‘Unified national effort’ looks to create a domestic ASML rival and memory capacity

TurboQuant forces KV cache to shed six-times the weight

The new program automatically promotes or demotes data based on live access patterns to ensure key data runs on the right memory hardware

And warns of upcoming price increases

VMware Cloud Foundation technologist argues customers are sitting on untapped capacity they've already paid for

Solidigm repivoted as Korean giant double dips on AI once more, this time in the U.S.

Hyperscaler leverages a two-tier Ethernet-based topology, custom AI Transport Layer & software tools to deliver a tightly integrated, low-latency platform

HBM4 win for Korean giant as it finds parity with SK Hynix

'Conditional memory' concept moves simple lookups from expensive GPU memory to cheaper CPU RAM

House China committee argues new laws don't go far enough

Researchers propose low-latency topologies and processing-in-network as memory and interconnect bottlenecks threaten inference economic viability

2026 projected to be an SSD dry year, especially with the boon of key value caching

Korean giant also working with SanDisk

Dell dominates both markets, but only one side is throwing a party

Samsung meanwhile looks to be on Nvidia's good list for HBM4

HBM4 memory reportedly pushed back into latter half of 2026

HPE and more expected to hike costs by 15%

Not all growth would be equal, however, with Europe significantly lagging behind Asia and the U.S.

Jensen Huang confident firm can handle AI supply chain issues

AI driving firm's latest company restructure plan

DRAM becomes a hot commodity as inference needs increase

Will start shipping in second quarter of 2024

Joining Micron and Samsung in trying to survive the memory market crash