The King Is Not Yet Dead - ARM Holdings Q1 FY3/26 Earnings Analysis

DISCLAIMER: This note is intended for US recipients only and, in particular, is not directed at, nor intended to be relied upon by any UK recipients. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Read our full disclaimer, here.

House View

I shall first repeat what I wrote in the AMD earnings note earlier today.

One day, $NVDA will no longer reign supreme. Empires don’t last forever, and they usually collapse under their own weight. Customers may praise NVDA in public but no customer relishes NVDA’s high prices and limited supply, which is what leads to the company’s huge margins and cash pile. This means that the largest customers will continually work with alternative suppliers - AMD, Intel, ARM, Broadcom, others - to try to undermine the more or less monopolistic claim that Nvidia has on the modern datacenter.

My own opinion, guided by nothing other than that most dangerous of things, feelings, is that $ARM will be the New King Of AI once the old king begins to weaken. I think the kingdom will be surrendered, slowly at first, then very quickly (as is the case with all tectonic shifts in technology). Why? Because power.

It is as absurd as it is impractical to expect grid power to grow at the rate required to keep the lights on and the fans humming in all the datacenters now popping up all around the world. It cannot be done. Electricity generation and transmission moves at a pace determined maybe 20-30 years ago at best, whereas each new datacenter needs THIS MUCH POWER AND RIGHT NOW.

The answers to this are:

• Build fewer datacenters (not going to happen, for now at least)
• Use off-grid power (not possible at this scale, and don’t talk to me about modular nuclear reactors until about 2035, OK?)
• Speed up the wayleaves, rezoning, transformer supply, cable supply, construction times and everything else required to get on-grid power to keep up (not going to happen, because this is utility companies we are talking about here)
• Build datacenters that have a fundamentally improved power/performance ratio.

Of the above, only one option is feasible and that is the final point. And the key to unlocking that is (i) core processor architecture and (ii) the software instruction set used to control the core processors. Both of these must be designed to use fundamentally lower power than the NVDA / GPU approach. And there is only one chip designer that has the chops to do this at scale and with a brand trusted by hyperscalers - that is $ARM.

Let’s get into it, and see where $ARM stock may be headed.