AWS Graviton5 is here and it's their biggest chip upgrade yet
The cloud giant is doubling down on ARM with 192 cores per chip and a focus on efficiency over raw speed.
Amazon Web Services unveiled Graviton5, its fifth-generation ARM-based processor, and the numbers are impressive: 192 cores in a single socket, five times more L3 cache than its predecessor, and up to 25% better performance than Graviton4.
But this isn't just about faster chips. It's about AWS trying to solve a fundamental problem in cloud computing—how to deliver more performance while actually using less energy and keeping costs down. For the third year running, more than half of AWS's new CPU capacity is powered by Graviton, with 98% of its top 1,000 EC2 customers already benefiting from the architecture's price-performance advantages.
That adoption rate tells you everything. Companies like Adobe, Airbnb, Atlassian, and SAP aren't just testing Graviton anymore—they're running critical workloads on it.
The technical shift matters, too. Unlike Graviton4, which used two 96-core processors linked together, Graviton5 consolidates everything into a single 192-core socket. The previous NUMA architecture introduced latency issues when cores needed to access memory across processors. By eliminating that bottleneck, AWS claims to have cut inter-core latency by roughly a third.
Emmanuel Oyedeji
Early testing backs up the hype. Companies report seeing 30% higher performance for databases, 35% faster web applications, and up to 60% improvement in specific analytical queries. Siemens saw 20% performance gains with 30% lower compute costs. Synopsys reported 35% faster runtimes for semiconductor design tools.
These aren't marginal improvements. They're the kind of numbers that make CFOs take notice.
AWS is rolling out Graviton5 in stages. The new M9g instances for general-purpose workloads are available in preview now. C9g instances for compute-intensive tasks and R9g instances for memory-heavy workloads are scheduled for 2026. All three use the same chip but adjust memory ratios depending on the use case—a strategy AWS has refined over multiple Graviton generations.
The chip itself is manufactured on TSMC's 3nm process, which helps explain the energy efficiency gains. Combined with AWS's Nitro 6 smartNICs that offload networking and storage tasks, the platform delivers up to 15% higher network bandwidth and 20% faster Amazon Elastic Block Store performance.
What makes Graviton interesting isn't just the specs—it's the philosophy behind it. AWS isn't chasing the highest clock speeds or the most specialized features. Instead, it's building versatile chips that can handle any workload, from high-performance computing in the morning to gaming servers at night. That flexibility means AWS can deploy these processors across its entire infrastructure without leaving expensive hardware idle.
It also puts pressure on Intel and AMD. With 192 cores, Graviton5 now matches AMD's highest core counts and exceeds Intel's 144-core offerings. And unlike those chips, which need to serve diverse customers across multiple markets, AWS can optimize Graviton specifically for cloud workloads.
The bet on ARM is paying off. Since launching Graviton in 2018, AWS has quietly become one of the most important drivers of ARM adoption in data centers. Now every major cloud provider offers ARM instances, and the architecture is even powering Windows laptops. But AWS remains the pioneer—and with Graviton5, it's maintaining that lead.
For customers, the value proposition is straightforward: run the same workloads faster while paying less and reducing energy consumption. In an industry where every percentage point of efficiency matters, that's hard to ignore.
Emmanuel Oyedeji
