The x86 CPU instruction set architecture (ISA) has been dominant in enterprise IT. Almost all business software has been written for x86 processors and almost all servers, PCs and macs run on x86 chips by Intel and AMD. However, the cracks in this impressive architecture are multiplying and many fear that the dam will break.
The decline of the x86 chip architecture is apparent. Apple is moving all its macs to its own chips based on a different ISA (called ARM) and Amazon Web Services has built its own server CPUs also running on ARM. These new chips run much faster and cooler than the x86 chips and I expect the performance delta to only grow with time.
If x86 gets disrupted it will have a huge impact on all players in the IT industry: hardware manufacturers, chip manufacturers, cloud providers and software companies (ISVs).
Vertical integration and technological vision combined with economies of scale
Jean-Louis Gassée’s latest Monday Note discusses the acquisition of Apple Spinoff NUVIA by Qualcomm. It is quite evident by now that x86 is waning. CISC, multithreading and all resulting issues like back channel attacks, inconsistent performance and high energy use have made the venerable ISA increasingly unattractive. It is therefore no surprise that a lot of players are trying to position themselves for the new future of both end-user devices and the data centre.
Although it is impossible to predict who will emerge victorious (who could have predicted that Apple and Google would dominate the mobile space in 2007) a few things are clear by now. Vertical integration and technological vision are as important as economies of scale. At the moment there are 2 players standing head and shoulders above the rest: Apple and AWS. These two are among the few that have the scale to create an attractive alternative ecosystem. Furthermore, they combine a technological vision with a commitment to vertical integration.
Apple is showing a consistent vision and drive in the way they are evolving Swift, their OS’s and their systems on chip. Steven Sinofsky’s reflections on Apple’s WWDC are thought provoking given his leadership history at Microsoft.
Similarly, AWS keeps moving their platform forward with forethought and consistency. Nitro en over firecracker hypervisors, Graviton chips (among a range of new silicon) and serverless. AWS is leading customers to a continuously more cost-efficient model of running workloads.
AWS’ Graviton and Apple’s M1 chips were not created in a vacuum. These chips are elements of a broader strategy.
Assuming x86 gets disrupted: what will be the impact on Google and Microsoft?
Google services are all modern workloads mostly running on Kubernetes. Migrating to a more efficient ISA makes total sense for them. It makes so much sense that it is surprising they did not come up with Graviton.
On the other hand, Google is a company that has missed many opportunities in the last 10 years. It was surprisingly late with public cloud services. It also wasted a huge window of opportunity with Google G-Suite (now Google Workspace) and allowed Microsoft to catch up with Office 365. I do not know whether Google can master the focus to develop its own chip but it will be more than happy to adopt an ARM chip built by a manufacturer.
Microsoft’s position is more difficult to discern. At the end point, Windows 10 is a legacy business for Microsoft. Nevertheless it ties up nicely with Office 365 and Intune and as such forms a bulwark against more modern alternatives. Moving Windows 10 to ARM must be very attractive to Microsoft in order to protect this legacy business. The problem Microsoft faces is the huge legacy installed base of software running on x86. Microsoft does not have the luxury of having an iOS installed base that can revitalise the mac. The only way for a successful transition of PCs to ARM would be a highly performant emulation layer (something like Rosetta 2).
On the Azure side, matters are even more complex. Azure has enjoyed phenomenal growth in the last years and based on what I see and hear my impression is that it has mainly been Windows workloads from the SME and enterprise segments. Microsoft is not actively pushing customers to move to modern serverless or platform as a service (PaaS) offerings that could benefit from ARM.
Keeping customers on VMs including Windows and SQL server licenses is great. You keep milking a familiar and profitable cow and you avoid difficult discussions with ISVs regarding the migration of x86 workloads to ARM. However, AWS keeps making ARM and Linux increasing attractive. Doing nothing makes you vulnerable. For example, last year AWS introduced Babelfish. This service is a clear and present danger because it is linked to AWS’ disruptive alternative solutions to Microsoft’s very profitable (and expensive) SQL Server business.
I see some very difficult tradeoffs for Microsoft here. This difficulty is apparent: despite all the nice talk, Microsoft has not changed and it values its captive legacy installed base. Also read this blogpost I wrote earlier about these topics.