Windows on Arm – Trump Card

Its battery life first and AI second.

• It turns out that as good as the Copilot+ features are, they need a bit more refining before they are widely adopted meaning that the main draw of an AI PC is its battery life where Arm wins hands down over x86.
• The early reviews of s running the X Elite and X Plus processors are in and by and large the reviews are pretty good for a 1^st generation product.
• The independent tests largely match what Qualcomm claimed at launch, and the vast majority of apps that users access on a PC run well directly on Arm or the emulator which works extremely well.
• The one exception is gaming where many games won’t run when downloaded from Steam, but this makes sense because these are not gaming machines as there is no version yet with a dedicated graphics chip.
• Hence, no one buying a gaming PC will buy one of these products (no one expected them to) meaning that, at the moment, this is not an issue.
• However, the one thing that has not gone entirely to plan is the rollout of Copilot+ which is still not available for the x86 processor and the where most useful feature, Recall, has been pulled at the last minute due to privacy concerns.
• This is not a huge surprise because Recall tracks everything the user does on the PC making it incredibly easy to find things items and events from the past which raised privacy concerns from the minute that it was announced.
• However, I am less concerned as I spend way too much time using the awful Outlook search, patchy Chrome history and my own data files for things I have accessed in the past but can’t find.
• Hence, when this is refined to fix the issues, I think that it will prove to be a popular product with content creators like me.
• Including Recall, there are 4 main AI services offered in Copilot+ which are Cocreator, Windows Studio Effects and Live Captions with Translation but none of them are quite as good or as useful as promised.
• I think that this will change with time as these are 1^st generation products which never really live up to the promises made until their creator has 2 or 3 generations under its belt.
• Voice assistants are a great example of this as we were promised voice assistants that could communicate fluently and naturally with the user back in 2016 and 2017 and what we got was useless, dumb voice bots.
• It is only now, 8 years later, that voice assistants are approaching anything close to what was promised, and this is why I think one of generative AI’s superpowers is the ability to use natural language as the man-machine interface.
• Hence, I suspect it will be a few iterations before Copilot+ comes into its own with the good news being that this can be updated on machines that are being sold now.
• Despite the issues (see here), device sales are reportedly quite good and the main reason that users are buying them seems to be for battery life rather than AI (see here (anecdotal at best)) which is not an enormous surprise.
• I think that this will work to Microsoft and Qualcomm’s advantage as users have PCs that can match Macbooks in terms of performance and battery life which will help seed the market with capable devices.
• This means that when Microsoft has got Copilot+ right, there will be an installed base of users that make use of the product without having to buy a new device.
• Over the last 6 months, there has been a significant shift in the marketing message from Qualcomm which at the launch of the X Elite and X Plus was all about AI but this has since moderated to battery life and AI.
• This is precisely the right approach and I think that we will see the message pivot more towards battery life as Microsoft tidies up Copilot+ and Qualcomm fixes some of the compatibility issues that have cropped up.
• This is the first time that Intel and AMD have had proper competition in this market and unless they can match Arm’s battery life, they are in big trouble.
• Either way, I still think that Arm is poised to take a large slice of this market very little of which is in Qualcomm or Arm’s medium-term earnings estimates.
• The best way to play this remains Qualcomm which has a larger exposure to this market (in terms of absolute profit generation), and it trades on a lower multiple.
• Hence, I continue to prefer Qualcomm as the best way to play this theme and I am also keeping an eye on MediaTek which I suspect is going to address this market sooner or later.