often follow good ones. In a typical fundraising cycle, investors apply considerable rigor to test an issuer’s assumptions and, if persuaded, deploy capital accordingly. This is where the market is now. As attractive returns are generated by validated business models and use cases, a number of late-arriving “me-too” investors pile in. This, in our view, is the classic sign of a bubble, but we are not there yet. Once AI’s version of pets.com receives a unicorn-type valuation, investors can then sound the alarm.

Unfurling the sails

Scientists and sailors are usually the only ones who understand the true power of the oceans and their tides. The rest of us tend to underestimate it. The same holds true for AI.

The management teams of leading AI companies have not ramped up CapEx on a whim; they understand both the opportunity and the competitive threat.

We maintain the view that this technology will impact nearly all aspects of the world’s economy, boosting productivity and discovery across the commercial sector, government, science, and other applications.

As illustrated by AI’s surprising inference stage, momentum around AI is cresting like a wave.  And there are myriad developments on the horizon – some anticipated, some inevitably not. We are monitoring developments in the neocloud, sovereign AI, distilled models, open-weight models, AI inflation, China, security, and the regulatory backdrop, including any public backlash.

A broadening conversation

Two years ago, colleagues from other sector teams sought our views on the latest developments in frontier models and semiconductor advancements. That still occurs, but it’s more of a two-way conversation. Our technology team now seeks insight from our industrials, utilities, and energy teams.

As hyperscalers increasingly tap debt markets to fund CapEx initiatives, we are collaborating more closely with investment-grade and high-yield credit colleagues. And with AI inflation rippling across the economy, we have plenty to talk about with our bond teams regarding the trajectory of industrial input and consumer prices.

These conversations illustrate the likely breadth of AI’s future reach. The upshot is that AI, in our view, is a generational theme. We believe investors should seek to understand how it could directly and indirectly impact portfolios on multiple levels: through raising the trajectory of corporate earnings growth, acting as a disruptive force, and upending entire industries and influencing demand – and thus prices – for key industrial inputs.

IMPORTANT INFORMATION

Artificial intelligence (“AI”) focused companies, including those that develop or utilize AI technologies, may face rapid product obsolescence, intense competition, and increased regulatory scrutiny. These companies often rely heavily on intellectual property, invest significantly in research and development, and depend on maintaining and growing consumer demand. Their securities may be more volatile than those of companies offering more established technologies and may be affected by risks tied to the use of AI in business operations, including legal liability or reputational harm.

Technology industries can be significantly affected by obsolescence of existing technology, short product cycles, falling prices and profits, competition from new market entrants, and general economic conditions. A concentrated investment in a single industry could be more volatile than the performance of less concentrated investments and the market as a whole.