Company Touch Tracker
This quarter our team engaged with 389 companies across the world, putting our trailing 12-month company touches at 1,714.
In Q1, team members traveled domestically to Oklahoma, Texas, and Virginia with multiple trips to California, Florida, and New York. Internationally, team members traveled to Japan, Italy, Mexico, and Sweden, and interacted, whether by phone or at a conference, with companies in 31 other countries. We track these interactions and share them on our Company Touch Tracker1.
Conference Recap
Phil Naylor, Portfolio Manager, traveled to Japan and California for conferences this quarter. Here are some of his thoughts on why attending a research conference can be a valuable way to interact with a large number of portfolio companies:
At Grandeur Peak, we prefer to visit companies at their headquarters. There is no substitute for walking the factory floor, sitting across from a CEO in their own conference room, and absorbing the culture of a business firsthand. Investment bank-hosted conferences offer something different, and in some ways, complementary. When we sit through 39 meetings in seven business days, as I did in March across the Daiwa Investment Conference in Tokyo and Roth’s 38th Annual Conference in Dana Point, California, we do not get as deep with any given company as we would in our typical on-site visits. Instead, we are listening for patterns. We are hearing how dozens of management teams answer the same questions, and we are noticing where the consensus view is forming, and where it might be creating opportunities.
This quarter, that pattern recognition produced a clear takeaway: the physical infrastructure layer of the AI buildout is a far more compelling place to invest than the software and services layer trying to figure out what AI means for their business. The hardware story still has legs. The software story is full of uncertainties and some wishful thinking.
Tokyo: Atoms Over Bits
We attended the Daiwa conference in Tokyo, with several off-site meetings at company headquarters across the city. Daniel Xu and I met with 19 companies over five days spanning semiconductors, materials, precision instruments, IT services, SaaS, and consumer names. The meetings that moved the needle were overwhelmingly in the first two categories.
One standout was a Japanese company that is the dominant global supplier of ceramic material used inside electronics, which we hold in several funds. The company holds 100% share in AI-related applications because no one else can match the thermal and expansion characteristics of their material. The company’s AI revenue has grown from ¥2 billion in 2023 to ¥9 billion in 2024 to ¥15 billion this fiscal year, and management expects ¥24-26 billion next year. They are building two new plants, including an R&D center dedicated to AI and next-generation applications. What I found most interesting, though, was the optionality beyond data centers. They are beginning to collaborate with leading vendors of humanoid robotics, and they are also already supplying products for satellite-mounted lasers. The volumes in space are small today, but the customer base’s rhetoric around increased volumes provides a compelling upside. The competitive moats here are real: multi-year co-design cycles, materials science that Chinese entrants are years away from replicating, and a product that represents a tiny fraction of the customer’s total cost. Price is one of the last things customers optimize for.
A Taiwanese liquid cooling specialist that we hold in several funds, reinforced the same theme from a different angle. They produce cold plate components that sit directly on top of AI chips to pull heat away using circulating liquid. This company is the leading supplier for NVIDIA’s downstream liquid cooling and has over 1,000 engineers, with more than 100 dedicated to NVIDIA alone. But the more interesting story, now, is their diversification into more customized applications, where customers are less price-sensitive than NVIDIA and the chip architecture adopted by the two largest custom chip designers requires twice as many cooling plates per chip. Management was refreshingly candid that their cold plate share will normalize from 70-80% to roughly 50% over time. They simply cannot expand production capacity as fast as their largest customer is growing, but the higher margins on custom chip work help offset that shift.
Like the ceramics company, this business has early-stage opportunities in space. They are developing specialized cooling solutions for low-orbit satellites with two major constellation operators, and the selling price per unit is ten to twenty times what a standard cooling component fetches. This is not meaningful revenue today, but it, along with robotics, illustrates how expertise in managing heat is finding applications far beyond the data center.
The contrast with the Japanese IT services and SaaS names was sharp. We met with seven companies in this category, ranging from cloud accounting to identity security to IT consulting, and, unfortunately, came away uninspired. The consistent pattern was management teams acknowledging that AI will disrupt their sector while insisting that their specific company will be a net beneficiary. Nearly every team told us, in several ways, that their competitors would struggle but they would be fine. When you hear the same line from seven different companies in three days, it stops being reassuring and starts being a warning sign. The Japanese IT services sector has not yet internalized what AI means for labor-based billing models and margin structures. Importantly, the culture in Japan is one where labor is tight and companies are not willing to reduce headcount if the need arises. For the service businesses, this could mean that they are left with an underutilized bench and a cost structure that will pressure their margin profiles. On the software front, we found a new willingness of customers willing to try AI-native products, abandoning what they had become used to in the past. Though just one small example, one translator we used expressed she would switch the accounting platform she used to run her freelancing business if another player could offer roughly the same technology at half the cost. I left Tokyo increasingly cautious on these types of companies.
There were a few businesses we met which are likely being unjustly caught in the AI sell off. One was a Japanese company specializing in the distribution of press releases, which we visited at their headquarters in Akasaka. Their competitive moat is distribution reach and audience ownership, not content creation, and in an AI world, that distinction matters considerably. We believe platforms that own the audience and their attention are more defensible than those selling labor, content, or software that AI can replicate.
Dana Point: Identifying Quality Growth in the U.S. Small-Cap Universe
Two weeks after Tokyo, Spencer Hackett, Dane Nielson, and I attended Roth’s conference in Dana Point, California, meeting with 20 companies over two days. The lineup was classic Grandeur Peak territory: U.S. small and mid-cap industrials, specialty consumer, and a handful of technology names.
There were multiple meetings that stood out, including a Canadian direct-to-consumer optical retailer, which is run by a founder, who has executed this playbook before. Their ability to manufacture glasses at dramatically lower costs than traditional optical retail and to deliver them to customers faster creates a compelling unit economic advantage. This is the kind of founder-led, structurally advantaged business we are looking for. Another example is a specialty food producer with a credible path to accelerating growth through their quick-service restaurant channel. Management’s confidence in the Quick Service Restaurant (QSR) pipeline was backed by specifics, and it was clear they are making strides to professionalize their efforts. Finally, an electro-mechanical timing semiconductor company continues to impress on design wins and their pipeline, but at current valuations, it remains more of a watchlist name than an actionable buy.
The common thread at Roth was that the U.S. small-cap universe continues to offer high-quality businesses at reasonable valuations, particularly outside of the AI hype cycle. Many of these companies are founder-led, capital-light, and are growing earnings through clean execution. They are exactly the profile that we seek to drive long-term returns in our portfolios.
Reflection
While nothing replaces our boots-on-the-ground strategy of directly visiting worksites, seeing a high volume of companies in a condensed period of time at conferences allows us to quickly identify sector themes and evaluate our convictions. When dozens of management teams answer the same questions in the span of a few days, patterns emerge that might otherwise take months to surface. For this reason, we find that attending select conferences, used deliberately and in combination with our primary research approach, can be a highly useful element of our research process.
Thirty-nine meetings in seven days across two continents reinforced a conviction we have been building for the past year or so: the AI investment story is a hardware and materials story first. The companies supplying the physical infrastructure (thermal management, ceramic substrates, precision semiconductor equipment, etc.) have tangible moats, visible demand, and pricing power. The software and services companies that claim to be AI beneficiaries are still telling a story rather than
[1] Grandeur Peak Global Advisors – About Us – Global Footprint, https://grandeurpeakglobal.com/global-footprint/