The Information Machine

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

Version 14

2026-06-07 18:11 UTC · 265 items

What

NVIDIA's Vera Rubin NVL72 cleared two independent rack validation milestones in late May — Dell/CoreWeave (L11, May 31) [1][2] and Microsoft/Foxconn (VR200 bring-up, June 1) [3] — while Jensen Huang traveled to Seoul on June 5 to align the supply chain for H2 2026 and described Vera Rubin as 'in full production' [5]. On the same day, NVIDIA open-sourced the Rubin NVSwitch Tray bill of materials, which SemiAnalysis disclosed contains an AMD EPYC 3151 embedded CPU — nine per VR200 rack — a cross-vendor dependency absent from NVIDIA's public architecture descriptions [6]. Two binding structural constraints remain: HBM4 shortage projected until 2028 [10] and the NVL72's 600kW per-rack power requirement mandating greenfield construction [11]. Jensen had separately stated at COMPUTEX that wafer-level mass production had started while rack-level mass production had not yet begun [3], leaving the actual large-scale delivery timeline unclear.

Why it matters

The AMD CPU disclosure in NVIDIA's own rack fabric is the most architecturally notable detail in the current period: NVIDIA's flagship AI infrastructure product depends on AMD silicon at a structural level, contradicting the image of a fully NVIDIA-integrated system. Whether rack-level mass production begins in time to fulfill H2 2026 deployment commitments from Microsoft, CoreWeave, Verda, and others is the practical question the production language ambiguity leaves open.

Open questions

  • Jensen said at COMPUTEX on June 1 that wafer-level mass production had started but rack-level had not yet begun [3], then described Vera Rubin as 'in full production' in Seoul on June 5 [5] — 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 [2][3], but full-cluster L12 validation with scale-out networking has not been reported by either [4] — when does the first L12 milestone occur?

  • The open-sourced NVSwitch Tray BoM shows AMD EPYC 3151 embedded CPUs — nine per VR200 rack [6] — raising the question of whether AMD silicon appears in other Rubin rack subsystems and what this implies for NVIDIA's supply chain dependencies.

  • Micron's IR materials assert high-volume HBM4 production for Vera Rubin [19], while multiple independent analyses conclude NVIDIA designated only Samsung and SK Hynix as HBM4 suppliers [20][21][22] — has Micron secured an actual production allocation?

Narrative

NVIDIA's Vera Rubin NVL72 completed the first two stages of its rack validation hierarchy through separate ODM chains in rapid succession. On May 31, Dell delivered the world's first fully validated VR200 NVL72 rack to CoreWeave, with L11 diagnostics confirming the rack's internal NVLink/IMEX scale-up domain operational [1][2]. The following day at COMPUTEX 2026, Jensen Huang announced Microsoft completed bring-up of its first Rubin VR200 NVL72 rack via Foxconn as ODM [3]. NVIDIA's three-stage validation hierarchy runs from L10 (single-server firmware) through L11 (single-rack scale-up domain) to L12 (full compute cluster with scale-out networking) [4]; neither operator has publicly cleared L12. Jensen disclosed at COMPUTEX that wafer-level mass production had started while rack-level mass production had not yet begun [3]. Five days later in Seoul, he described Vera Rubin as 'in full production' while aligning NVIDIA's supply chain for H2 2026 [5] — language that appears to reference wafer-level status but is less precise than the COMPUTEX distinction.

A disclosure on June 5 added an unexpected detail to the Rubin rack architecture. NVIDIA open-sourced the NVSwitch Tray bill of materials and circuit diagram, and SemiAnalysis reported the design includes an AMD EPYC 3151 embedded CPU [6]. Each VR200 rack contains nine NVSwitch Trays, placing nine AMD CPUs in every rack. The disclosure shows that NVIDIA's flagship AI infrastructure product depends on AMD silicon at a structural level — a cross-vendor dependency absent from NVIDIA's public architecture descriptions. Jensen also separately articulated a market thesis for the Vera CPU: in the agentic AI era, CPUs shift from traffic-cop schedulers to active orchestration layers, which Jensen framed as a $200B market opportunity independent of GPU sales [7].

Two hardware constraints set binding ceilings on deployment pace. HBM4 supply is dominated by SK Hynix (approximately 70% of NVIDIA's orders) and Samsung, whose entire 2026 allocation has sold out, with rack prices at $8.8M and shortage projected until 2028 [8][9][10]. The NVL72's 600kW per-rack power requirement is incompatible with most existing data center infrastructure, requiring greenfield construction [11][12]. Deployment commitments continue to accumulate — including 130,000 Rubin GPUs via Nscale for Microsoft [13] and H2 2026 VR200 NVL72 plans from Verda (formerly DataCrunch) across Europe, the US, and APAC [14] — but the wafer-to-rack production gap means committed volumes cannot yet ship at scale. NVIDIA holds approximately $674M in equity in Nscale [15], making co-announced deployment figures between the two companies non-arm's-length transactions [16].

Critical assessments temper the deployment narrative. SemiAnalysis rated Jensen's COMPUTEX keynote 'F tier,' finding no new AI datacenter products and arguing that the Windows-on-NVIDIA-ARM transition is structurally unlike Apple's x86-to-M1 switch [17]. SemiAnalysis also documents that Rubin FP4/FP8 FLOPs scale approximately 3.5x over GB200 while FP16 gains are only ~1.6x and HBM capacity is flat [18] — making the stated efficiency advantage workload-specific rather than uniform. The Micron HBM4 question remains open: Micron's IR press release asserts high-volume HBM4 production for Vera Rubin [19], while multiple independent sources conclude NVIDIA designated only Samsung and SK Hynix as suppliers [20][21][22].

Timeline

  • 2026-01-05: NVIDIA debuts Rubin chip at CES: 336 billion transistors, 50 petaflops AI performance. [45]
  • 2026-01: Jensen Huang announces at CES 2026 that Vera Rubin NVL72 is in full production. [46][47]
  • 2026-02: SK Hynix begins HBM4 mass production shipments to NVIDIA, holding approximately 70% of NVIDIA's HBM4 orders. [48][8]
  • 2026-03-17: Nscale acquires 8GW Monarch Compute Campus in West Virginia; Microsoft signs 1.35GW LOI co-announced with NVIDIA and Caterpillar. [49][50][16][51]
  • 2026-05: Samsung sells out entire 2026 HBM4 supply; rack prices reach $8.8M; shortage projected until 2028. [40][10][41][9]
  • 2026-05: NVL72's 600kW per-rack power requirement documented as incompatible with existing data centers, requiring greenfield construction. [11][12][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. [44][15][13]
  • 2026-05-18: First Vera CPUs hand-delivered to OpenAI, Anthropic, and other leading AI labs. [52][53][54]
  • 2026-05-21: NVIDIA reports Q1 2026 earnings: $81.6B revenue, up 85% year-over-year. [24][55]
  • 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. [56][57][58][59][28][25]
  • 2026-05-31: Dell delivers world's first fully validated Vera Rubin NVL72 rack to CoreWeave; L11 diagnostics confirm scale-up domain operational. [1][29][2][4][30][31][32]
  • 2026-06-01: Jensen Huang at COMPUTEX announces Microsoft completed bring-up of its first Rubin VR200 NVL72 rack via Foxconn; wafer-level production started, rack-level mass production not yet begun. [3]
  • 2026-06-01: SemiAnalysis rates Jensen's COMPUTEX keynote 'F tier': no new AI datacenter products; Windows-on-NVIDIA-ARM transition framed as unlikely to replicate Apple's success. [17]
  • 2026-06-02: NVIDIA NemoClaw industrial AI agent blueprint announced with adoptions from Cadence, Siemens, Dassault Systèmes, and Synopsys. [27]
  • 2026-06-04: Jensen Huang identifies agentic AI as a $200B market, arguing CPUs shift to active orchestration layers and Vera CPU gains an independent demand driver. [7]
  • 2026-06-05: Jensen Huang visits Seoul, describes Vera Rubin as 'in full production,' and aligns AI supply chain for H2 2026. [5]
  • 2026-06-05: NVIDIA open-sources Rubin NVSwitch Tray BoM; SemiAnalysis reports design includes AMD EPYC 3151 embedded CPU — nine per VR200 rack. [6]

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 [5]; NVLink Fusion, NemoClaw, and a $200B agentic AI CPU thesis extend the platform narrative beyond GPU-centric infrastructure [7].

Evolution: The Seoul trip adds supply-chain coordination language and the 'in full production' phrasing is slightly stronger than the COMPUTEX wafer/rack distinction, but the direction is consistent.

[23][24][3][25][26][27][7][5]

SemiAnalysis

Rubin FP4/FP8 gains are ~3.5x over GB200 while FP16 is only ~1.6x and HBM capacity is flat — efficiency is workload-specific. COMPUTEX keynote rated F tier. Open-sourced NVSwitch Tray BoM discloses AMD EPYC 3151 embedded CPUs (nine per VR200 rack) as an unexpected cross-vendor dependency [6].

Evolution: Extended: the AMD CPU disclosure adds a new hardware dependency finding to SemiAnalysis's existing critical coverage.

[18][17][6]

Microsoft / Foxconn

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

Evolution: Consistent.

[28][16][13][3]

Dell / CoreWeave

First-mover operators: 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.

[1][29][2][30][31][32][33]

Marvell / NVLink Fusion partners

Official Marvell press release frames NVLink Fusion as enabling customers to build 'semi-custom AI infrastructure' by integrating Marvell custom silicon with NVIDIA's NVLink interconnect fabric.

Evolution: Consistent.

[34][35][36][26][37][38]

Micron

Officially asserts high-volume HBM4 production specifically designed for NVIDIA Vera Rubin, directly contradicting industry reports that NVIDIA designated only Samsung and SK Hynix as HBM4 suppliers.

Evolution: Consistent; the contradiction with independent sources remains unresolved.

[19][39][20][21][22]

Memory and supply chain analysts

HBM4 shortage is the binding structural constraint: SK Hynix holds ~70% of NVIDIA's orders, Samsung has sold out its 2026 supply, rack prices are $8.8M, and shortage is projected until 2028.

Evolution: Consistent.

[8][9][40][10][41][3]

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 second binding structural bottleneck alongside HBM4 supply.

Evolution: Consistent.

[11][12][42]

Tensions

  • Micron's official IR press release states high-volume HBM4 production for NVIDIA Vera Rubin [19], while multiple independent analyses conclude NVIDIA designated only Samsung and SK Hynix as HBM4 suppliers [20][21][22] — two claims mutually incompatible unless they refer to different allocation tiers. [19][20][21][22]
  • 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 only ~1.6x and HBM capacity is flat [18] — the efficiency claim is workload-specific, not uniform. [18][10][11]
  • Official NVLink Fusion press materials frame the partnership as enabling 'semi-custom AI infrastructure' for third-party silicon [26], but the structural terms require those chips to adopt NVIDIA's NVLink interconnect — leaving unresolved whether this represents genuine openness or an ecosystem control mechanism. [34][26][25]
  • NVIDIA holds approximately $674M in Nscale equity [15] while describing Nscale publicly as a commercial partner, making large deployment announcements co-publicized by both companies non-arm's-length transactions [16]. [43][44][15][16]
  • Jensen stated at COMPUTEX on June 1 that wafer-level production had started but rack-level mass production had not yet begun [3], then described Vera Rubin as 'in full production' five days later in Seoul [5] — the two descriptions use different production language, leaving actual rack delivery timelines ambiguous. [3][5]
  • NVIDIA's promotional architecture descriptions present the NVL72 as a NVIDIA-integrated system, but the open-sourced NVSwitch Tray BoM, reported by SemiAnalysis, discloses AMD EPYC 3151 embedded CPUs in the switch fabric — nine per VR200 rack [6] — a cross-vendor dependency absent from NVIDIA's public materials. [6]

Sources

  1. [1] 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)
  2. [2] 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)
  3. [3] 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)
  4. [4] 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)
  5. [5] Seoul Purpose: How NVIDIA and South Korea Are Building the Future of AI — NVIDIA Blog (2026-06-05)
  6. [6] BREAKING NEWS: NVIDIA HAS JUST OPEN SOURCED THEIR RUBIN NVSWITCH TRAY BoM & DIAGRAM & IT INCLUDES AMD EYPC 3151 … — SemiAnalysis Twitter (2026-06-05)
  7. [7] Jensen Huang just identified the next $200 billion market (Save this). — Milk Road AI Twitter (2026-06-04)
  8. [8] SK Hynix Secures 70% of Nvidia's HBM4 Orders - Semicon — reactive:nvidia-vera-computex-launch
  9. [9] SK Hynix Surges 15% to New High: HBM Shortage Until 2028, How Much Longer Can AI Memory King Rise? — reactive:nvidia-vera-computex-launch
  10. [10] Nvidia's memory costs soar 485%, latest AI systems now cost $7.8 ... — reactive:nvidia-vera-computex-launch
  11. [11] The Data Center Isn't Ready. NVIDIA's Vera Rubin platform ships in… — reactive:nvidia-vera-computex-launch
  12. [12] NVIDIA Vera Rubin: 600kW Racks by 2027 | Introl Blog — reactive:nvidia-vera-computex-launch
  13. [13] 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
  14. [14] 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
  15. [15] UK AI Infrastructure Startup Nscale Receives $674 Million (£500 ... — reactive:nvidia-vera-computex-launch
  16. [16] Nscale acquires 8GW Monarch Compute Campus, Microsoft signs on for 1.35GW of compute - DCD — reactive:nvidia-vera-computex-launch
  17. [17] 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)
  18. [18] for more details on Nvidia's VR NVL72 Oberon and future roadmap, check out our article from February: — SemiAnalysis Twitter (2026-05-31)
  19. [19] Micron in High-Volume Production of HBM4 Designed for NVIDIA ... — reactive:nvidia-vera-computex-launch
  20. [20] Micron Is Locked Out of HBM4 in NVIDIA's Vera Rubin Systems — reactive:nvidia-vera-computex-launch
  21. [21] NVIDIA to Use SK hynix and Samsung HBM4 for "Vera Rubin" Without Micron | TechPowerUp — reactive:nvidia-vera-computex-launch
  22. [22] Why Nvidia Snubbed Micron For Samsung, SK Hynix - Dailymotion — reactive:hbm-memory-supply-squeeze
  23. [23] NVIDIA CEO Jensen Huang at Dell Technologies World: ‘Demand Is Going Parabolic, Utterly Parabolic’ — NVIDIA Blog (2026-05-18)
  24. [24] NVIDIA just dropped $81.6B in Q1 revenue up 85% YoY 🤯 — reactive:nvidia-vera-computex-launch (2026-05-21)
  25. [25] 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)
  26. [26] NVIDIA Unveils NVLink Fusion for Industry to Build Semi-Custom AI Infrastructure With NVIDIA Partner Ecosystem | NVIDIA Newsroom — reactive:nvidia-vera-computex-launch
  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] 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)
  30. [30] CoreWeave Completes Industry-First Bring-Up And Validation Of NVIDIA Vera Rubin NVL72 — reactive:nvidia-vera-computex-launch
  31. [31] HPCwire - Since 1987 – Covering the Fastest Computers in the World and the People Who Run Them — reactive:nvidia-vera-computex-launch
  32. [32] CoreWeave Completes Industry-First Bring-Up and Validation of NVIDIA Vera Rubin NVL72 - Las Vegas Sun News — reactive:nvidia-vera-computex-launch
  33. [33] CoreWeave completes validation of Nvidia Vera Rubin NVL72 — reactive:nvidia-vera-computex-launch
  34. [34] Marvell and NVIDIA to Provide Custom Solutions for Advanced AI Infrastructure — reactive:nvidia-vera-computex-launch
  35. [35] Marvell and NVIDIA partner on NVLink Fusion | Marvell Technology posted on the topic | LinkedIn — reactive:nvidia-vera-computex-launch
  36. [36] Marvell today announced it is teaming with NVIDIA to offer NVLink ... — reactive:nvidia-vera-computex-launch
  37. [37] NVIDIA Unveils NVLink Fusion for Industry to Build Semi-Custom AI ... — reactive:nvidia-vera-computex-launch
  38. [38] NVIDIA invests $ 2 billion in Marvell Technology in silicon photonics ... — reactive:nvidia-vera-computex-launch
  39. [39] Micron Singapore - Facebook — 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-Backed Nscale Plans Huge Data Center Cluster in West ... — reactive:nvidia-vera-computex-launch
  44. [44] Nvidia-backed UK AI firm Nscale raises $1.1 billion funding round — reactive:nvidia-vera-computex-launch
  45. [45] Nvidia debuts Rubin chip with 336B transistors and 50 petaflops of AI performance - SiliconANGLE — reactive:nvidia-vera-computex-launch
  46. [46] Nvidia CEO confirms Vera Rubin NVL72 is now in production — reactive:nvidia-vera-computex-launch
  47. [47] NVIDIA Vera Rubin AI Platform Hits Full Production CES 2026 ... — reactive:nvidia-vera-computex-launch
  48. [48] SK Hynix set to ship HBM4 for Nvidia's Vera Rubin this month — reactive:nvidia-vera-computex-launch
  49. [49] 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
  50. [50] 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
  51. [51] Nscale acquisition includes plan to build AI facility in Mason County — reactive:nvidia-vera-computex-launch
  52. [52] Vera Arrives: NVIDIA’s First CPU Built for Agents Lands at Top AI Labs — NVIDIA Blog (2026-05-18)
  53. [53] NVIDIA hand-delivers first 1.2 TB/s Vera CPUs to OpenAI, Anthropic ... — reactive:nvidia-vera-computex-launch
  54. [54] 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
  55. [55] "Demand has gone parabolic. The reason is simple: Agentic AI has arrived." — reactive:nvidia-vera-computex-launch (2026-05-21)
  56. [56] NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI — NVIDIA Blog (2026-05-21)
  57. [57] NVIDIA Vera Rubin NVL72 wins Computex 2026 awards for AI ... — reactive:nvidia-vera-computex-launch
  58. [58] Meta Builds AI Infrastructure With NVIDIA — reactive:nvidia-vera-computex-launch
  59. [59] NVIDIA GTC 2026: Google Cloud Deepens Partnership for AI ... — reactive:nvidia-vera-computex-launch