|
||||
|
|
||||
The Loudest Room in Tech Has a Quiet Powerhouse
Every gold rush has its shovel sellers. While the investment world debates the fortunes of AI darlings like Nvidia and Microsoft, a quieter story is unfolding in the infrastructure layer beneath them — in the custom silicon, optical interconnects, and data centre networking chips that make the whole machine run.
Marvell Technology is one of the companies doing exactly that work. And understanding why it matters is a masterclass in how transformative technology cycles can create value in unexpected places.
Market Context: AI Infrastructure Is Entering Its Next Phase
We're roughly three years into the mainstream AI investment narrative. The first phase was about large language models, cloud computing capacity, and — above all else — GPUs. Nvidia became the poster child of the supercycle, and rightly so.
But here's where it gets interesting for attentive investors: AI infrastructure is maturing. Hyperscalers — the Googles, Amazons, and Microsofts of the world — are no longer content to simply buy general-purpose chips. They are increasingly designing their own custom silicon, known as ASICs (Application-Specific Integrated Circuits), tailored precisely to their AI workloads.
This shift is significant. Custom chips can be dramatically more efficient than off-the-shelf alternatives for specific tasks — and efficiency translates directly into cost savings at hyperscale. According to industry analysis, the custom silicon market for AI accelerators could grow into the hundreds of billions of dollars over the next decade.
Marvell sits at the centre of this trend. The company is a leading supplier of custom ASIC design services and co-packaged optics — the high-speed optical connections that allow data to move between chips at the blistering pace that modern AI demands. It also provides the networking chips that stitch together entire data centres.
In short: Marvell doesn't make the GPUs. It makes the nervous system that connects them.
The Educational Concept: What Is "Pick-and-Shovel" Investing?
If you've spent any time in investing circles, you've likely heard the phrase "pick-and-shovel investing" — but it's worth unpacking properly, because it's directly relevant here.
The metaphor comes from the California Gold Rush of the 1840s. Most prospectors who rushed west in search of gold came home empty-handed. But the merchants selling picks, shovels, denim jeans, and provisions? They made reliable money regardless of whether any individual prospector struck gold.
In investing terms, a pick-and-shovel play means focusing on the enablers of a trend rather than betting on which end-product winner will emerge. The logic is compelling for several reasons:
Reduced binary risk. You don't have to predict who "wins" the AI race. You just need the infrastructure to keep being built.
Diversified demand. A company supplying essential components to Google, Amazon, and Microsoft benefits from all of their growth simultaneously.
Stickier relationships. Custom chip design, in particular, creates deep, long-term partnerships. Once a hyperscaler co-develops an ASIC with a supplier, switching costs are enormous.
The trade-off? Infrastructure companies can be harder to understand, less headline-grabbing, and sometimes more cyclical. They also face their own competitive risks — this is not a risk-free category by any means.
Actionable Takeaway: How to Think About the Infrastructure Layer
You don't need to make any specific investment decisions based on this piece — but here are three genuinely useful ways to apply this framework to your own research:
Map the value chain. For any transformative technology you're excited about — AI, quantum computing, autonomous vehicles — draw out the entire value chain. Who supplies the suppliers? Who makes the tools that make the tools? These second and third-order beneficiaries are often less covered and, occasionally, less richly valued by the market.
Understand switching costs. One of the most durable sources of competitive advantage in business is how painful it is for a customer to leave. Custom chip design creates exactly that kind of stickiness. When researching any company, ask: how hard would it be for their biggest customers to walk away?
Watch capital expenditure announcements. When hyperscalers announce multi-billion-dollar data centre investment programmes — as they have repeatedly done throughout 2024 and into 2025 — that spending flows downstream. Following the capex is a legitimate way to identify sectors likely to benefit.
The broader lesson is this: in every great technology supercycle, the picks-and-shovels businesses deserve a seat at the research table alongside the marquee names. They are frequently underappreciated until they aren't.
Until Next Week
The AI story is nowhere near over — but it is evolving. The investors who thrive over the next decade will likely be those who look beyond the obvious headlines and into the infrastructure that quietly makes everything possible.
Do your own digging. Ask better questions. And as always — stay curious.
— The Wealth Runway Team
This newsletter is provided for educational and informational purposes only and does not constitute financial, investment, tax, or legal advice. Nothing in this publication should be construed as a recommendation to buy, sell, or hold any security or other financial instrument. Always conduct your own research and consult a licensed, regulated financial advisor before making any investment decision. Past performance is not indicative of future results.
/