AWS announces new AMD EPYC-powered cloud instances, purportedly 10% less expensive than other instance types, available today
Paul Alcorn / Tom's Hardware :
Context & Ripple Effects
This 2018 launch slots into AWS's slow build-out of non-Intel options across EC2: the company had already opened the door to FPGA instances for video processing and machine learning work two years earlier, and within a year it would push further with its own Arm-based Graviton2 line claiming 40% better price-performance against comparable x86 instances.
For AMD, the announcement is the marquee hyperscale endorsement its server comeback needed — the same EPYC line that later added Google and Twitter as customers when the second-generation chips arrived with better performance at lower power.
First-order effects
- AWS customers get a drop-in x86 option priced roughly 10% below other instance types, available immediately with no code changes required.
- AMD converts AWS from prospect to flagship cloud customer, giving EPYC a reference deployment to sell against Intel's Xeon install base.
Second-order effects
- Intel faces direct price competition inside its largest cloud channel for the first time, pressuring Xeon pricing and discounting for every hyperscaler renewal.
- Rival clouds watching the playbook — Oracle among them, which later refreshed its compute tier with Arm processors and updated pricing — gain cover to add alternative silicon of their own.
Third-order effects
- If the pattern holds, cloud compute pricing decouples from any single CPU vendor's roadmap: AWS's own Graviton2 followed by third-gen EPYC Milan show providers treating silicon as a swappable procurement lever rather than a fixed dependency.
- Server economics shift toward a two-supplier x86 market plus in-house Arm designs, weakening the premium Intel historically commanded in data center CPUs.
The trend: Hyperscale clouds are diversifying CPU supply — merchant x86 alternatives plus in-house Arm — to convert chip competition into sustained per-instance price cuts.