6 Comments
User's avatar
Stockdrifts Research's avatar

What do you think of GSIT and its compute in memory. I know that the delay is an issue, however do you think it can pull it off?

Jamey Garner's avatar

Yeah I’m interested in this as well

Damnang's avatar

GSIT is technically interesting, but I would separate the technology from execution risk.

Compute in memory is a real long term direction, especially as memory bandwidth and data movement become bigger bottlenecks.

The issue for GSIT is whether they can turn that technical promise into customer adoption, software ecosystem support, and meaningful revenue.

I also wrote about why PIM(CIM) adoption is difficult in my previous article, “Memory Is No Longer a Cyclical Industry.” Reading that section may help make the GSIT execution risk easier to understand.

So I would say it is worth watching, but I would be cautious until there is clearer evidence of design wins or commercial traction.

Jamey Garner's avatar

Thank you!

Stockdrifts Research's avatar

I also loved your take on the edge AI article. I am wondering if you think BZAI's low power architecture is something to look at? They too got the execution risk looming over them and H2 might be revenue heavy. But do you think their approach might be the right approach of the future?

Stockdrifts Research's avatar

Thanks for the response