BreakingTech retrospective archive — event of August 31, 2026.
AWS has made EC2 R9g and R9gd instances based on Graviton5 processors available. According to Amazon, the new generation delivers up to 25% higher compute performance compared to R8g and is designed primarily for databases, in-memory caches, and real-time analytics.
Graviton has become one of the most successful cases of custom silicon in the cloud. Amazon designs the processors in-house to optimize them for its infrastructure and the most common workloads.
The cloud builds its own chips
For years, cloud providers purchased nearly all their compute capacity from a few external manufacturers. Today, AWS, Google, and Microsoft increasingly design proprietary components, because controlling the chip allows them to optimize costs, power consumption, and performance.
The same dynamic is even more evident in AI, with Trainium and Inferentia. Graviton demonstrates that custom silicon is not only about generative models: it is also becoming normal for traditional enterprise workloads.
The economic advantage
A proprietary processor enables the cloud provider to reduce the marginal cost of infrastructure and differentiate its offering. For the customer, however, it also increases the risk of dependency on specific architectures.
Nonetheless, the spread of ARM in servers demonstrates that the historical monopoly of the x86 architecture is no longer a given. The cloud is transforming the processor market from the inside.
