We're entering a golden age of Arm-based processors. It used to be that when you decided to save 20+% on your cloud compute bill with Graviton processors, purists could come out of the woodwork and tell you "you're locking yourself in to AWS by compiling your software for Graviton! You're ensuring that you won't be able to move your most critical workloads ever." I actually heard from one of these people today, which provoked me to write this rant.

In 2024, it is absolutely no longer the case that compiling for Graviton locks you in to anything. Leaving aside the fact that Arm-based Ampere chips have been available on many major clouds for a while (including Azure and Google Cloud), with the announcements of Azure's Cobalt chip and Google's Axion chip, the number of options available to run Arm-based workloads in the cloud has exploded.

These new chips are all built using flavors of Arm Neoverse, making it almost certain that your Arm-based workloads will port with no issues, unless you're using very specific processor features (in which case, you knew exactly what you were doing when you used those features in the first place).

At this point it's a no-brainer to use Graviton for your AWS compute workloads, unless you're using some arcane x86 features. You're going to get significant savings across the board, which will likely only get better as competitive pressures mount. You'll also reduce energy use significantly, which seems poised to become even more important in this era of energy-guzzling GPU clusters.

At this point using Graviton for your compute workloads is all upside, and anyone telling you different is living in the past.