Highlights from T. Rowe Price’s annual technology tour

For nearly two decades now, a large group of T. Rowe Price portfolio managers and analysts have traveled to Silicon Valley each year to meet with U.S. technology companies. This annual tour continues to be one of our most energizing and collaborative outings for identifying key trends and developments in the leading global center for high technology and innovation.

Below are highlights from our most recent tech tour in December 2024, which included both equity and credit analysts from T. Rowe Price Investment Management, Inc.

Key equity takeaways 

The market is craving use cases for artificial intelligence (AI) given the massive semiconductor infrastructure spending on AI seen over the last 18 months. One example that came up several times on our trip was autonomous driving.

Acceleration of AI computing and improvements in hardware systems have this technology on the cusp of scaling commercially. We expect autonomous driving to take another significant step forward in 2025. We believe there are certain companies in this space that are favorably positioned with potentially strong use cases and that already are in the process of scaling commercially.

We believe the U.S. regulatory environment will be favorable, and unit economics appear attractive based on our conversations. We see compelling long‑term value here.

There remains an insatiable demand for AI computing and hardware. Based on our meetings, the frenzied pace of infrastructure investment is not slowing down and out-year estimates of the total addressable market remain large. Opportunities continue to emerge as the big hyperscale customers focus even more on networking and their own custom silicon hardware programs.  

While AI has driven incredible demand for hardware in the early days, monetization on the enterprise software and application side has demanded some patience from investors. That said, based on our due diligence and meetings with companies, we think 2025 will be the year when monetization starts to come through the profit and loss statements in a more meaningful way. We expect this trend will ramp up through 2026 and beyond.

The major leap here will be the move from pure copilots to “agents,” allowing AI to act more autonomously and boosting productivity for end users. Highly sticky and mission-critical software providers all appear well positioned to capture this AI upside.

Key credit takeaways

Software

General environment: The overall tone from our meetings with software companies was generally positive, with comments regarding the demand environment remaining healthy.

A declining rate environment and the incoming Trump administration both should equate to a more favorable mergers and acquisition (M&A) environment for the software industry. Several companies we spoke with over the past few weeks say they are actively assessing M&A prospects, although valuations have not been cooperative. Given this backdrop, we expect increased deal flow from both strategic and private equity buyers in 2025. 

In the high yield market, declining interest rates should help highly levered software companies generate the most substantial free cash flow (FCF) that we have seen in the past two to three years.

That said, given the higher interest rate environment and resulting lack of FCF seen in recent years, sponsors also have been less capable of extracting value via dividends, asset sales, etc. We expect efforts to monetize private equity investments to pick up in 2025.

AI: As of today, clearly defined, business‑ready use cases for generative AI (genAI) are limited. However, early areas of adoption include software developer automation tools and efficiency offerings for the contact center/customer experience.

The latter of these two examples illustrates an interesting bifurcation in AI use cases that was discussed in several of our meetings. While some tools simply supplement workflows and drive efficiency for existing employees (call summarization in contact centers, for example), others are meant to displace workers (such as contact center agents).

Our discussions with software vendors highlighted the fact that chief information officers (CIOs) as potential buyers are grappling with the following questions when it comes to adding genAI capabilities to their contracts:

• How do I measure the return on investment of the proposed tool to justify the spending dollars? Should success be measured by reduced head count, increased efficiency per seat, or some other measure? If my employee saves two hours per day using their new contact center call summarization feature, will they spend that time taking additional calls or opt to take longer lunches?
• How much of the genAI functionality is just “table stakes” (call summarization, for example) versus incremental value that we should be comfortable paying for (such as call center agents)?

Several of the vendors that we spoke with are seeking to take away these CIO challenges by offering their genAI models on a free or heavily discounted basis for an introductory period. This strategy provides a low-stakes avenue for buyers to install the genAI tools and to take time to assess whether the value added is substantive enough to justify incremental spending.

From the software vendor’s perspective, this approach offers an opportunity to drive usage and show the potential efficiency/employee experience gains from the technology, thereby likely strengthening their case and offering optionality when it comes time to monetize.

While we’ve described some of the strategic barriers to the application of genAI technologies, there are also functional barriers that must be overcome—most notably, data today are highly dispersed across the enterprise tech estate, i.e., spread across on-premises and cloud environments globally. Enterprises must modernize their data to take best advantage of coming AI advancements.

Finally, large language models and related genAI applications are only as good as the datasets upon which they are run. Management teams at several vertical software companies claimed to have deep, industry-specific data advantages that they believe will translate to differentiated inference outcomes.

Hardware/Electronics/Networking

Hyperscaler cloud service providers are currently designing their next-generation data center architectures, which will cater to growing AI workloads while also serving traditional use cases. The design cycles that are expected to happen over the next few years on these reimagined data centers should offer major opportunities for content wins for vendors with innovative networking, storage, cooling, and logic/memory semiconductor solutions.

Hyperscalers generally rip and replace entire data centers rather than replacing components that have been designed out. This means that replacement of legacy technologies will happen in brand‑new data centers and those that are being refurbished at the end of their useful lives.

Due to the explosion of GPU usage in next-generation data centers, energy consumption will be a key constraint. Components that help reduce energy consumption will be at an advantage in the next wave of data center design.

Several companies that we spoke with agreed that the market is seeing demand pull forward today for AI-related components as hyperscalers, enterprises, and sovereign entities combat concerns that they will be left behind in the current arms race. This rhetoric points to a high likelihood that the AI-related hardware/semiconductor/networking component players could see a demand reversion as buyers digest their purchased supplies. However, the timing of this is uncertain.

Similarly to the software vendors we spoke with, companies in the hardware/electronics/networking spaces also appear to be keeping an active roster of M&A candidates.