In Volume VIII of Talking Shop, a little over two years ago, we discussed the emergence of artificial intelligence, catalyzed by the release of ChatGPT, which quickly became a viral sensation in early 2023. Since then, the pace of innovation has only accelerated. While OpenAI has continued to introduce better versions of ChatGPT, other players have introduced models such as Claude, Gemini, Grok, LLaMA, Perplexity, among others. China-based DeepSeek caused waves with the introduction of a much lower-cost model back in January.   

To stay at the leading edge of understanding this trend, we recently attended ServiceNow’s Knowledge 2025 conference – which was entirely dedicated to advancements in AI. 

Over the past two years, the largest technology companies have invested aggressively in the infrastructure needed to support AI. We estimate that the top four “hyperscalers” (Microsoft, Amazon, Meta, and Google) spent a combined $300 billion in 2023 and 2024, and are expected to spend another $300 billion in 2025 on AI-related data center infrastructure. That’s $600 billion in just three years, in the U.S. alone. 

This spending goes far beyond advanced chips. It includes high-end networking gear, electrical componentry, fiber and copper interconnects, HVAC systems for cooling, and large-scale construction. It’s a full-stack infrastructure buildout. 

To put it in perspective: the combined cost of the Apollo, Gemini, and Mercury space programs totaled around $330 billion, adjusted for inflation. The interstate highway system cost $500-$600 billion in today’s dollars. The dot-com internet buildout of the late 1990’s is estimated to have cost $300-400 billion domestically.  

While much of the early buzz around AI focused on writing essays, generating images, or answering questions, the term Agentic AI has emerged and along with it a new wave of real-world applications. “Agents” are the next evolution of AI, having developed into systems that not only analyze information, but take action on it. ServiceNow’s Knowledge 2025 conference was squarely focused on making this future a reality today.  

During the keynote, Nvidia CEO Jensen Huang called ServiceNow the “AI operating system for the enterprise.” ServiceNow is uniquely positioned to offer AI agents, given its software is sold into most functions of an enterprise already, and critically, has access to the data needed to make these agents function.  

In a simple example, traditional AI might tell you how to book a flight. An AI agent will compare prices, choose the best option based on your preferences, and book it automatically. Booking.com is working on this.  

For ServiceNow, the conference showcased many use cases and examples of this technology from customer service and sales to human resources and IT. In many instances, the agents worked alongside employees, freeing them up to spend more time on higher value tasks. Just two years ago, AI was used primarily to summarize customer complaints – now, it’s resolving them. Witnessing these capabilities firsthand was incredibly valuable.  

With 25,000 individuals in attendance, the excitement was palpable. When the keynote ended, hundreds of attendees rushed to the “build your own agent” exhibit. The line looked more like a movie premiere than a tech demo. 

It’s still unclear how fast this technology will be adopted and how useful it will prove to be in practice. But one thing is certain: hundreds of billions of dollars are being spent and customers and investors will expect a return. 

We believe Agentic AI is part of that answer, unlocking new levels of productivity, automation, and innovation across industries. As that unfolds, the demand for supporting infrastructure will only grow. 

Making use of the market volatility earlier this year, we’ve leaned into this trend by adding or initiating positions in companies we believe are well-positioned to benefit from the AI infrastructure buildout such as Arista Networks, a provider of high-end networking switches and routers essential for AI data centers, Amphenol, a supplier of copper and fiber optic interconnects that form the data backbone, AECOM, a key player in large-scale construction and engineering services for data center expansion, and Trane Technologies, a global leader in HVAC equipment and services.  

Even with substantial investment already made, we believe this is just the beginning. As more AI-driven applications emerge, the need for advanced infrastructure will continue to accelerate. We remain focused on identifying and investing in the companies driving and enabling this transformation. 

Best Regards,  

The Sandhill Research Team 

Disclaimer: This commentary is for informational purposes only and does not constitute investment advice, a recommendation, or an offer to buy or sell any securities. Sandhill Investment Management (“Sandhill”) is a registered investment adviser with the U.S. Securities and Exchange Commission. Views expressed reflect Sandhill’s opinions as of the commentary date and are subject to change without notice. Forward-looking statements are speculative and involve risks and uncertainties that may cause actual outcomes to differ materially. All investments involve risk, including the potential loss of principal. Economic and market commentary is based on publicly available information believed to be reliable, but its accuracy or completeness cannot be guaranteed. Investors should carefully consider their individual circumstances, objectives, and risk tolerance before making any investment decisions, and consult with financial, legal, or tax professionals as needed.