Positioning for the AI Revolution

Artificial intelligence has exploded from a niche topic to an everyday conversation in the public square. Politicians, CEOs, financial analysts, teachers, writers, and actors all have urgent questions. There’s enthusiasm, and there’s terror. While ChatGPT and other generative AI techniques can feel like magic, the foundational progress for this technology has been building for some time. At Eagle, we’ve been wrestling with the AI roadmap for years, attuned both to new developments and evolutions that have been quietly humming in the background. The roadmap is still a bit hazy—think Google Maps circa 2005—and we are curious as well as cautious as we anticipate what’s to come.

Our portfolio has been an early beneficiary of AI. Several of Eagle’s core holdings have invested in this area for the past decade and are well-positioned for the future. Companies like Alphabet and Amazon have historically been cheaper than they screened because earnings were burdened by R&D spent to future-proof the business and build portfolios of call options. As with any new technology, the pace of change and the precise way it is utilized is difficult to forecast. One of us rode in a Google self-driving car as part of a management meeting almost a decade ago. Our next research meeting was spent debating how self-driving vehicles might soon disrupt many industries. We were early. By contrast, generative AI emerged with stunning results that felt instantaneous. In between these extremes has been incremental progress. Think text messages that autocorrect, personalized entertainment, and smart thermostats.

It is helpful to organize AI into three categories: 1) Automation, 2) Personalization, and 3) Generative AI. Automation enables machines to “think” and do human management and maintenance tasks. Smart HVAC, data center power management, self-driving cars, and credit card fraud management all fall into this category. The technology is highly utilized by certain digital leaders but still has a long runway in the broader economy. Personalization enables the customization of digital offerings to consumers. This includes movie recommendations, targeted advertising, and individualized shopping experiences. This has been, and continues to be, a source of advantage for digital leaders. Generative AI is the “creation” we see with services such as ChatGPT for textual content or DALL-E for visual. This is at an embryonic stage.

In the coming years, the deployment of AI will impact most digital businesses and spread broadly into other sectors. While Eagle won’t get every twist and turn precisely right, change creates good stock selection opportunities. In some cases, we can invest in companies with undervalued exposure to new growth prospects. In other cases, the path will be to invert: in a market facing higher creative destruction, avoiding losers is as powerful as picking winners. We want to share several of our frameworks and how the portfolio is positioned.

Distribution is defense

The Internet was disruptive to distribution monopolies. Retailers, media networks, radio stations, and newspapers controlled local markets. The emergence of internet distribution, typically with greater scale efficiencies, upended these monopolies as we knew them. AI, in turn, will significantly impact intellectual property production. If “content was king” as the internet scaled, “distribution is defense” within this next wave of disruption.

For example, Meta benefits from content creators having better tools. Instagram continues to siphon market share from other entertainment sources. The platform is well-suited for experimenting with new tools, forms, and styles of delivery. Meta’s management understands this and has chosen to become a leading provider of open-source generative AI software, effectively lowering the barrier to entry. As its AI algorithms curate personalized feeds across its vast distribution, Meta—with three billion daily users—is the biggest beneficiary of this new mode of content creation. Meta’s plan to offer AI customer-service-agents to businesses using WhatsApp is another opportunity.

Alphabet has similar advantages. Google Search, Chrome, Maps, Android, and YouTube are among the most distributed products in the world, with billions of direct user relationships. While AI introduces uncertainty and risk into several of its businesses, the company has levers to fend off attacks. Google can add innovation to its distribution faster than a disruptor can build user scale. An early example is the fading hype around Microsoft Bing. Meanwhile, Google has introduced more sophisticated AI into Search without a noticeable loss of market share.

Without distribution control, Hollywood studios, music labels, and video game publishers face graver risks from the deflationary force of generative AI. The ongoing strikes by writers and actors in Hollywood have inherent drama, but news coverage of their impact overlooks the situation’s nuance and uncertainty. AI might well benefit studios in the short term while also introducing unforeseen challenges in years to come—if an emergent technology lowers barriers to entry, the role of the studio will have diminishing returns.

Netflix is a hybrid situation. It has the largest and most valuable distribution for high-quality video content, with comprehensive user data enabling it to customize feeds. However, it is also a scaled content producer, and the billions spent to “be in the game” are a competitive moat. Sensitive to these complexities, we’re watching this develop.

Data moats

Large proprietary data sets are increasing in value. Data is the raw material used to train machine learning models. Digital leaders like Alphabet and Meta have amassed massive troves of commercial and consumer data, which they have harnessed to assume AI leadership positions. While both companies must ward off challenges, access to this data bestows enormous advantages.

What gets less attention is that there are also non-digital incumbents with robust data sets that now have an opportunity to create and capture additional value. UnitedHealth, Aon, Capital One, and Bayer have data leadership positions in their respective industries.

UnitedHealth has the broadest, most extensive set of patient and medical data. The complicated regulatory structure and poor connectivity between disparate systems across the industry make it difficult to harness this data fully. That said, machine learning tools now offer enough incremental value that it is worth making the investment. United has significant opportunities to drive lower costs and better patient outcomes.

Aon is one of two global insurance brokers that sees enough underwriting to have differentiated data. The leadership team has spent over a decade migrating the company’s systems into more integrated platforms. This data advantage is now leading to a product advantage. Aon can build custom products to match commercial needs with non-traditional pools of capital. For example, a pension fund can allocate capital to bespoke Aon insurance products. This provides the pension fund with a new source of uncorrelated return and provides the insurance buyer with better pricing. For years, smaller brokerages have gained share by poaching talent from the big firms. This may reverse as machine learning and other data-driven tools enhance Aon’s offerings to talent and clients.

Capital One is one of the largest credit card issuers in the U.S. and has technology weaved into its DNA. Its single-digit earnings multiple reflects the market’s dour view of the company as an old-line business. Yet, it continues to innovate ahead of the industry. Several years ago, it closed its last data center, having moved its workloads to Amazon Web Services (“AWS”). It is one of Snowflake’s largest customers, if not the largest. Its peers are years behind. A single view of customer data allows Capital One to make swift and accurate risk and marketing decisions, driving best-in-class returns on equity and potentially widening its advantage.

Bayer is the largest crop science company in the world. Its seeds and traits business has been built over decades, with entrenched intellectual property and yield data from around the globe. The innovation model is that as Bayer improves the yield of its seeds, it prices to value, sharing the incremental economics with the farmer. If a better seed increases farmer profitability by $1, Bayer can raise the price by $0.50, and both parties win. Advances in machine learning may help extend and enhance the company’s R&D engine. This can drive both better earnings growth and improved farmer outcomes.

Atoms versus bits

“Atoms businesses” operate in the physical world, whereas “bits businesses” are primarily digital—a manufacturer, for instance, versus a software company. We expect AI to unleash a new wave of creative destruction on many bits business models. As the range of outcomes widens, investors should be better compensated for taking on this uncertainty. Atoms businesses should be comparatively less impacted.

Eagle owns several leading aerospace companies and hotel brands. While digital technology is an enabler, they are primarily atoms businesses. AI will be a tool that they deploy, but it is unlikely to fundamentally change the nature of these companies or how they compete. We have all experienced situations wherein newly introduced “smart tech”—a hotel room’s thermostat with a mind of its own, contactless check-in with digital keys that fail to turn—overcomplicates user experience and creates more friction than the analog method it was meant to improve. While these systems will be refined with time, their introduction does not represent a significant emerging disruption risk.

Over the past year, Eagle built core positions in ConocoPhillips and Shell. Energy production is one of the most fundamental atoms business. Despite AI’s roadmap to improve human intelligence, it is unlikely to revolutionize the nature of this market. As we invert by avoiding areas facing rising pressure, the energy industry stands out as comparatively undisturbed. In fact, to the extent that AI computing is enormously energy-intensive, it will likely provide a modest tailwind to demand.

Bits versus atoms

Amazon, in which Eagle has a position, has an unusual combination of strengths in both bits and atoms. Its digital storefront drives incredible organic traffic without paying the same level of “tax” to Alphabet or Meta that peers must. Its digital advertising business is now one of the world’s largest. At the same time, its physical logistics footprint has now surpassed both UPS and FedEx.

Advances in AI benefit Amazon because its most formidable competitors are brick-and-mortar firms like Walmart and Target. In many industries, deployment of AI will be zero-sum as any temporary advantage is competed away. But in Amazon’s case, it can offer increasingly personalized shopping and advertising experiences that can’t be replicated in the physical world. We see this as one of several levers driving improving margins and returns on capital across Amazon’s retail assets.

Application software versus infrastructure software

AI is changing the way software is built and consumed. Assistant tools such as Microsoft’s GitHub Copilot are increasing engineer productivity. Soon, there will be software writing software, with only high-level human direction. This will both grow the market and introduce deflationary forces.

A possible analog is the history of the semiconductor market. The blistering pace of Moore’s law innovation drove demand, but it also weakened pricing power. Only the most innovative and competitively advantaged companies survived. This is vastly different from what the software industry is accustomed to, where being average has still been pretty good for profits. Our early hypothesis is that many small-to-medium-sized application software providers are vulnerable to looming creative destruction. Without access to the broad data sets integral to AI, they may be pushed out by core providers like Microsoft or commoditized by cheaper alternatives.

By contrast, we think hyper-scale cloud platforms such as AWS, Microsoft Azure, and Google Cloud Platform are well positioned. These companies provide the scaled infrastructure that modern computing is being built on. AI-driven acceleration in this trend is a tailwind to growth. The nature of training machine learning models is episodic. Cloud infrastructure is better suited to handle this lumpy demand than on-premise data centers. While the hyper-scalers have a small share of legacy IT spend, they will operate the vast majority of AI workloads.

The road ahead

It’s early, and we remain sober. As long-term investors, we must be imaginative when looking around corners about what might go right or wrong. At the same time, we can’t succumb to futurology unanchored by facts on the ground. Some of the enthusiasm around AI this year relates to actual new product development. Much is simply a recognition of trends that were already in place. We expect a rich opportunity set in the coming years as a new period of change unfolds.

As always, if you have any questions or would like to discuss anything herein, please call us at (212) 293-4040. Also, if your financial situation or investment objectives have changed, if your IPO eligibility or U.S. Person status has changed, or if you would like to modify or discuss any investment restrictions or guidelines, please reach out to your contact on the client team or email ClientServices@eaglecap.com.¹