If you only follow the headlines, you’d think the AI story begins and ends with Nvidia. And to be fair, Nvidia has earned its place. It built the picks and shovels for the first phase of the AI boom. GPUs became the default engine for training giant models, and Nvidia rode that wave brilliantly. But markets move on. And right now, something important is happening under the surface.
Big Tech is taking control of its own chips
Microsoft just confirmed it’s rolling out its second-generation custom AI chip, Maia 200.
Maia 200 will power Copilot for businesses and support large AI models, including ChatGPT. In other words, Microsoft is designing silicon specifically for how it runs AI, rather than relying entirely on off-the-shelf GPUs.
This is part of a much bigger shift.
Amazon has Trainium and Inferentia. Google has tensor processing units (TPUs). Meta is doing its own designs. Everyone wants tighter control over performance, power efficiency, and cost.
Custom silicon is no longer optional. It’s strategic.
The market is already telling us this
Look at the share prices.
Nvidia hasn’t collapsed – far from it. But it’s gone basically nowhere for six months. No major selling pressure. But no real buying pressure either. The stock is stuck in neutral.
Meanwhile, Taiwan Semiconductors (TSMC) – the company that actually manufactures these custom chips is on fire – up roughly 40% over the same period. New all-time highs.
That divergence matters.
It tells us the AI spend isn’t slowing…it’s shifting.
From general-purpose chips to highly specialised silicon. From “one size fits all” to “built exactly for my workload.”
And this is where the opportunity for investors comes in…
There’s a chip company that sits right in the middle of the custom silicon revolution.
It helps hyperscalers design custom AI accelerators, networking chips, and data-movement silicon that connects everything together inside massive data centres.
And here’s the key point many miss…
Custom AI chips don’t work in isolation.
They need ultra-fast connectivity. They need advanced interconnects. They need specialised silicon to move data efficiently between chips, racks, and servers.
As Microsoft, Amazon, and others roll out more Maia-style chips, the amount of custom networking and infrastructure silicon required explodes.
More custom chips mean more complexity, which leads to more demand for this chip maker’s expertise.
Not a subscriber to Money Morning?
You can get free daily recommendations like these with Money Morning eletter. Just sign up here.
