Jul 16, 2026 · 25 min · 6 segments
Three AWS Summits across the US — **LA on June 10, NYC on June 17, and Washington DC on June 30–July 1** — brought a wave of announcements that all orbited a single thesis: **the age of agentic AI is…
Geries AbouAyashGuest
Malini ChatterjeeHost
Amit NarayananHostWhy does CPU matter so much in this agentic AI era? Everybody's talking about GPUs, right? But Graviton5 seems to suggest that CPU is becoming even more and more important for agentic AI.
You know, We're asking a different question today than we were even a year ago.
For the last few years, we have naturally focused on GPU because that's where model training and inference happens.
But as we move into an agentic world, the challenge shifts from generating intelligence to operationalizing intelligence.
Some are retrieving information, others are calling APIs, some are coordinating workflows, others are interacting with enterprise applications or collaborating with other agents to complete business processes.
And that's exactly why I think Rabbiton 5 is such an important announcement.
We're talking about 192 cores with 33% lower inter-core latency, allowing thousands of concurrent agent sessions to communicate efficiently.
As you know, Malini, our customers can see up to 25% higher compute performance than Graviton4.
35% faster web applications, 35% faster inference, and approximately 30% faster database performance.
What we've really done is optimize the entire platform for modern AI workloads, not just one component.
Why does CPU matter so much in this agentic AI era? Everybody's talking about GPUs, right? But Graviton5 seems to suggest that CPU is becoming even more and more important for agentic AI.
You know, We're asking a different question today than we were even a year ago.
For the last few years, we have naturally focused on GPU because that's where model training and inference happens.
But as we move into an agentic world, the challenge shifts from generating intelligence to operationalizing intelligence.
Some are retrieving information, others are calling APIs, some are coordinating workflows, others are interacting with enterprise applications or collaborating with other agents to complete business processes.
And that's exactly why I think Rabbiton 5 is such an important announcement.
We're talking about 192 cores with 33% lower inter-core latency, allowing thousands of concurrent agent sessions to communicate efficiently.
As you know, Malini, our customers can see up to 25% higher compute performance than Graviton4.
35% faster web applications, 35% faster inference, and approximately 30% faster database performance.
What we've really done is optimize the entire platform for modern AI workloads, not just one component.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.
