The Information Machine

NVIDIA Launches Vera CPU and Vera Rubin NVL72 at COMPUTEX / GTC Taipei

Synthesis history

19 versions, newest first.

  1. Version 19 2026-06-19 02:26 UTC · 300 items

    The new items this pass (30397, 30398, 31276, 31324, 31325) are social media amplifications of the OpenAI CFO / Fall 2026 Vera Rubin training run story already covered via item 30179 in the prior synthesis. Item 31324 i…

  2. Version 18 2026-06-16 18:22 UTC · 295 items

    Three developments are new this pass. First, the SemiAnalysis SOCAMM report is now substantively resolved in content: the claim is that NVIDIA halved SOCAMM memory in Rubin racks due to supply shortages [^28956], with S…

  3. Version 17 2026-06-14 02:29 UTC · 286 items

    The main addition this pass is the AgentPerf benchmark story [^28128][^28325]: NVIDIA published the first agentic AI throughput benchmark on June 12 showing GB300 NVL72 running 20x more agents per megawatt than H200, ad…

  4. Version 16 2026-06-11 08:28 UTC · 280 items

    The main development this pass is primary-source confirmation of the three-vendor HBM4 qualification. Bloomberg reported on June 5 — citing the NVIDIA CEO — that Samsung, SK Hynix, and Micron are all certified to supply…

  5. Version 15 2026-06-09 02:38 UTC · 272 items

    Two substantive new developments this pass. First, analyst Rohan Paul reported on June 8 that NVIDIA qualified all three major HBM4 suppliers — including Micron — for Vera Rubin with all in full production [^26575]; thi…

  6. Version 14 2026-06-07 18:11 UTC · 265 items

    Two substantive new developments this pass. First, NVIDIA open-sourced the Rubin NVSwitch Tray BoM on June 5, and SemiAnalysis reported it contains AMD EPYC 3151 embedded CPUs — nine per VR200 rack [^25105] — revealing …

  7. Version 13 2026-06-05 02:51 UTC · 258 items

    One substantive new angle: Jensen Huang articulated a market thesis that agentic AI shifts CPUs from traffic-cop schedulers to active orchestration layers, framing Vera CPU as a $200B market opportunity [^24686] — this …

  8. Version 12 2026-06-03 18:39 UTC · 250 items

    New items this pass are largely confirmatory rather than materially new. Official press releases from Marvell and NVIDIA [^23460][^23463] add direct sourcing to the NVLink Fusion 'semi-custom AI infrastructure' framing,…

  9. Version 11 2026-06-02 08:31 UTC · 237 items

    Three developments materially update the prior synthesis. First, Jensen Huang announced at COMPUTEX that Microsoft has completed bring-up of its first Rubin VR200 NVL72 rack via Foxconn as ODM [^23025], adding a second …

  10. Version 10 2026-06-01 02:17 UTC · 228 items

    Two developments materially update the prior synthesis. First, Dell delivered the world's first fully validated Vera Rubin NVL72 rack to CoreWeave on May 31, clearing L11 diagnostics (single-rack scale-up domain operati…

  11. Version 9 2026-05-27 02:31 UTC · 221 items

    Three developments materially update the prior synthesis. First, Phoronix benchmark results for the Vera CPU are now available via NVIDIA's blog [^21655], partially resolving the open question about independent performa…

  12. Version 8 2026-05-25 19:48 UTC · 203 items

    The most significant new development is the WSJ's characterization of Nscale as 'NVIDIA-backed' [^20519], which reframes the Nscale-Microsoft-NVIDIA deployment relationship from a commercial supply chain into one where …

  13. Version 7 2026-05-25 12:38 UTC · 185 items

    The West Virginia deal has gained significant precision: Data Center Dynamics reporting [^19856] reveals the campus is named 'Monarch Compute Campus,' has a total capacity of 8GW (not just the 1.35GW Microsoft signed fo…

  14. Version 6 2026-05-25 06:11 UTC · 168 items

    The dominant new development is the Nscale-Microsoft relationship, which has moved from a general European GPU commitment to two named, site-specific deployments: 66,000+ Rubin GPUs for Microsoft at Start Campus in Port…

  15. Version 5 2026-05-24 19:49 UTC · 159 items

    The dominant new developments are the rapid expansion of the Vera Rubin commercial ecosystem — Google Cloud [^17532], Microsoft Azure [^17533], and Nscale (100,000+ GPUs to Europe in 2027) [^17534] join Meta and Nebius …

  16. Version 4 2026-05-24 10:11 UTC · 135 items

    The dominant new developments are the crystallization of HBM4 supply chain specifics — SK Hynix securing ~70% of NVIDIA's orders [^13384], Samsung selling out its entire 2026 HBM4 supply [^12210], a 435% memory price su…

  17. Version 3 2026-05-23 04:17 UTC · 87 items

    The dominant new development is a report that Rubin GPU mass production targets have been lowered [^11387], which materially sharpens the memory supply chain tension that was previously surfaced only through Huang's own…

  18. Version 2 2026-05-22 19:23 UTC · 69 items

    The dominant new development is NVIDIA's Q1 2026 earnings ($81.6B, +85% YoY), which converts the previously narrative-only 'parabolic demand' claim into hard financial evidence. Two new supply-side tensions emerged that…

  19. Version 1 2026-05-22 02:17 UTC · 3 items

    NVIDIA is executing a coordinated hardware offensive centered on two products: the Vera CPU — its first processor purpose-built for agentic AI — which began shipping to leading AI labs on May 18 [^7726], and the Vera Ru…