Amazon Web Services: EC2 Graviton Strategy

This is a snippet of the transcript, sign up to read more.

What's the endgame here? If they own the full stack, where they can design the chip, design the server, they can build the micro service on top, the customer workloads, what is the endgame you're going to have? You could have hundreds of different mini Gravitons that serve different customer groups, effectively.

I think that's exactly right. You can eventually build very hyper specific processors based on your observations of your customer patterns. And that's probably where they're headed. I have a tale from Amazon Apocrypha, the lore of the company. Eventually, Amazon started building their own top of rack switches, and building the ASICs inside of these because they were outgrowing the capabilities of what Cisco could provide them. One of the things that was interesting is, back when they were using Cisco gear – Juniper at the time – they would basically say, we have this issue, can you help us replicate it? The problem was, the vendor couldn't replicate the scale that AWS was operating at. They really needed to start building their own stuff that was hyper specialized for what they needed to do, at the scale they needed to operate at and allowed them to debug the full stack in a way that they couldn't previously. It just gives them the, as you mentioned, ability to own the whole stack. It's very interesting that they can be very, very mission focused on a specific workload or a specific customer type, even down to the hardware level, which is really unique.

This is a snippet of the transcript, sign up to read more.