The Information Machine

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

Version 8

2026-05-25 19:48 UTC · 203 items

What

NVIDIA's Vera Rubin NVL72 platform has moved from product launch into industrial-scale deployment, with Microsoft as the primary customer and European cloud provider Nscale — characterized by the Wall Street Journal as 'NVIDIA-backed' [22] — as the infrastructure intermediary. Dell, Supermicro [14], and HPE [15] have all announced Vera Rubin NVL72 system support, forming a complete OEM ecosystem around the platform. Analysts project NVIDIA's Vera CPU will outperform Intel and AMD x86 chips by 1.5x at Computex 2026 [11]. Two structural bottlenecks — HBM4 memory supply (shortage projected through 2028 [33]) and 600kW per-rack power requirements that exceed most existing data center capacity [34][35] — remain the primary constraints on deployment timelines.

Why it matters

The WSJ's characterization of Nscale as 'NVIDIA-backed' [22] raises a structural question about the supply chain: NVIDIA is simultaneously the chip designer, a reported financial backer of the cloud operator deploying those chips, and the named co-party in the West Virginia gigawatt-scale infrastructure deal — a concentration of interests unprecedented in the AI hardware buildout. If NVIDIA holds equity in Nscale, the deployment commitments and pricing that both companies have co-publicized are no longer arm's-length commercial relationships, which affects how investors and regulators should read the announcements.

Open questions

  • The WSJ describes Nscale as 'NVIDIA-backed' [22] — what is the size and structure of NVIDIA's financial stake in Nscale, and does it constitute a conflict of interest in Nscale's role as Microsoft's infrastructure intermediary for the Portugal and West Virginia deployments?

  • The West Virginia deal is structured as a Letter of Intent [27], not a binding contract — does NVIDIA's reported financial backing of Nscale [22] affect the firmness of Microsoft's 1.35GW commitment or Nscale's capital position to fund the full 8GW Monarch Compute Campus buildout?

  • Micron's official IR page states high-volume HBM4 production for Vera Rubin [36] while multiple industry reports maintain Micron was excluded [37][38] — with Supermicro and HPE now confirmed as OEM partners [14][15], does the expanded manufacturing ecosystem clarify the actual two-supplier versus three-supplier HBM4 allocation?

  • Analysts project Vera CPU will outperform Intel and AMD x86 by 1.5x at Computex 2026 [11] — will the June 1–5 COMPUTEX event produce the first independent third-party benchmarks from production hardware, and if not, when does validation of this claim emerge?

Narrative

NVIDIA's Vera Rubin NVL72 — 336 billion transistors, 50 petaflops of AI performance [1] — was announced as being in full production at Jensen Huang's CES 2026 keynote [2][3] and won the Computex 2026 Best Choice Golden Award [4]. The Vera CPU, an 88-core design delivering 1.2 TB/s of memory bandwidth, was hand-delivered to OpenAI, Anthropic, and other leading AI labs on May 18, 2026 [5][6]. NVIDIA's Q1 2026 financial results — $81.6B revenue, up 85% year-over-year [7][8] — validated Huang's characterization of AI demand as 'parabolic,' and Huang's May 18 keynote at Dell Technologies World projected a $3–4 trillion AI infrastructure buildout by 2030 [9] while explicitly naming memory supply as the primary production bottleneck [10]. Analysts at Computex 2026 project the Vera CPU will outperform Intel and AMD x86 chips by 1.5x [11], with industry forecasters projecting 4 million Vera CPU shipments in FY2027 and approximately two-thirds of the x86 server CPU market [12].

An OEM ecosystem has consolidated around the Vera Rubin platform. Dell positioned its AI Factory as the primary enterprise channel for Vera Rubin NVL72 [13]. Supermicro officially announced support for both the Vera Rubin NVL72 and the smaller HGX Rubin NVL8 variant, with expanded liquid-cooled rack-scale manufacturing capacity [14]. HPE published a dedicated resource on deploying Vera Rubin NVL72 at gigascale [15]. At the hyperscaler and cloud level, Meta [16], Google Cloud [17][18], and Microsoft Azure [19] each formalized Vera Rubin alignment through official channels at GTC Taipei in May 2026, while Nebius committed to Vera Rubin NVL72 deployment from H2 2026 via SEC filing [20][21].

European cloud provider Nscale — described by the Wall Street Journal as 'NVIDIA-backed' [22], a characterization that adds a financial dimension to what was previously described as a commercial relationship — has become the primary infrastructure intermediary for Microsoft's Vera Rubin deployments at scale. Nscale is confirmed as the first European company to deploy NVIDIA Vera Rubin [23] and is delivering 66,000+ Rubin GPUs to Microsoft at Start Campus in Portugal, a 200MW deployment [24][25]. In West Virginia, Nscale acquired the 8GW Monarch Compute Campus in Mason County [26], and Microsoft signed a Letter of Intent for 1.35GW of that capacity in a deal co-announced with NVIDIA and Caterpillar [27][28]. The LOI structure — as distinct from a binding contract — combined with NVIDIA's reported financial backing of Nscale, are both unresolved context factors in assessing the independence and firmness of the infrastructure commitments.

Two structural constraints threaten to pace how quickly the publicized deployment commitments can materialize. The primary documented constraint is memory supply: HBM4 costs have surged 485%, rack prices have reached $8.8 million [29][30], Samsung has sold out its entire 2026 HBM4 supply [31], SK Hynix holds approximately 70% of NVIDIA's HBM4 orders [32], and the shortage is projected to persist until 2028 [33]. The second constraint is physical infrastructure: Vera Rubin's 600kW per-rack power requirement exceeds what most existing data centers can support, making greenfield builds a requirement at deployment scale [34][35]. Against this backdrop, Micron's official investor relations press release states the company is in 'high-volume production of HBM4 designed for NVIDIA Vera Rubin' [36] — directly contradicting industry reports that NVIDIA designated Samsung and SK Hynix as sole HBM4 suppliers with Micron excluded [37][38] — a discrepancy that remains publicly unresolved.

Timeline

  • 2026-01-05: NVIDIA debuts Rubin chip at CES: 336 billion transistors, 50 petaflops AI performance [1]
  • 2026-01: Jensen Huang announces at CES 2026 keynote that Vera Rubin NVL72 is in full production [2][3]
  • 2026-01: HBM4 mass production delayed to end of Q1 2026 due to spec upgrades and Samsung yield challenges [48][49][50]
  • 2026-02: SK Hynix begins HBM4 mass production shipments to NVIDIA; holds approximately 70% of NVIDIA's HBM4 orders [51][32]
  • 2026-03-17: Nscale acquires 8GW Monarch Compute Campus in West Virginia; Microsoft signs LOI for 1.35GW in a deal co-announced with NVIDIA and Caterpillar [52][28][27][26]
  • 2026-05-18: First Vera CPUs hand-delivered to OpenAI, Anthropic, and other leading AI labs [40][5][6]
  • 2026-05-18: Jensen Huang keynotes at Dell Technologies World: projects $3–4T AI infrastructure buildout by 2030, flags memory supply constraint, delivers 'Buy Dell' endorsement [39][10][9][53]
  • 2026-05-21: NVIDIA reports Q1 2026 earnings: $81.6B revenue, up 85% year-over-year [7][8]
  • 2026-05-21: NVIDIA GTC Taipei at COMPUTEX: Vera Rubin NVL72 wins Best Choice Golden Award; Meta, Google Cloud, and Microsoft Azure partnerships formalized [46][4][16][17][19]
  • 2026-05: Nscale confirmed as first European company to deploy Vera Rubin; delivers 66,000+ Rubin GPUs to Microsoft at Start Campus Portugal (200MW) [23][24][25]
  • 2026-05: Supermicro announces support for Vera Rubin NVL72 and HGX Rubin NVL8 with expanded liquid-cooled rack manufacturing; HPE announces Vera Rubin NVL72 gigascale AI support [14][15]
  • 2026-05: Micron officially announces high-volume HBM4 production for Vera Rubin, contradicting industry reports of supplier exclusion [36][37][38]
  • 2026-05: Samsung sells out entire 2026 HBM4 supply; memory costs surge 485%; rack prices reach $8.8M; shortage projected to persist until 2028 [31][29][30][33]
  • 2026-05: Analysts predict Vera CPU will outperform Intel and AMD x86 by 1.5x at Computex 2026; NVIDIA projected to ship 4M Vera CPUs in FY2027 and capture two-thirds of x86 server market [11][12]
  • 2026-05: Multiple technical analyses document Vera Rubin NVL72's 600kW per-rack power requirement as incompatible with existing data center infrastructure, requiring greenfield builds [34][35][44]
  • 2026-05: Nebius formalizes full-stack AI cloud partnership with NVIDIA via SEC filing; plans Vera Rubin NVL72 deployment in US and Europe from H2 2026 [20][21]
  • 2026-06-01: COMPUTEX 2026 main event scheduled (June 1–5) [54]

Perspectives

NVIDIA / Jensen Huang

Maximally bullish: AI demand is 'parabolic,' Vera CPU and Vera Rubin NVL72 represent generational leaps in inference economics, and Q1 2026 earnings ($81.6B, +85% YoY) validate the demand trajectory. Huang acknowledged memory supply chain constraints at Dell Technologies World. Official NVIDIA press releases confirm Meta, Google Cloud, Microsoft, and Nebius partnerships, and NVIDIA is named co-party in the West Virginia AI Factory deal.

Evolution: Consistent bullish framing reinforced by record earnings; the WSJ's 'NVIDIA-backed' characterization of Nscale adds a potential equity stake to the supply chain relationships NVIDIA has publicly described only in commercial terms.

[39][40][7][9][10][16][17][19][28][22]

Microsoft

Anchor customer for Vera Rubin NVL72 across two continents via Nscale: receiving 66,000+ Rubin GPUs at Start Campus Portugal (200MW) and holding a 1.35GW LOI for the West Virginia Monarch Compute Campus, while Azure published a formal datacenter planning post enabling large-scale NVIDIA Rubin deployments.

Evolution: Scale of commitment is precisely documented; the LOI structure and NVIDIA's reported financial backing of Nscale are unresolved context factors for assessing the West Virginia commitment's firmness.

[19][24][25][27][22]

Nscale

Confirmed as the first European company to deploy Vera Rubin and the primary infrastructure intermediary for Microsoft's GPU deployments on two continents; the Wall Street Journal characterizes Nscale as 'NVIDIA-backed,' indicating a financial relationship beyond commercial supply that Nscale has not publicly described in detail.

Evolution: The 'NVIDIA-backed' characterization reframes Nscale from a commercial customer to a financially connected deployment partner — the most significant framing shift in this pass.

[23][27][24][25][26][22]

OEM ecosystem (Dell, Supermicro, HPE)

Dell positioned as NVIDIA's primary enterprise Vera Rubin channel via AI Factory; Supermicro officially announced support for Vera Rubin NVL72 and the smaller HGX Rubin NVL8 variant with expanded liquid-cooled rack-scale manufacturing; HPE published a dedicated gigascale deployment resource — together forming a complete multi-vendor OEM tier around the platform.

Evolution: Supermicro and HPE are new OEM additions, completing the ecosystem alongside Dell and validating Vera Rubin as a broad multi-vendor platform.

[13][14][15]

Hyperscaler and cloud customers (Meta, Google Cloud, Nebius)

Meta confirmed as Vera Rubin customer via official NVIDIA press release; Google Cloud deepened its partnership at GTC 2026 and published a dedicated Vera Rubin NVL72 systems guide; Nebius committed to Vera Rubin NVL72 deployment from H2 2026 via SEC filing.

Evolution: Consistent; commitments formalized through investor-relations-grade channels rather than new structural announcements this pass.

[16][17][18][20][21]

Micron

Officially asserts high-volume HBM4 production specifically designed for NVIDIA Vera Rubin, announced via investor relations press release and promoted heading into GTC 2026 — directly contradicting industry reports of Micron's exclusion from the Vera Rubin supply chain.

Evolution: Consistent; the contradiction with exclusion-narrative reporting remains publicly unresolved.

[36][41][42][37][38]

Memory and supply chain analysts

Converge on HBM4 shortage as the binding structural constraint: SK Hynix holds ~70% of NVIDIA's HBM4 orders, Samsung has sold out its 2026 supply, rack prices have surged 485% to as high as $8.8M, and the shortage is projected to persist until 2028.

Evolution: Consistent; the gigawatt-scale deployment commitments now documented increase the stakes of whether memory supply can support announced timelines.

[32][33][31][29][30][43]

Data center infrastructure analysts and technical commentators

Multiple independent analyses characterize Vera Rubin NVL72's 600kW per-rack power requirement as a fundamental incompatibility with existing data center infrastructure — not a retrofit problem but a greenfield requirement — establishing a second structural bottleneck alongside HBM4 supply.

Evolution: Consistent with prior synthesis; the infrastructure readiness problem is documented across multiple independent sources and reinforced by additional LinkedIn technical commentary.

[34][35][44][45]

Tensions

  • Micron's official IR press release states high-volume HBM4 production for NVIDIA Vera Rubin [36], while TechPowerUp and a dedicated Substack analysis maintain NVIDIA designated Samsung and SK Hynix as sole suppliers with Micron excluded [37][38] — two claims that are mutually incompatible unless they refer to different production generations or allocation tiers. [36][37][38][43]
  • NVIDIA markets Vera Rubin on a 10x cost-per-token reduction [39][46] while real rack economics show a 485% memory price surge with prices reaching $8.8M [29][30] and 600kW power requirements that make most existing data centers physically incompatible [34][35] — with no independent production benchmark yet available to validate the efficiency claim. [39][46][29][30][34][35]
  • The WSJ characterizes Nscale as 'NVIDIA-backed' [22] while NVIDIA's public communications describe Nscale only as a commercial partner — if NVIDIA holds equity in Nscale, the independence and arm's-length nature of the deployment announcements that both companies have co-publicized is open to question. [22][27][28]
  • Analyst projections place NVIDIA on track for 4 million Vera CPU shipments, 1.5x performance over Intel/AMD x86, and two-thirds x86 server market share [12][11], while AMD argues publicly that agentic AI favors EPYC architecture [47] — two irreconcilable assessments of the server CPU market trajectory, with hyperscaler procurement now providing the live test. [12][11][47]
  • The West Virginia Monarch Compute Campus deal is structured as a Letter of Intent [27] rather than a binding contract, while Nscale and NVIDIA have co-publicized it in terms suggesting a firm 1.35GW commitment — the gap between LOI language and the deployment scale being marketed has not been publicly addressed. [27][28][26]

Sources

  1. [1] Nvidia debuts Rubin chip with 336B transistors and 50 petaflops of AI performance - SiliconANGLE — reactive:nvidia-vera-computex-launch
  2. [2] Nvidia CEO confirms Vera Rubin NVL72 is now in production — reactive:nvidia-vera-computex-launch
  3. [3] NVIDIA Vera Rubin AI Platform Hits Full Production CES 2026 ... — reactive:nvidia-vera-computex-launch
  4. [4] NVIDIA Vera Rubin NVL72 wins Computex 2026 awards for AI ... — reactive:nvidia-vera-computex-launch
  5. [5] NVIDIA hand-delivers first 1.2 TB/s Vera CPUs to OpenAI, Anthropic ... — reactive:nvidia-vera-computex-launch
  6. [6] 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
  7. [7] NVIDIA just dropped $81.6B in Q1 revenue up 85% YoY 🤯 — reactive:nvidia-vera-computex-launch (2026-05-21)
  8. [8] "Demand has gone parabolic. The reason is simple: Agentic AI has arrived." — reactive:nvidia-vera-computex-launch (2026-05-21)
  9. [9] Jensen Huang today: Memory demand >> supply chain capacity. “Supply chain needs to be ready.” AI memory supercycle... — reactive:nvidia-vera-computex-launch (2026-05-18)
  10. [10] NVIDIA AI - Jensen Huang Says “Buy Dell” - LinkedIn — reactive:nvidia-vera-computex-launch
  11. [11] NVIDIA's Vera CPUs Set to Outrun Intel and AMD x86 Chips by 1.5x at Computex 2026, Analysts Predict — reactive:nvidia-vera-computex-launch
  12. [12] 'Nvidia is already on track' to deliver 4 million Vera CPUs in FY2027 — reactive:nvidia-vera-computex-launch
  13. [13] Michael Dell, Jensen Huang: Boldest Statements From Dell Technologies World 2026 — reactive:nvidia-vera-computex-launch
  14. [14] Supermicro Announces Support for Upcoming NVIDIA Vera Rubin ... — reactive:nvidia-vera-computex-launch
  15. [15] Pioneering the next era of gigascale AI with NVIDIA Vera Rubin ... — reactive:nvidia-vera-computex-launch
  16. [16] Meta Builds AI Infrastructure With NVIDIA — reactive:nvidia-vera-computex-launch
  17. [17] NVIDIA GTC 2026: Google Cloud Deepens Partnership for AI ... — reactive:nvidia-vera-computex-launch
  18. [18] Google Cloud GTC 2026: Vera Rubin NVL72 Systems Guide — reactive:nvidia-vera-computex-launch
  19. [19] Microsoft's strategic AI datacenter planning enables seamless, large ... — reactive:nvidia-vera-computex-launch
  20. [20] NVIDIA and Nebius Partner to Scale Full-Stack AI Cloud - SEC.gov — reactive:nvidia-vera-computex-launch
  21. [21] Nebius to offer NVIDIA Vera Rubin NVL72 in US and Europe from H2 2026 | Corporate - EQS News — reactive:nvidia-vera-computex-launch
  22. [22] Nvidia-Backed Nscale Plans Huge Data Center Cluster in West ... — reactive:nvidia-vera-computex-launch
  23. [23] NVIDIA Vera Rubin Deployment in Europe Announced — reactive:nvidia-vera-computex-launch
  24. [24] Nscale to Deliver 66,000+ NVIDIA Rubin GPUs to Microsoft at Start Campus' Site in Portugal — reactive:nvidia-vera-computex-launch
  25. [25] 66,000+ NVIDIA Rubin GPUs. 200MW of AI infrastructure. One ... — reactive:nvidia-vera-computex-launch
  26. [26] Nscale acquisition includes plan to build AI facility in Mason County — reactive:nvidia-vera-computex-launch
  27. [27] Nscale acquires 8GW Monarch Compute Campus, Microsoft signs on for 1.35GW of compute - DCD — reactive:nvidia-vera-computex-launch
  28. [28] 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
  29. [29] Nvidia's memory costs soar 485%, latest AI systems now cost $7.8 ... — reactive:nvidia-vera-computex-launch
  30. [30] 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
  31. [31] Samsung sells out of 2026 HBM4 supply as memory resurgence ... — reactive:aws-garman-a100-demand
  32. [32] SK Hynix Secures 70% of Nvidia's HBM4 Orders - Semicon — reactive:nvidia-vera-computex-launch
  33. [33] SK Hynix Surges 15% to New High: HBM Shortage Until 2028, How Much Longer Can AI Memory King Rise? — reactive:nvidia-vera-computex-launch
  34. [34] The Data Center Isn't Ready. NVIDIA's Vera Rubin platform ships in… — reactive:nvidia-vera-computex-launch
  35. [35] NVIDIA Vera Rubin: 600kW Racks by 2027 | Introl Blog — reactive:nvidia-vera-computex-launch
  36. [36] Micron in High-Volume Production of HBM4 Designed for NVIDIA ... — reactive:nvidia-vera-computex-launch
  37. [37] Micron Is Locked Out of HBM4 in NVIDIA's Vera Rubin Systems — reactive:nvidia-vera-computex-launch
  38. [38] NVIDIA to Use SK hynix and Samsung HBM4 for "Vera Rubin" Without Micron | TechPowerUp — reactive:nvidia-vera-computex-launch
  39. [39] NVIDIA CEO Jensen Huang at Dell Technologies World: ‘Demand Is Going Parabolic, Utterly Parabolic’ — NVIDIA Blog (2026-05-18)
  40. [40] Vera Arrives: NVIDIA’s First CPU Built for Agents Lands at Top AI Labs — NVIDIA Blog (2026-05-18)
  41. [41] Micron Singapore - Facebook — reactive:nvidia-vera-computex-launch
  42. [42] Micron's HBM4 Enters High-Volume Production, Boosting Bandwidth | Mark Hirsch posted on the topic | LinkedIn — reactive:nvidia-vera-computex-launch
  43. [43] SK hynix, Samsung, and Micron fighting for NVIDIA supply contracts for new 16-Hi HBM4 orders — reactive:nvidia-vera-computex-launch
  44. [44] Nvidia's Vera Rubin GPU: Redesigning Data Centres for 600kW Racks — reactive:nvidia-vera-computex-launch
  45. [45] 600kW racks in 2027 | Maciek Szadkowski — reactive:nvidia-vera-computex-launch
  46. [46] NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI — NVIDIA Blog (2026-05-21)
  47. [47] Agentic AI Changes the CPU/GPU Equation - AMD — reactive:agentic-compute-cpu-gpu
  48. [48] HBM4 Mass Production Delayed to End of 1Q26 By Spec Upgrades ... — reactive:nvidia-vera-computex-launch
  49. [49] HBM4 Mass Production Delayed According to TrendForce ... — reactive:nvidia-vera-computex-launch
  50. [50] Samsung delays HBM4 rollout to 2026 due to yield challenges, all ... — reactive:aws-garman-a100-demand
  51. [51] SK Hynix set to ship HBM4 for Nvidia's Vera Rubin this month — reactive:nvidia-vera-computex-launch
  52. [52] 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
  53. [53] Jensen Huang Says “Buy Dell” | Dell Tech World 2026 | BuilderBase — reactive:nvidia-vera-computex-launch
  54. [54] NVIDIA GTC Taipei at COMPUTEX 2026 | June 1-5 — reactive:nvidia-vera-computex-launch