Our latest Disruptor SeriesTM episode examines the trajectory of AI—from build-out to physical AI.
In part one of the AB Disruptor Series “Mind in the Machine” episode covering artificial intelligence (AI), we explored the supply chain that powers AI. In part two (watch the full episode here), we looked at the big spenders in the AI build-out as well as its beneficiaries and the companies innovating in the AI realm. Here’s a brief summary of the episode.
The AI “S” Curves: Infrastructure, Inferencing and Physical AI
It helps to think of the road map of AI as a series of three S curves, which illustrate the typical pattern of growth for the adoption of new technologies, products or processes. The first S curve, infrastructure and training, is the build-out of the backbone to support AI capabilities: data centers, servers, graphic processing units and other hardware components.
The second S curve is for inferencing—the adoption and application of trained AI models to a myriad of real-economy use cases, whether it’s producing content, answering questions or making predictions. AI has the potential to disrupt every industry, but mass adoption is needed for its impact to be maximized. With physical AI, the third S curve, AI is being deployed to develop physical systems, such as autonomous robots and drones. Still in the very early stages, this stage will likely be the most transformative aspect of AI.
The companies operating in the AI ecosystem are often grouped into two high-level buckets: semiconductor and hardware providers on one hand and software firms on the other. We think it’s helpful to think about them a little differently—reframing them as AI spenders and AI beneficiaries.
Titans Power the Spending Surge of the AI Build-Out
To this point, the infrastructure and training S curve has seen the most progress and is in the rapid growth phase, with a massive wave of capital spending under way from the titans of AI. The inference S curve is just getting started, as monetization of the models begins to pick up. Spending priorities aren’t monolithic, though. Hyperscalers are the core buyers of AI data-center infrastructure. Microsoft, Google, Amazon and Meta Platforms are forecast to spend nearly $400 billion this year. Oracle and AI cloud specialists are also increasingly relevant.
Frontier model developers are major demand drivers, striving to develop large language models that are more accurate and useful for personal and professional needs. The next wave of AI demand will likely come from government buyers, including the US and European Union, seeking to ramp up their AI capabilities. We’ll also likely see more demand from enterprises modernizing private or “hybrid” AI stacks to run large language models on proprietary data. Collectively, the wave of AI spending is creating a windfall for other firms.
Potential Beneficiaries of the AI Build-Out Are Wide-Ranging
The AI infrastructure build-out is well under way and capital commitments stretch into the 2030s. The immediate beneficiaries of this rising tide are the companies that provide what’s needed to bring data centers to life and keep them humming. NVIDIA, Broadcom and Taiwan Semiconductor Manufacturing are the poster children, as chip spending skyrockets to add computing power. There are needs in building and maintaining data centers, too, whether it’s networking components, efficient power supply and cooling systems, or physical and cybersecurity measures. Hardware and semiconductor firms benefit if the pie continues to grow.
Software firms stand to benefit from AI adoption, too, developing AI applications, infrastructure software, security software and tailored applications for business verticals. It’s a very dynamic and heterogeneous space, and we believe there will be winners and losers. AI is following the pattern of most of history’s technology cycles: it’s led by an infrastructure build-out today (spending on data centers and chips) but will be monetized over time via software. Beneficiaries will include incumbents that successfully monetize AI incrementally as well as a wave of AI-native start-ups that will be an increasingly disruptive force.
For small- and mid-cap investors, a broad field of companies provide both software and the semiconductors and hardware “plumbing” that enable AI. Often, they’re nimbler and more focused than the mega-cap titans, and some have unique value propositions—the trailblazers. They carry more potential upside if they win but also a sizable downside, particularly if they’re narrowly focused. It’s a very dynamic space where we see three categories of companies: those that will thrive, those that will survive and those that will slowly fade away.
We think firms at risk of fading away will be those most vulnerable to AI’s automation of human seats (like customer-support software or contact-center software). Also vulnerable could be vendors at risk of cannibalizing their own businesses with AI, along with other slow-moving firms. As AI evolves from ChatGPT prompts to AI agents that execute tasks for users, it should create potential for companies to thrive. AI agents create new bottlenecks, including ensuring secure access to those AI agents, which could benefit security companies. This shift also requires more computing capacity and ultimately more networking, which could stoke demand for firms that provide AI interconnect offerings through optical lasers or cables.
Robots and Drones: Physical AI and the Road Ahead for AI
The physical AI S curve is still in the early stages. If inferencing means humans using AI are able to be more efficient, physical AI is in a way a “supernova” for AI. With physical AI, AI is integrated into robots, assembly lines, drones and other machines. While NVIDIA seems like a winner on both the infrastructure build phase for physical AI and the deployment phase as robots move into the real world, other firms such as Amazon should be able to double dip as the S curve from physical AI ramps up. The company seems firmly entrenched in the AI build-out and eventual payoff from mass AI adoption, and we think its physical infrastructure is also highly likely to benefit from AI robotics.
There’s a rousing debate over the productivity impact of AI and its potential to eliminate jobs while creating new ones. But its innovative potential is undeniable, as more and more companies—both new economy and old economy—deploy it to boost efficiency and make better products. For investors, there are opportunities in AI spenders and beneficiaries, but also in traditional, non-AI firms that use AI to get a leg up on the competition.
AB’s Disruptor Series provides distinctive perspectives on critical issues facing the capital markets today.