{"id":22013,"date":"2026-03-17T02:45:24","date_gmt":"2026-03-17T09:45:24","guid":{"rendered":"https:\/\/dev.phison.com\/?p=22013"},"modified":"2026-05-05T09:26:25","modified_gmt":"2026-05-05T16:26:25","slug":"phison-rescales-local-ai-inferencing-with-flash-memory-expansion-2","status":"publish","type":"post","link":"https:\/\/www.phison.com\/ja\/phison-rescales-local-ai-inferencing-with-flash-memory-expansion-2\/","title":{"rendered":"Phison Rescales Local AI Inferencing with Flash Memory Expansion"},"content":{"rendered":"

Pascari aiDAPTIV\u2122 technology enables larger-model inference on AI devices with intelligent flash tiering to extend retention and reduce recompute<\/em><\/strong><\/p>\n

San Jose, CA \u2014 GTC 2026 \u2014March 16, 2026<\/strong>\u00a0\u2014\u00a0Phison Electronics<\/a>\u00a0(8299TT), a global leader in NAND flash controllers and storage solutions, today announced its GTC showcase at booth 119, demonstrating how multi-tier memory architecture supports larger models and long-context inference on NVIDIA-powered local AI platforms.<\/p>\n

The industry is facing a growing memory constraint while demand for\u00a0AI-ready platforms<\/a>\u00a0continues to surge. Fine-tuning and inference on proprietary data require massive compute and memory resources, creating investment challenges for organizations. These rising solution costs and workflow bottlenecks are slowing time-to-market for revenue-generating innovation. To address this challenge, Phison introduced\u00a0aiDAPTIV\u2122 technology<\/a>\u00a0for local and edge AI use cases. By leveraging Pascari SSDs as a new AI memory tier, aiDAPTIV technology intelligently extends and manages AI working memory across GPU memory, system RAM and flash.<\/p>\n

Today\u2019s announcement showcases how aiDAPTIV applies these\u00a0multi-tier memory<\/a>\u00a0architecture principles to local AI systems as NVIDIA AI infrastructure advances GPU memory capabilities to support\u00a0inference workloads in data center environments<\/a>. Built on high-endurance flash optimized for sustained paging and context retention, aiDAPTIV supports memory-intensive inference and fine-tuning workloads under fixed hardware configurations. The aiDAPTIV flash-based memory tier enables organizations to support these evolving workloads on local systems while maintaining data privacy and improving long-term infrastructure efficiency.<\/p>\n

\u201cConventional memory management was never designed for AI, and today\u2019s AI infrastructure can no longer rely on generic memory management,\u201d said Michael Wu, President and GM, Phison US. \u201cWith aiDAPTIV technology, we\u2019ve built an AI-aware architecture that extends effective memory across multiple tiers. This enables larger models and long-context inference on local AI platforms without requiring additional GPU hardware, helping organizations keep AI workloads local while planning infrastructure investments more predictably.\u201d<\/p>\n

At NVIDIA GTC 2026, Phison is showcasing demonstrations of partner notebooks, workstations and systems powered by\u00a0NVIDIA GB10 Grace Blackwell processors<\/a>\u00a0as well as\u00a0NVIDIA GeForce RTX 50 Series<\/a>\u00a0and\u00a0NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition GPUs<\/a>.<\/p>\n

Demonstrations highlight long-context inference, agentic AI workflows leveraging KV cache reuse, and memory-intensive fine-tuning on large-scale models, showcasing how aiDAPTIV extends effective memory across GPU memory, system RAM and flash to support workloads that would otherwise exceed available system capacity.<\/p>\n

To request an NVIDIA GTC 2026 meeting with Phison to receive an aiDAPTIV demo, please reach out to\u00a0sales@phison.com<\/a>. To explore Phison\u2019s full portfolio, visit\u00a0www.phison.com<\/a>.<\/p>\n

For more information about the announcement and solutions, visit\u00a0the media kit.<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

Pascari aiDAPTIV\u2122 technology enables larger-model inference on AI devices with intelligent flash tiering to ex […]<\/p>\n","protected":false},"author":9,"featured_media":22016,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[396],"tags":[],"years":[163],"class_list":["post-22013","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-press-releases","years-163"],"_links":{"self":[{"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/posts\/22013","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/users\/9"}],"replies":[{"embeddable":true,"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/comments?post=22013"}],"version-history":[{"count":1,"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/posts\/22013\/revisions"}],"predecessor-version":[{"id":22017,"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/posts\/22013\/revisions\/22017"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/media\/22016"}],"wp:attachment":[{"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/media?parent=22013"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/categories?post=22013"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/tags?post=22013"},{"taxonomy":"years","embeddable":true,"href":"https:\/\/www.phison.com\/ja\/wp-json\/wp\/v2\/years?post=22013"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}