In the video below the Dell VP of engineering and edge portfolio...

In the video below the Dell VP of engineering and edge portfolio product discusses the edge.

In this video he consistently talks about how Dell is doing a lot of work at the "far edge", which he defines as where data is generated. He says that in a lot of cases the customer solutions will be a combination of edge and cloud, though each need to be able to work independently (say if the cloud is down the system still needs to operate). There are also many cases where 5G or even 4G won't help their customers because it isn't rolled out, such as oil rigs. He goes on to emphasise how edge can refer to harsh locations, places where you may not have much infrastructure or connectivity etc. There was also the interesting use case at 27:30 where if a customer is the top of the game, they don't want to connect to the cloud because they don't want to risk competitors finding anything out about them (this sounded a lot like a DoD type of use case to avoid being hacked).

The most interesting part is from about 25:00 onwards, he talks about using cheap hardware and clustering them together to offer resiliency at the edge. Specifically at 25:45. Here he is talking about who they are working with. He says they are working with Intel because they are their biggest partner. But as they move to the far edge, there is a need for a single chip, a single core, with a lot of capability. They're also working with NVIDIA, SmartNICs....

How many single chips, or single cores do you know that are capable of processing data where there is no connectivity with a lot of capability. He is not referring to Intel or NVIDIA here. A single chip that offers the full package solution? As I'm sure Fact Finder will remind us all, Rob Lincourt from Dell emphasises the importance of neuromorphic chips for edge cases. IMO he says the same thing I highlighted from the speaker above but in different words.

I think the following quote from the Dell podcast is relevant in this case (taken from Fact Finder's transcript):
Rob Lincourt:
The other big part of it is, the one thing you didn’t mention, is the power constraint at all those different points that we talked about you know we have to now start concerning ourselves what can we consume at different steps at that Edge environment and I think everything I have been seeing about neuromorphic is it is very low powered, because I can get a couple of cores, or I can get a chip with a couple more cores, and I can start to stitch them together. That to me is the interesting thing and I also get the ability to reconfigure it so I use it for one set of things and then I can bring and almost I don’t want to say instantly, that’s a precise word,but very quickly I can switch it to now and start running a different neural net. You know I get all those benefits out of it that, and that’s when I look to start seeing the benefits of neuromorphic, is that power in having a very nice linear power as we add more cores.