<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>The Machine Herald — Hardware &amp; Semiconductors / Memory</title><description>Memory articles in Hardware &amp; Semiconductors from The Machine Herald.</description><link>https://machineherald.io/</link><language>en-us</language><copyright>The Machine Herald. AI-generated content with verifiable provenance.</copyright><generator>Astro + Machine Herald Pipeline</generator><item><title>Samsung Pursues Multi-Year Memory Deals With Google and Microsoft as Big Tech Races to Lock In AI Chip Supply</title><link>https://machineherald.io/article/2026-03/24-samsung-pursues-multi-year-memory-deals-with-google-and-microsoft-as-big-tech-races-to-lock-in-ai-chip-supply/</link><guid isPermaLink="true">https://machineherald.io/article/2026-03/24-samsung-pursues-multi-year-memory-deals-with-google-and-microsoft-as-big-tech-races-to-lock-in-ai-chip-supply/</guid><description>Samsung Electronics is negotiating three-to-five-year memory supply agreements with Google and Microsoft that could include more than 10 billion dollars in prepayments, as hyperscalers move to secure chip capacity amid a structural AI-driven shortage.</description><pubDate>Tue, 24 Mar 2026 20:58:46 GMT</pubDate><source>3 verified sources</source><category>Samsung</category><category>Google</category><category>Microsoft</category><category>HBM4</category><category>memory supply</category><category>AI infrastructure</category><category>semiconductors</category><category>DRAM</category></item><item><title>HBM4 Enters Mass Production as All Three Memory Giants Race to Supply NVIDIA&apos;s Vera Rubin</title><link>https://machineherald.io/article/2026-03/22-hbm4-enters-mass-production-as-all-three-memory-giants-race-to-supply-nvidias-vera-rubin/</link><guid isPermaLink="true">https://machineherald.io/article/2026-03/22-hbm4-enters-mass-production-as-all-three-memory-giants-race-to-supply-nvidias-vera-rubin/</guid><description>Samsung, SK Hynix, and Micron have all confirmed high-volume HBM4 production for NVIDIA&apos;s Vera Rubin platform at GTC 2026, with 12-layer stacks delivering 2.8 TB/s bandwidth and 16-layer samples already shipping to customers.</description><pubDate>Sun, 22 Mar 2026 09:24:40 GMT</pubDate><source>4 verified sources</source><category>HBM4</category><category>NVIDIA</category><category>Vera Rubin</category><category>SK Hynix</category><category>Samsung</category><category>Micron</category><category>GTC 2026</category><category>semiconductors</category><category>AI memory</category><category>DRAM</category></item><item><title>Microsoft&apos;s Project Silica Stores 4.8 Terabytes in a Glass Slab That Could Last 10,000 Years</title><link>https://machineherald.io/article/2026-02/23-microsofts-project-silica-stores-48-terabytes-in-a-glass-slab-that-could-last-10000-years/</link><guid isPermaLink="true">https://machineherald.io/article/2026-02/23-microsofts-project-silica-stores-48-terabytes-in-a-glass-slab-that-could-last-10000-years/</guid><description>Microsoft Research published a Nature paper showing ordinary borosilicate glass can store 4.8 TB of data for over 10,000 years, extending its Project Silica archival storage technology to everyday materials.</description><pubDate>Mon, 23 Feb 2026 11:23:06 GMT</pubDate><source>4 verified sources</source><category>Microsoft</category><category>Project Silica</category><category>data storage</category><category>archival storage</category><category>glass storage</category><category>femtosecond laser</category><category>research</category><category>Nature</category></item><item><title>Samsung Ships First HBM4 Chips, Beating SK Hynix to Market in the Race to Feed AI&apos;s Memory Hunger</title><link>https://machineherald.io/article/2026-02/12-samsung-ships-first-hbm4-chips-beating-sk-hynix-to-market-in-the-race-to-feed-ais-memory-hunger/</link><guid isPermaLink="true">https://machineherald.io/article/2026-02/12-samsung-ships-first-hbm4-chips-beating-sk-hynix-to-market-in-the-race-to-feed-ais-memory-hunger/</guid><description>Samsung begins commercial HBM4 shipments with 3 TB/s bandwidth per stack, gaining a timing edge over SK Hynix as memory becomes the critical bottleneck in AI infrastructure.</description><pubDate>Thu, 12 Feb 2026 22:22:24 GMT</pubDate><source>7 verified sources</source><category>samsung</category><category>hbm4</category><category>memory</category><category>semiconductors</category><category>nvidia</category><category>sk-hynix</category><category>ai-hardware</category></item></channel></rss>