The Information Machine

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

Version 15

2026-06-09 02:38 UTC · 272 items

What

NVIDIA's Vera Rubin NVL72 cleared its first two rack validation stages through separate ODM chains in late May and early June, with Dell/CoreWeave at L11 on May 31 [3] and Microsoft/Foxconn on June 1 [4]. On June 8, analyst Rohan Paul reported NVIDIA qualified HBM4 from all three major suppliers — SK Hynix (~60–70%), Samsung (~25–30%), and Micron (remainder) — with all three in full production for Vera Rubin [9], directly addressing what had been the most contested open question about the supply chain. Separately, SemiAnalysis published a 'Vera SOCAMM' memory report that drew accusations of inaccuracy; SemiAnalysis defended it by citing physical evidence at the SK Hynix Computex booth [11], though the substantive content of the report remains incompletely described. The previously disclosed AMD EPYC 3151 embedded CPUs in the open-sourced NVSwitch Tray BoM [12] received wide secondary coverage but added no new technical detail.

Why it matters

If the three-vendor HBM4 qualification is confirmed by a primary source, the supply ceiling for Vera Rubin production is meaningfully higher than the two-vendor picture implied, reducing the most-cited constraint on deployment pace. The Vera SOCAMM angle, if the SemiAnalysis report is substantiated, would add a third architectural detail to track alongside the AMD CPU dependency and the 600kW power requirement.

Open questions

  • Rohan Paul's June 8 report claims NVIDIA qualified Micron as a third HBM4 supplier alongside Samsung and SK Hynix [9], but this comes from a single analyst tweet — has this been confirmed by NVIDIA, Micron, or another primary source?

  • SemiAnalysis published a Vera SOCAMM note that drew strong pushback and defended it with SK Hynix Computex booth evidence [11] — what exactly does the SOCAMM report claim, and what are the architectural or supply-chain implications if correct?

  • Jensen said wafer-level production had started but rack-level mass production had not yet begun at COMPUTEX on June 1 [4], then described Vera Rubin as 'in full production' in Seoul on June 5 [6] — when does rack-level mass production actually begin, and does the gap create delivery risk against H2 2026 commitments?

  • Both CoreWeave and Microsoft achieved single-rack L11 bring-up [3][4], but L12 cluster validation with scale-out networking has not been reported — when does the first L12 milestone occur?

Narrative

NVIDIA's Vera Rubin NVL72 is a high-density AI compute rack built around 72 Rubin GPUs interconnected via NVLink at the rack level, requiring 600kW of power per rack and HBM4 memory [1]. Dell delivered the first fully validated VR200 NVL72 rack to CoreWeave on May 31, clearing L11 diagnostics that confirmed the rack's internal NVLink/IMEX scale-up domain operational [2][3]. On June 1 at COMPUTEX, Jensen Huang announced Microsoft completed bring-up of its first Rubin VR200 NVL72 via Foxconn as ODM, and disclosed that wafer-level production had started while rack-level mass production had not yet begun [4]. NVIDIA's three-stage validation hierarchy runs L10 (single-server firmware) through L11 (single-rack scale-up domain) to L12 (full compute cluster with scale-out networking) [5]; neither operator has publicly cleared L12. Five days after COMPUTEX, Jensen described Vera Rubin as 'in full production' during a Seoul supply-chain alignment trip [6] — language that sits in mild tension with the more precise COMPUTEX distinction.

Two supply-chain constraints have framed deployment expectations since launch. HBM4 supply had been understood as dominated by SK Hynix and Samsung, with Samsung selling out its entire 2026 allocation and rack prices at $8.8M [7][8]. On June 8, analyst Rohan Paul reported that NVIDIA qualified all three major HBM4 suppliers — SK Hynix at approximately 60–70% share, Samsung at 25–30%, and Micron with the remainder — with all three in full production for Vera Rubin [9]. If confirmed by a primary source, this expands the effective supply ceiling and partially resolves what had been an open question about whether Micron secured an allocation. The second structural constraint — the 600kW per-rack power requirement — remains unchanged, mandating greenfield data center construction for most deployments [1][10].

A separate SemiAnalysis report on 'Vera SOCAMM' — a memory module format associated with HBM packaging — attracted pushback characterized as accusations of inaccuracy. SemiAnalysis defended the report by citing physical evidence visible at the SK Hynix booth at Computex [11]. The full content of the SOCAMM report has not been described in detail in available sources, leaving the architectural and supply-chain implications unclear. Earlier, NVIDIA open-sourced the Rubin NVSwitch Tray bill of materials, which SemiAnalysis reported includes AMD EPYC 3151 embedded CPUs — nine per VR200 rack [12]. This cross-vendor dependency, absent from NVIDIA's public architecture descriptions, received broad secondary amplification [13][14][15][16][17] without new technical detail.

Deployment commitments continue to accumulate from Nscale (130,000 Rubin GPUs for Microsoft, with NVIDIA holding approximately $674M in Nscale equity) [18][19] and Verda across Europe, the US, and APAC [20]. SemiAnalysis has rated Jensen's COMPUTEX keynote 'F tier,' finding no new AI datacenter products, and documents that Rubin FP4/FP8 gains are approximately 3.5x over GB200 while FP16 gains are only approximately 1.6x and HBM capacity is flat [21][22] — making stated efficiency advantages workload-specific rather than uniform.

Timeline

  • 2026-01-05: NVIDIA debuts Rubin chip at CES: 336 billion transistors, 50 petaflops AI performance. [43]
  • 2026-01: Jensen Huang announces at CES 2026 that Vera Rubin NVL72 is in full production. [44][45]
  • 2026-02: SK Hynix begins HBM4 mass production shipments to NVIDIA, holding approximately 70% of NVIDIA's HBM4 orders. [46][39]
  • 2026-03-17: Nscale acquires 8GW Monarch Compute Campus in West Virginia; Microsoft signs 1.35GW LOI co-announced with NVIDIA and Caterpillar. [47][48][29][49]
  • 2026-05: Samsung sells out entire 2026 HBM4 supply; rack prices reach $8.8M; shortage projected until 2028. [40][8][41][7]
  • 2026-05: NVL72's 600kW per-rack power requirement documented as incompatible with existing data centers, requiring greenfield construction. [1][10][42]
  • 2026-05: NVIDIA equity stake in Nscale confirmed at approximately $674M; Microsoft's Rubin GPU deployment via Nscale revised upward to 130,000 units. [50][19][18]
  • 2026-05-18: First Vera CPUs hand-delivered to OpenAI, Anthropic, and other leading AI labs. [51][52][53]
  • 2026-05-21: NVIDIA reports Q1 2026 earnings: $81.6B revenue, up 85% year-over-year. [25][54]
  • 2026-05-21: GTC Taipei: Vera Rubin NVL72 wins Computex Best Choice Golden Award; Meta, Google Cloud, and Microsoft formalize partnerships; $2B NVLink Fusion investment in Marvell announced. [55][56][57][58][28][26]
  • 2026-05-31: Dell delivers world's first fully validated Vera Rubin NVL72 rack to CoreWeave; L11 diagnostics confirm scale-up domain operational. [2][30][3][5][31][32][33]
  • 2026-06-01: Jensen Huang at COMPUTEX announces Microsoft completed bring-up of first Rubin VR200 NVL72 via Foxconn; wafer-level production started, rack-level mass production not yet begun. [4]
  • 2026-06-01: SemiAnalysis rates Jensen's COMPUTEX keynote 'F tier': no new AI datacenter products; Rubin FP4/FP8 gains documented at ~3.5x over GB200 while FP16 gains are ~1.6x. [22][21]
  • 2026-06-04: Jensen Huang identifies agentic AI as a $200B market, framing Vera CPU as an active orchestration layer rather than a scheduler. [23]
  • 2026-06-05: Jensen Huang visits Seoul, describes Vera Rubin as 'in full production,' and aligns AI supply chain for H2 2026. [6]
  • 2026-06-05: NVIDIA open-sources Rubin NVSwitch Tray BoM; SemiAnalysis reports design includes AMD EPYC 3151 embedded CPU — nine per VR200 rack. [12]
  • 2026-06-08: Analyst Rohan Paul reports NVIDIA qualified HBM4 from all three suppliers — SK Hynix (~60–70%), Samsung (~25–30%), and Micron — with all in full production for Vera Rubin. [9]
  • 2026-06-08: SemiAnalysis defends Vera SOCAMM report against critics calling it fake news, citing evidence at SK Hynix Computex booth. [11]

Perspectives

NVIDIA / Jensen Huang

Q1 2026 earnings ($81.6B, +85% YoY) validate AI demand; Vera Rubin is 'in full production' as of June 5 with supply chain aligned for H2 2026 [6]; NVLink Fusion, NemoClaw, and a $200B agentic AI CPU thesis extend the platform narrative beyond GPU-centric infrastructure [23].

Evolution: Consistent; the Seoul 'full production' phrasing is slightly stronger than the COMPUTEX wafer/rack distinction but the direction is unchanged.

[24][25][4][26][27][23][6]

SemiAnalysis

Rubin FP4/FP8 gains are ~3.5x over GB200 while FP16 is ~1.6x and HBM capacity is flat — efficiency is workload-specific. COMPUTEX keynote rated F tier. NVSwitch Tray BoM discloses AMD EPYC 3151 embedded CPUs (nine per VR200 rack). A Vera SOCAMM report drew fake-news accusations; SemiAnalysis defends it with Computex booth evidence [11].

Evolution: Extended: the Vera SOCAMM defense adds a new contested architectural claim to existing critical coverage.

[21][22][12][11]

Microsoft / Foxconn

First hyperscaler to complete Rubin VR200 NVL72 bring-up via Foxconn as ODM; also anchor customer via Nscale (130,000 GPUs, 1.35GW LOI for West Virginia).

Evolution: Consistent.

[28][29][18][4]

Dell / CoreWeave

Dell delivered the world's first fully validated Vera Rubin NVL72 rack to CoreWeave on May 31, clearing L11 diagnostics with the scale-up domain confirmed operational.

Evolution: Consistent.

[2][30][3][31][32][33]

Micron

Officially asserts high-volume HBM4 production for NVIDIA Vera Rubin. Analyst Rohan Paul now reports Micron is formally qualified alongside Samsung and SK Hynix, with all three in full production [9] — consistent with Micron's own IR claims, though NVIDIA has not confirmed the three-vendor picture directly.

Evolution: The Micron-as-qualified-supplier question appears to be resolving in Micron's favor, but awaits primary-source confirmation.

[34][35][36][37][38][9]

Memory and supply chain analysts

HBM4 shortage had been framed as a binding structural constraint, but Rohan Paul's June 8 three-vendor qualification report [9] suggests the supply ceiling is higher than the two-vendor picture implied; the shortage-until-2028 projection [8] may need revision if confirmed.

Evolution: Softened: multi-vendor qualification, if confirmed, partially addresses the constraint that had been characterized as binding.

[39][7][40][8][41][9]

Data center infrastructure analysts

Vera Rubin NVL72's 600kW per-rack power requirement is a fundamental incompatibility with existing data center infrastructure, establishing greenfield construction as a binding structural bottleneck independent of HBM4 supply.

Evolution: Consistent.

[1][10][42]

Tensions

  • Analyst Rohan Paul reports NVIDIA qualified all three major HBM4 suppliers including Micron for Vera Rubin [9], while prior independent analyses concluded NVIDIA designated only Samsung and SK Hynix [36][37][38] — two views of the supplier set that cannot both be current. [9][36][37][38]
  • SemiAnalysis published a Vera SOCAMM note that critics called fake news; SemiAnalysis defends it by citing physical evidence at SK Hynix's Computex booth [11] — the substantive content of the report is disputed and incompletely described in available sources. [11]
  • Jensen stated at COMPUTEX on June 1 that wafer-level production had started but rack-level mass production had not yet begun [4], then described Vera Rubin as 'in full production' five days later in Seoul [6] — two statements using different production language that leave actual rack delivery timelines ambiguous. [4][6]
  • NVIDIA markets Vera Rubin on a 10x cost-per-token reduction, but SemiAnalysis documents FP4/FP8 gains of ~3.5x over GB200 while FP16 gains are ~1.6x and HBM capacity is flat [21] — the efficiency claim is workload-specific, not uniform. [21]
  • NVIDIA's promotional materials present the NVL72 as a NVIDIA-integrated system, but the open-sourced NVSwitch Tray BoM discloses AMD EPYC 3151 embedded CPUs — nine per VR200 rack [12] — a cross-vendor dependency absent from NVIDIA's public architecture descriptions. [12]
  • NVIDIA holds approximately $674M in Nscale equity [19] while describing Nscale publicly as a commercial partner, making large deployment announcements co-publicized by both companies non-arm's-length transactions [29]. [19][29]

Sources

  1. [1] The Data Center Isn't Ready. NVIDIA's Vera Rubin platform ships in… — reactive:nvidia-vera-computex-launch
  2. [2] BREAKING NEWS: COREWEAVE & DELL IS THE FIRST CLOUD TO ANNOUNCE THAT THEY HAVE RUBIN VR200 NVL72 WITH FULLY PASSING L… — SemiAnalysis Twitter (2026-05-31)
  3. [3] Notably, passing L11 diags means that this rack is up and running, including the IMEX channels on the NVL72 scale-up dom… — SemiAnalysis Twitter (2026-05-31)
  4. [4] BREAKING NEWS: JENSEN JUST ANNOUNCED MICROSOFT HAS FINISHED BRING UP ON THEIR FIRST RUBIN VR200 NVL72 RACK with their OD… — SemiAnalysis Twitter (2026-06-01)
  5. [5] At L10 your Firmware/BIOS and OS works on a single server, at L11 a single rack or scale-up domain works, and then at L1… — SemiAnalysis Twitter (2026-05-31)
  6. [6] Seoul Purpose: How NVIDIA and South Korea Are Building the Future of AI — NVIDIA Blog (2026-06-05)
  7. [7] SK Hynix Surges 15% to New High: HBM Shortage Until 2028, How Much Longer Can AI Memory King Rise? — reactive:nvidia-vera-computex-launch
  8. [8] Nvidia's memory costs soar 485%, latest AI systems now cost $7.8 ... — reactive:nvidia-vera-computex-launch
  9. [9] Nvidia just cleared the memory bottleneck significantly for Vera Rubin by qualifying HBM4 from Samsung, SK Hynix, and Mi… — Rohan Paul Twitter (2026-06-08)
  10. [10] NVIDIA Vera Rubin: 600kW Racks by 2027 | Introl Blog — reactive:nvidia-vera-computex-launch
  11. [11] Our Vera SOCAMM note is causing a bit of a stir. As always some folks are jumping to the wrong conclusions. Those saying… — SemiAnalysis Twitter (2026-06-08)
  12. [12] BREAKING NEWS: NVIDIA HAS JUST OPEN SOURCED THEIR RUBIN NVSWITCH TRAY BoM & DIAGRAM & IT INCLUDES AMD EYPC 3151 … — SemiAnalysis Twitter (2026-06-05)
  13. [13] NVIDIA Open Sources BlueField NVSwitch Tray Bill of Materials — reactive:nvidia-vera-computex-launch
  14. [14] NVIDIA Open-Sources Rubin NVSwitch Tray Bill of Materials with ... — reactive:nvidia-vera-computex-launch
  15. [15] BREAKING NEWS‼️‼️ NVIDIA RUBIN USES AMD CPU ... — reactive:nvidia-vera-computex-launch
  16. [16] SemiAnalysis' Post - LinkedIn — reactive:nvidia-vera-computex-launch
  17. [17] NVIDIA HAS JUST OPEN SOURCED THEIR RUBIN NVSWITCH ... — reactive:nvidia-vera-computex-launch
  18. [18] 130,000 Rubin GPUs Are Being Deployed at Nscale For Microsoft, Further Showing Massive Interest In NVIDIA's Next-Gen AI Chips — reactive:nvidia-vera-computex-launch
  19. [19] UK AI Infrastructure Startup Nscale Receives $674 Million (£500 ... — reactive:nvidia-vera-computex-launch
  20. [20] Verda to deploy NVIDIA VR200 NVL72 and R200 in H2 2026 across Europe, the US, and APAC — Blog — Verda (formerly DataCrunch) — reactive:nvidia-vera-computex-launch
  21. [21] for more details on Nvidia's VR NVL72 Oberon and future roadmap, check out our article from February: — SemiAnalysis Twitter (2026-05-31)
  22. [22] F TIER KEYNOTEMAX: Jensen ComputeX presentation was one of the worst keynotes he has done. He announced nothing new on t… — SemiAnalysis Twitter (2026-06-01)
  23. [23] Jensen Huang just identified the next $200 billion market (Save this). — Milk Road AI Twitter (2026-06-04)
  24. [24] NVIDIA CEO Jensen Huang at Dell Technologies World: ‘Demand Is Going Parabolic, Utterly Parabolic’ — NVIDIA Blog (2026-05-18)
  25. [25] NVIDIA just dropped $81.6B in Q1 revenue up 85% YoY 🤯 — reactive:nvidia-vera-computex-launch (2026-05-21)
  26. [26] The CEO of NVIDIA, looked at Matt Murphy and said "The next trillion dollar company, ladies and gentlemen." (Save this). — Milk Road AI Twitter (2026-06-02)
  27. [27] Industrial Software Leaders Build Secure, Autonomous AI Engineers With NVIDIA NemoClaw — NVIDIA Blog (2026-06-02)
  28. [28] Microsoft's strategic AI datacenter planning enables seamless, large ... — reactive:nvidia-vera-computex-launch
  29. [29] Nscale acquires 8GW Monarch Compute Campus, Microsoft signs on for 1.35GW of compute - DCD — reactive:nvidia-vera-computex-launch
  30. [30] Dell just made history this weekend and it is the culmination of an execution streak that no other company in enterprise… — Milk Road AI Twitter (2026-05-31)
  31. [31] CoreWeave Completes Industry-First Bring-Up And Validation Of NVIDIA Vera Rubin NVL72 — reactive:nvidia-vera-computex-launch
  32. [32] HPCwire - Since 1987 – Covering the Fastest Computers in the World and the People Who Run Them — reactive:nvidia-vera-computex-launch
  33. [33] CoreWeave Completes Industry-First Bring-Up and Validation of NVIDIA Vera Rubin NVL72 - Las Vegas Sun News — reactive:nvidia-vera-computex-launch
  34. [34] Micron in High-Volume Production of HBM4 Designed for NVIDIA ... — reactive:nvidia-vera-computex-launch
  35. [35] Micron Singapore - Facebook — reactive:nvidia-vera-computex-launch
  36. [36] Micron Is Locked Out of HBM4 in NVIDIA's Vera Rubin Systems — reactive:nvidia-vera-computex-launch
  37. [37] NVIDIA to Use SK hynix and Samsung HBM4 for "Vera Rubin" Without Micron | TechPowerUp — reactive:nvidia-vera-computex-launch
  38. [38] Why Nvidia Snubbed Micron For Samsung, SK Hynix - Dailymotion — reactive:hbm-memory-supply-squeeze
  39. [39] SK Hynix Secures 70% of Nvidia's HBM4 Orders - Semicon — reactive:nvidia-vera-computex-launch
  40. [40] Samsung sells out of 2026 HBM4 supply as memory resurgence ... — reactive:aws-garman-a100-demand
  41. [41] Price of Nvidia's Vera Rubin NVL72 racks skyrockets to as much as $8.8 million apiece, but server makers' margins will be tight — Nvidia is moving closer to shipping entire full-scale systems — reactive:nvidia-vera-computex-launch
  42. [42] Nvidia's Vera Rubin GPU: Redesigning Data Centres for 600kW Racks — reactive:nvidia-vera-computex-launch
  43. [43] Nvidia debuts Rubin chip with 336B transistors and 50 petaflops of AI performance - SiliconANGLE — reactive:nvidia-vera-computex-launch
  44. [44] Nvidia CEO confirms Vera Rubin NVL72 is now in production — reactive:nvidia-vera-computex-launch
  45. [45] NVIDIA Vera Rubin AI Platform Hits Full Production CES 2026 ... — reactive:nvidia-vera-computex-launch
  46. [46] SK Hynix set to ship HBM4 for Nvidia's Vera Rubin this month — reactive:nvidia-vera-computex-launch
  47. [47] Nscale and Microsoft Announce Collaboration with NVIDIA and Caterpillar to Deliver 1.35GW of NVIDIA Vera Rubin NVL72 GPUs at Flagship AI Factory Campus in West Virginia — reactive:nvidia-vera-computex-launch
  48. [48] Nscale and Microsoft Announce Collaboration with NVIDIA and Caterpillar to Deliver 1.35GW of NVIDIA Vera Rubin NVL72 GPUs at Flagship AI Factory Campus in West Virginia — reactive:nvidia-vera-computex-launch
  49. [49] Nscale acquisition includes plan to build AI facility in Mason County — reactive:nvidia-vera-computex-launch
  50. [50] Nvidia-backed UK AI firm Nscale raises $1.1 billion funding round — reactive:nvidia-vera-computex-launch
  51. [51] Vera Arrives: NVIDIA’s First CPU Built for Agents Lands at Top AI Labs — NVIDIA Blog (2026-05-18)
  52. [52] NVIDIA hand-delivers first 1.2 TB/s Vera CPUs to OpenAI, Anthropic ... — reactive:nvidia-vera-computex-launch
  53. [53] Nvidia unveils details of new 88-core Vera CPUs positioned to compete with AMD and Intel – new Vera CPU rack features 256 liquid-cooled chips that deliver up to a 6X gain in CPU throughput | Tom's Hardware — reactive:nvidia-vera-computex-launch
  54. [54] "Demand has gone parabolic. The reason is simple: Agentic AI has arrived." — reactive:nvidia-vera-computex-launch (2026-05-21)
  55. [55] NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI — NVIDIA Blog (2026-05-21)
  56. [56] NVIDIA Vera Rubin NVL72 wins Computex 2026 awards for AI ... — reactive:nvidia-vera-computex-launch
  57. [57] Meta Builds AI Infrastructure With NVIDIA — reactive:nvidia-vera-computex-launch
  58. [58] NVIDIA GTC 2026: Google Cloud Deepens Partnership for AI ... — reactive:nvidia-vera-computex-launch