Artificial Intelligence (AI) has the potential to be a transformative technology that impacts how we all live and do business. With some creativity, and time for current offerings to improve, it is not difficult to imagine how in the future these systems could enhance or even replace much of the work that we humans are currently doing. Late in 2025, none of this is breaking news and the relevant questions for investors are ultimately: what is the utility of these new tools in dollar terms and when should we expect to see the benefit? More use cases uncovered, or benefits realized sooner than expected would add additional tailwinds to the AI theme, while the opposite also holds. At the index level, this is what markets are grappling with — the excitement surrounding how impactful (and ideally profitable) these technologies could be in the future, tempered by the concern that the expectations implied by current valuations are too high.Hyperscaler Capital ExpendituresSource: LPL Research, Bloomberg (Consensus) 11/05/25The AI hyperscalers, included in the “Hyperscaler Capital Expenditures” chart above, are fully convinced that this is a race worth winning, which is reflected in their capital expenditures (capex) growth. Across the AI theme, the investments for the supporting infrastructure (semiconductors/datacenters/energy/talent/compute) required to “stay in the game” are staggering, and the large language model business that is leveraging these new technologies is currently unprofitable. While companies like OpenAI are not profitable today, they are projecting strong growth and a path to profitability that has kept investors interested. This has allowed chipmakers, most notably NVIDIA (NVDA), to benefit from booming demand while end users work on a path to profitability. As a result, NVDA briefly achieved a $5 trillion market cap in October, has plenty of cash coming in, and can invest in its customers.These investments could be interpreted as a vote of confidence that the users downstream will crack the profitability code, and an investment that reflects the desire to participate in the resulting growth. A more pessimistic take would be that this circular financing is being used to buttress the financial position of unprofitable business lines to maintain demand for chips. The AI ecosystem has many such relationships, as outlined in the graphic below.How NVIDIA and OpenAI Fuel the AI Money MachineSource: LPL Research, Bloomberg 10/08/25In aggregate, investors have welcomed AI dealmaking and driven share prices higher following deal announcements. That said, we are watching for signs that enthusiasm may be waning. For example, we wrote about Remaining Performance Obligations (RPOs) in September when a large deal was announced between OpenAI and Oracle (ORCL). Following the announcement, ORCL shares rose sharply but are now trading closer to pre-announcement levels.To keep up with the AI race, companies are leveraging multiple sources of capital, including Special Purpose Vehicles (SPVs). These SPVs are separate companies with their own balance sheets created by a parent company to isolate and share financial risk with other investors and creditors. While on its own, this is not nefarious, history is littered with the corporate remains of financial engineers finding clever ways to manage (i.e. hide) balance sheet risk. SPV’s catch our attention as this is how Enron hid its fraud and how banks took on much higher leverage during the great financial crisis. Regardless, it is another development that we are monitoring as companies are going to greater lengths to keep investing.While the number of items on our list of AI theme concerns and the amount of dollars involved has grown, this general line of reasoning with respect to AI is not new. We ourselves have written similarly titled notes with similar messages over the past few years, and the AI bellwether stocks have continued to march impressively higher. All of that to say, while we believe these concerns are legitimate and each dollar invested in AI raises the bar for the expected payoff, the momentum behind this theme is strong and it is not something we are ready to step in the way of today. Earnings expectations are still growing; the synergies between AI and quantum computing remain unknown, and the current administration appears aligned with AI initiatives.