Computex 2026 Day 3 – Wintel

More convincing needed

Intel – A temporary breathing space

• After being repeatedly bashed by its competitors, it was Intel’s chance to set the record straight, but unfortunately, it said nothing to dissuade me from the view that x86 becomes more obsolete with every passing conference.
• There is no doubt in my mind that Intel will still have a large share of PC and data centre CPUs in a few years’ time, but crucially, that share will be meaningfully lower than it is today.
• Intel’s latest release, the Core Ultra Series 3, claims to close the power efficiency gap with 27 hours of streaming video, but these sorts of tasks are not very processor intensive.
• Hence, I suspect that with the CPU under high load, Arm and the M5 in particular will continue to outpace x86, but with its 18A process, Intel has certainly closed the gap somewhat.
• When it comes to software, ecosystem and gaming, Intel still leads, but Apple has already completely fixed this problem, and Qualcomm has made significant progress.
• The data centre is another story where the standard for AI is already clearly Arm-based CPUs, meaning that the legacy software moat that Intel had has evaporated.
• However, demand in the data centre is so strong and capacity so limited that Intel is going to see good demand for its CPUs regardless, and its foundry business is also likely to fill up with those that could not get space at TSMC or Samsung.
• This is going to lift short-term earnings, the proceeds of which need to be re-employed in fixing the fundamental obsolescence of x86 rather than doubling down on it.
• Hence, I remain deeply concerned about Intel’s future as a chipmaker, although the foundry is looking a lot better.
• Foundry is not enough to carry the valuation, and so I remain a seller at current levels.

Microsoft – Bad timing

• Microsoft made the mistake of hosting its developer conference when everyone else was focused on Computex, meaning that its efforts to get back into the AI race got less attention than they would normally have received.
• To make matters worse, it did not have that much to say, which does little to dispel the notion that it is losing the enterprise AI space to Anthropic.
• Microsoft has been struggling with its AI products for some time, but now that it is free of its ties to OpenAI, it has been racing to develop leading products of its own.
• By its own admission, AI is the future of its business, which means that if it ends up being merely a distributor of AI from Anthropic or OpenAI, then its future as an enterprise technology leader will be over.
• This is why it has released another seven models, which aim to push back against Anthropic’s current total domination of the enterprise.
• The scale of this dominance has surprised me at Computex, as all the demos use Claude and none that I have seen use OpenAI, Google, Microsoft, or anyone else.
• This could be existential for Microsoft because if all the value in the enterprise goes towards AI service providers, then Microsoft’s days of high profitability are over.
• This is why the seven new models, which Microsoft claims are as good as Opus 4.6, are so important, and Microsoft badly needs to get users writing code with its models.
• Here, it has an advantage as Claude remains seriously capacity constrained and suffered an outage last night.
• By contrast, Microsoft has added more capacity in the last 18 months than in the previous 10 years, and so as long as its coding tools are comparable with Anthropic’s, then it has a chance as it can meet demand that Anthropic can’t.
• This is a must-win for Microsoft and represents the greatest challenge that Satya Nadella has faced since becoming CEO.
• Mr Nadella faultlessly pivoted Microsoft to the enterprise away from its obsession with owning the platform, and now he has to do it all over again.
• The result is that the stock has derated from mid-30’s 12-month forward PER to 22.7x FY2027 PER, but I can’t see any firm evidence yet that the market has got this one wrong, and so I am happy to stay away for now.