Beyond the Chips: What Silicon Valley Reveals About the Next Phase of the AI Boom

18 November 2026

Beyond the Chips: What Silicon Valley Reveals About the Next Phase of the AI Boom

Photo: Audun Wickstrand Iversen © DNB

By Audun Wickstrand Iversen, Portfolio Manager at DNB Asset Management

The twenty interviews our investment team recently conducted with Silicon Valley companies point to a new phase in the AI boom. Demand for computing power remains strong. For investors, though, what happens around the chips is becoming increasingly decisive: inference, memory, networking and optical technologies are growing in importance, and artificial intelligence is beginning to move beyond the digital world.

Anyone travelling through Silicon Valley today will see little sign that the wave of AI investment has peaked. Our conversations with these twenty companies, spread across the technology value chain, pointed instead to another trend: for many suppliers, the challenge is not demand but keeping up with it.

Nvidia told us that its growth will remain constrained by component availability and production capacity for the foreseeable future. Applied Materials has already raised its growth forecast for semiconductor manufacturing equipment three times this year, most recently to nearly 40%. Seagate, meanwhile, has effectively sold out its production and is asking customers to commit to volumes several years in advance. For investors, however, another development is at least as interesting: the AI value chain is broadening.

Networking and optics move into focus

In the first phase of the AI boom, much of the investment went into training ever more capable models. The emphasis is now shifting to inference, the actual use of those models. According to Nvidia, the turning point came roughly a year to eighteen months ago, and most customer spending is now directed toward inference.

The economic consequences are significant. AI agents and reasoning models require far more computing power than simple chatbots. Inference also places different demands on infrastructure: beyond graphics processing units (GPUs), fast memory, central processing units (CPUs), storage and high-performance networking are becoming increasingly important.

As a result, the number of companies set to benefit from this investment is growing. Alongside the chipmakers, networking providers such as Arista Networks, optics specialists such as Lumentum and storage companies such as Seagate and NetApp are attracting more attention.

AI becomes a cost item in its own right

AI is also becoming a significant expense for companies. Some of the businesses we spoke to said that tokens, the units used to bill for AI model usage, already account for around 10% of their costs. AI budgets are rising while headcount is growing more slowly.

This is pushing companies to scrutinise the return on their AI infrastructure more closely. Nutanix, for example, has invested around $20 million in its own Nvidia servers so that it can run open-source models in-house, with the aim of covering about 80% of its needs this way. By the company’s calculations, the investment should pay for itself within a year.

AI steps off the screen

Another development may prove even more consequential: AI is starting to leave the data centre and enter the physical world. In San Francisco, driverless cars are now a familiar part of everyday life. Since April 2025, Aurora has been operating driverless trucks on Texas highways. The company aims to have 200 autonomous trucks on the road by the end of the year. Because its trucks don’t need rest breaks, it says they could cover roughly twice as many miles a year as trucks with human drivers.

We are seeing similar developments in industry and healthcare. Lam Research uses AI-assisted robots for maintenance work, and Intuitive Surgical already uses machine learning in robotic surgery.

For investors, this shift changes the game. The key question is no longer just who builds the most powerful AI model. Increasingly, the challenge is to identify where bottlenecks will emerge as AI is deployed at scale in the real world. If our time in Silicon Valley left one clear conclusion, it is this: the range of AI investment opportunities is not shrinking. It is expanding.

Find all our Strategic Case articles

 

Recommandé pour vous