Could AI Be a Disrupter to Big Tech?

By: Justin Larson, CFA Cue the eye rolls. Cue the “you don’t get it” crowd and the “this is different” gang. If you’ve found your way to this article, perhaps you’re expecting a prognostication of grand proportions. Clickbait, even. This piece is not that – it’s an interrogation of the current market narrative. A ground swell of skeptical and thoughtful investors are beginning to question the massive (read MASSIVE) spending commitments on AI technology by the world’s largest tech companies (and governments). It’s critical that the prudent, long-term investor begin asking questions, too.

Goldman Sachs estimates that over $1 trillion in capital expenditures will be made in coming years to build out infrastructure for AI use cases, with a cool $300 billion+ coming from the biggest US tech names. To their defense, several of these companies recently indicated demand is already here to support those investments, and the lack of capacity actually detracted from recent earnings potential. We think prudent investors would be wise to invite some skepticism into their minds.

On the Landscape

Let’s first understand where these companies (Alphabet, Amazon, Microsoft, Nvidia, Meta, etc.) sit in the AI value chain. In large part, they’re building the infrastructure – the cloud services or the semiconductor chips or the data centers to process unfathomable amounts of data that feed into AI models and their cloud computing services. They are constructing the building blocks of generative AI. Revenues from the infrastructure buildouts have been strong, in large part due to the fever pitch around the technology leading to insatiable demand. AI models, even more basic ones, are incredibly expensive to build and train, and as long as that remains the case with demand at ample levels, the revenues are likely to flow in. If these companies are able to continue scaling infrastructure buildouts, profitability from those endeavors looks consistent in the near term, but it’s not necessarily an innovative venture. Further, developers building out the use cases on these infrastructure platforms need to see revenue (and a lot of it) if they’re to continue building and refining their models with continued investment. According to Coatue Management, the cost to run an AI (LLM) model decreased by nearly 10x from January 2023 to June 2023, but data also suggests it still takes roughly $20 in cost to run one single Chat GPT prompt. If the price reduction continues, it’s great news for innovators building models, but doesn’t explicitly bring down costs to the infrastructure providers and certainly doesn’t improve profitability. Put simply, building an AI use case should become cheaper over time, but building a physical data center and additional power plants to support that compute power is unlikely to see the same benefit.

Most of these companies have AI applications as well (Gemini from Google, Meta AI from Meta) or have invested in AI applications (Microsoft into OpenAI). Revenues from AI applications have been nonexistent, exceptionally small, or not disclosed at all. The companies claim their AI applications are enhancing profitability in other areas of their businesses, such as creating more targeted advertising techniques or making their cloud platforms more desirable. OpenAI’s Chat GPT generated less than $4 billion in revenue for all of 2024 according to financial data and New York Times reporting. It’s not bad for the platform that was the fastest to reach 100 million users in history, but it’s certainly not been a cash cow, let alone profitable. So for the end use cases of AI, where does the revenue come from? And when? If you’ve used one of the many chatbots now available, odds are you’ve been somewhat impressed, but perhaps not at all. A convenient time saver, not a life changer. Hallucinations, poor sourcing, and an inability to contextualize makes them closer to a search engine than anything truly revolutionary. For now, at least. How useful and economically viable will these be in the near term? According to Census Bureau data from August 2024, on average, fewer than 10% of companies across industries had adopted or had plans to adopt AI tools in their businesses by now. This could mean the technology is either not quite there yet, or most just aren’t ready to use it or trust its outputs. Either way, if mass adoption is not right around the corner, how long can these massive investments go unrewarded before markets begin questioning the economics of this evolving, yet untested technology?

On Competition

Massive investments into internet technology was needed in the early days of the internet. And the returns came, though not immediately and not to companies that were expected to realize them. It was newcomers that built innovative ideas on top of what incumbents had spent incredible amounts of money laying ground work for. Those newcomers weren’t burdened by the sunken costs – they were free to innovate with next to no cost of capital, no hurdle rate. No expectations elevated by a successful past and a massive investment in a new future. History is littered with similar stories – in fact, coming for an incumbent, innovating your way to relevance - is as American as baseball (until football unseated that incumbent). From oil lamps and the radio to Blockbuster, MySpace, and AOL, regular disruption and innovation is commonplace.We think a major flaw in the AI narrative is the idea that the company that spends the most money will reap the greatest long term reward. The idea that the companies that are dominant now will continue to be the most dominant amidst a new era of innovation. What if this monstrous spending does nothing but spawn the next generation of innovators? What does that mean for the returns on these capital intensive investments behemoth tech companies are making? Will investors continue to put a massive premium on their stock prices?

On Value Accretion

One of the biggest questions investors should be asking if attempting to capture long term benefits of this AI revolution is: What companies/industries stand to gain the most from AI use cases? Rarely are the builders of new technologies the biggest beneficiaries. It is the new opportunities that technological progress ushers in where the greatest long term economic value lies. Those that built out the foundations of the internet saw excitement and short term profitability boosts, but in short order markets realized the true value was in what could be done with the technology, not its construction. In fact, the TCP/IP protocol that functions as the backbone of the modern internet and essentially allows computers to communicate with servers is not even owned or maintained by a company, but by a global non-profit that is run by members and volunteers. The companies that did and continue to lay submarine cables around the world’s oceans to boost connectivity don’t necessarily do so because it’s a profitable endeavor, but rather what that connectivity can then create for society and their businesses.

Let’s say some of the promising use cases come to fruition with AI. Perhaps not now, but let’s say in the next two decades. Will the societal and economic value be greater from the development of software from a big tech firm that can read and respond to your emails just as you would or is the benefit greater for the healthcare startup that can use advanced AI inference models in patient diagnostics to improve diagnostic accuracy and potentially save lives with early detection? Will the value of additional ad revenue for Meta enabled by targeted AI models be greater than for the materials company building AI-enabled smart highways that could all but eliminate traffic fatalities? These are futuristic use cases that are years from reality, if feasible at all, but the point remains: we rarely see the evolution of technology early in its development. Be ready for surprises and be ready for change.

In Conclusion

With this backdrop we must carefully reflect on current pricing in markets for these behemoth tech companies, along with considering the feasibility of excess returns on these massive investments in AI infrastructure. Perhaps this time is truly different and those that build out this new future will also be the ones capturing its most prolific benefits, but that would be the exception to history. We believe the entrepreneurial and innovative spirit that underlies our economy will continue to show its face. Society will adapt, new companies will emerge with new promise, and those that looked most skeptically at what markets viewed as certain in this moment, will stand to benefit from that prudence.

If you have any questions or would like more information on topics such as these, don’t hesitate to reach out to our team.

Click Here to Schedule

Disclosures

Content in this material is for general information only and not intended to provide specific recommendations for any individual.