The Information Machine

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

Version 9

2026-05-27 02:31 UTC · 221 items

What

NVIDIA's Vera Rubin platform has advanced on three fronts simultaneously: Phoronix benchmark testing of the Vera CPU — reported via NVIDIA's blog — confirms a 1.5x overall performance advantage over latest-generation x86 processors [10]; Microsoft's Rubin GPU deployment via Nscale has been revised upward to 130,000 units [14]; and NVIDIA's equity stake in Nscale is now documented as approximately $674M within a $1.1B funding round [12][13], confirming that NVIDIA holds substantial equity in the cloud provider deploying its chips at scale. An expanded OEM tier now includes Compal [23], PEGATRON [21], and Aivres [24] building HGX Rubin NVL8 systems, while NVIDIA's own DGX Rubin NVL8 reference system supports Intel Xeon 6 as an alternative host CPU alongside Vera [11]. HBM4 supply shortage projected to 2028 [29] and 600kW per-rack power requirements [30] remain the binding structural constraints.

Why it matters

The Phoronix benchmark results [10] provide the first third-party quantification of Vera CPU's performance advantage over x86, substantiating claims that had previously rested on analyst projections — though they arrive via NVIDIA's promotional blog rather than as an independent Phoronix publication. The confirmed ~$674M NVIDIA equity stake in Nscale [13] means NVIDIA is simultaneously chip designer, financial backer of the deployment operator handling Microsoft's gigawatt-scale commitments, and co-party to those infrastructure deals — a concentration of interests that is structurally significant for regulators and investors assessing the independence of the AI infrastructure buildout.

Open questions

  • Phoronix results show a 1.5x overall Vera CPU advantage and a 10% geometric-mean improvement over the AMD EPYC 9575F 5.0 GHz [10], but the results appear through NVIDIA's promotional blog rather than as an independent Phoronix report — when does Phoronix publish its own unmediated review, and does the benchmark weighting favor memory-bound workloads where Vera's LPDDR5X architecture excels?

  • NVIDIA's own DGX Rubin NVL8 reference system supports Intel Xeon 6 as a host CPU option alongside Vera [11] — does this reflect deliberate market segmentation between the two CPUs, or does it signal that Vera's software ecosystem readiness still lags x86 for certain production workloads?

  • Microsoft's Rubin GPU deployment via Nscale is now reported as 130,000 units [14], up from 66,000+ previously documented for the Portugal deployment [15] — does the higher figure combine the Portugal and West Virginia LOI commitments, or does it represent a revised estimate of the Portugal deployment alone?

  • Micron's IR press release asserts high-volume HBM4 production for Vera Rubin [32] while TechPowerUp, a dedicated Substack analysis, and a video explainer all confirm NVIDIA designated Samsung and SK Hynix as sole suppliers [33][34][35] — has Micron secured an actual production allocation or is the IR claim aspirational?

Narrative

NVIDIA's Vera Rubin platform encompasses an 88-core Vera CPU with 1.2 TB/s of memory bandwidth and the NVL72 system — 336 billion transistors and 50 petaflops of AI performance [1] — announced in full production at Jensen Huang's CES 2026 keynote [2][3]. The Vera CPU was hand-delivered to OpenAI, Anthropic, and other leading AI labs on May 18, 2026 [4][5], and NVIDIA's Q1 2026 financials — $81.6B revenue, up 85% year-over-year [6][7] — validated Huang's characterization of AI demand as 'parabolic.' Industry forecasters project 4 million Vera CPU shipments in FY2027 and approximately two-thirds of the x86 server CPU market [8], and the Vera Rubin NVL72 won the Computex 2026 Best Choice Golden Award [9].

Phoronix benchmark testing of the Vera CPU, reported through NVIDIA's promotional blog, provides the first third-party quantification of the platform's competitive position: a 1.5x overall performance advantage over the latest-generation 128-core x86 processors, 90% of rated peak memory bandwidth in STREAM TRIAD tests (the highest of any CPU Phoronix has tested), and a 10% geometric-mean improvement over the AMD EPYC 9575F 5.0 GHz processor [10]. The LPDDR5X memory operates at under 30 watts versus over 100 watts for traditional DDR5 systems. Phoronix founder Michael Larabel described Vera as 'the most formidable competition to Intel and AMD x86_64 processors ever realized from non-x86 architecture.' The gap between the 1.5x headline and the 10% geometric mean reflects benchmark weighting — Vera leads decisively on memory-bound workloads and more narrowly on compute-bound tasks. NVIDIA's own DGX Rubin NVL8 reference system simultaneously supports Intel Xeon 6 as an alternative host CPU option [11], implicitly acknowledging continued x86 relevance within the platform.

European cloud provider Nscale operates as the primary infrastructure intermediary for Microsoft's Vera Rubin deployments, and the financial relationship with NVIDIA is now documented: CNBC reported a $1.1B funding round with NVIDIA backing [12], and CaproAsia reported approximately $674M as a direct NVIDIA equity investment in Nscale [13]. The GPU deployment scope for Microsoft has been revised upward to 130,000 Rubin GPUs [14], compared to the 66,000+ previously reported for Start Campus Portugal (200MW) [15][16]. Nscale also holds the 8GW Monarch Compute Campus in West Virginia, against which Microsoft signed a 1.35GW Letter of Intent co-announced with NVIDIA and Caterpillar [17][18]. NVIDIA's confirmed equity stake in Nscale means it is simultaneously chip designer, financial backer of the deployment operator, and co-party to the infrastructure commitments — a concentration of interests without prior precedent in the AI hardware buildout.

A complete OEM tier has formed around both Vera Rubin form factors. For the NVL72, Dell, Supermicro [19], HPE [20], and PEGATRON [21][22] have all announced system support. For the smaller HGX Rubin NVL8, Compal [23], PEGATRON [21], Aivres [24], and Supermicro [19] announced systems at GTC 2026. Two structural constraints persist: HBM4 memory supply, with Samsung having sold out its entire 2026 allocation [25], SK Hynix holding approximately 70% of NVIDIA's orders [26], rack prices reaching $8.8M [27][28], and the shortage projected to last until 2028 [29]; and the NVL72's 600kW per-rack power requirement that makes most existing data center infrastructure physically incompatible, mandating greenfield construction [30][31].

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 due to spec upgrades and Samsung yield challenges [56][57]
  • 2026-02: SK Hynix begins HBM4 mass production shipments to NVIDIA; holds approximately 70% of NVIDIA's HBM4 orders [58][26]
  • 2026-03-17: Nscale acquires 8GW Monarch Compute Campus in West Virginia; Microsoft signs 1.35GW LOI co-announced with NVIDIA and Caterpillar [59][18][17][43]
  • 2026-05-18: First Vera CPUs hand-delivered to OpenAI, Anthropic, and other leading AI labs [60][4][5]
  • 2026-05-18: Jensen Huang keynotes Dell Technologies World: projects $3–4T AI infrastructure buildout by 2030 and flags memory supply as primary bottleneck [36][61][37]
  • 2026-05-21: NVIDIA reports Q1 2026 earnings: $81.6B revenue, up 85% year-over-year [6][7]
  • 2026-05-21: NVIDIA GTC Taipei: Vera Rubin NVL72 wins Computex Best Choice Golden Award; Meta, Google Cloud, and Microsoft formalize partnerships [62][9][38][39][40]
  • 2026-05: Samsung sells out entire 2026 HBM4 supply; memory costs surge 485%; rack prices reach $8.8M; shortage projected to persist until 2028 [25][27][28][29]
  • 2026-05: Multiple analyses document Vera Rubin NVL72's 600kW per-rack power requirement as incompatible with existing data centers, requiring greenfield construction [30][31][52]
  • 2026-05: Micron officially announces high-volume HBM4 production for Vera Rubin, contradicting industry reports of supplier exclusion [32][33][34]
  • 2026-05: NVIDIA equity stake in Nscale confirmed at approximately $674M within a $1.1B funding round [12][13][44]
  • 2026-05: Microsoft's Rubin GPU deployment via Nscale revised upward to 130,000 units, including Start Campus Portugal (200MW) [14][42][15][16]
  • 2026-05: Compal, PEGATRON, Aivres, and Supermicro announce HGX Rubin NVL8 systems at GTC 2026; NVIDIA DGX Rubin NVL8 supports Intel Xeon 6 as alternative host CPU alongside Vera [24][23][21][11][19]
  • 2026-05: Dell, Supermicro, HPE, and PEGATRON confirm Vera Rubin NVL72 system support, completing the enterprise OEM tier [45][46][20][21]
  • 2026-05-26: Phoronix benchmarks of Vera CPU published via NVIDIA blog: 1.5x overall x86 advantage, 90% peak bandwidth utilization, 10% geometric mean improvement over AMD EPYC 9575F [10]
  • 2026-06-01: COMPUTEX 2026 main event scheduled (June 1–5) [63]

Perspectives

NVIDIA / Jensen Huang

Maximally bullish: Q1 2026 earnings ($81.6B, +85% YoY) validate 'parabolic' AI demand; Phoronix benchmark results are promoted as confirming Vera CPU's generational leadership over x86; NVIDIA is co-party to the West Virginia infrastructure deal and confirmed equity holder in Nscale.

Evolution: The confirmed ~$674M equity stake in Nscale adds a financial ownership dimension to relationships NVIDIA had previously described only in commercial terms.

[36][6][37][38][39][40][18][41][10]

Microsoft

Anchor customer for Vera Rubin NVL72 across two continents via Nscale: 130,000 Rubin GPUs at Start Campus Portugal and a 1.35GW LOI for the West Virginia Monarch Compute Campus, co-announced with NVIDIA.

Evolution: Deployment scope revised upward to 130,000 GPUs; the LOI structure and NVIDIA's confirmed equity in Nscale are unresolved context factors for assessing commitment firmness.

[40][15][16][17][14]

Nscale

First European company to deploy Vera Rubin and primary GPU infrastructure intermediary for Microsoft; confirmed recipient of approximately $674M NVIDIA equity investment, giving NVIDIA a direct financial stake in Nscale's role as the deployment operator.

Evolution: The equity confirmation reframes Nscale from commercial customer to financially integrated NVIDIA deployment arm — the most structurally significant relationship clarification in the arc.

[42][17][15][16][43][12][13][44]

OEM ecosystem (Dell, Supermicro, HPE, Compal, PEGATRON, Aivres)

A complete multi-vendor OEM tier now covers both the NVL72 (Dell, Supermicro, HPE, PEGATRON) and the smaller HGX Rubin NVL8 (Compal, PEGATRON, Aivres, Supermicro), validating Vera Rubin as a broad platform rather than a single-vendor solution.

Evolution: Compal, PEGATRON, and Aivres are new additions to the NVL8 tier, and NVIDIA's own DGX Rubin NVL8 reference system supports Intel Xeon 6 as a host CPU option alongside Vera, an implicit hedge within NVIDIA's own product line.

[45][46][20][24][23][19][21][11][22]

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

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

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

[38][39][47][48][49][50]

Micron

Officially asserts high-volume HBM4 production specifically designed for NVIDIA Vera Rubin via an investor relations press release, directly contradicting industry reports and a Dailymotion explainer concluding NVIDIA designated Samsung and SK Hynix as sole suppliers.

Evolution: Consistent; the Dailymotion explainer [35] adds a third source to the exclusion narrative, deepening the contradiction with Micron's IR claims.

[32][51][33][34][35]

Memory and supply chain analysts

Converge on HBM4 shortage as the binding structural constraint: SK Hynix holds ~70% of NVIDIA's 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 scale of deployment commitments now documented increases the stakes of whether memory supply can support announced timelines.

[26][29][25][27][28]

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 binding structural bottleneck alongside HBM4 supply.

Evolution: Consistent; the constraint is reinforced by additional technical commentary and remains unaddressed in NVIDIA's public deployment announcements.

[30][31][52][53]

Tensions

  • Micron's official IR press release states high-volume HBM4 production for NVIDIA Vera Rubin [32], while TechPowerUp, a dedicated Substack analysis, and a Dailymotion explainer all conclude NVIDIA designated only Samsung and SK Hynix as HBM4 suppliers [33][34][35] — two claims that are mutually incompatible unless they refer to different allocation tiers. [32][33][34][35]
  • NVIDIA markets Vera Rubin on a 10x cost-per-token reduction while real rack economics show a 485% memory price surge with prices reaching $8.8M [27][28] and 600kW power requirements that make most existing data centers physically incompatible [30][31] — with no independent third-party production benchmark yet available to validate the efficiency claim at scale. [27][28][30][31]
  • NVIDIA confirmed as holding approximately $674M in Nscale equity [13] while publicly describing Nscale only as a commercial partner — if NVIDIA holds substantial equity in the cloud provider deploying its chips at scale, the deployment announcements and pricing co-publicized by both companies are not arm's-length commercial transactions. [41][12][13][17][18]
  • NVIDIA positions Vera CPU as the purpose-built replacement for x86 in agentic AI workloads [10], while NVIDIA's own DGX Rubin NVL8 reference system supports Intel Xeon 6 as a host CPU option alongside Vera [11] — implicitly acknowledging continued x86 relevance within NVIDIA's own platform architecture. [10][11]
  • The West Virginia Monarch Compute Campus deal is structured as a Letter of Intent [17] rather than a binding contract, while Nscale and NVIDIA have co-publicized it in terms implying a firm 1.35GW commitment — the gap between LOI language and the deployment scale being marketed has not been publicly addressed, and NVIDIA's equity in Nscale adds a further complication to assessing the independence of the parties. [17][18][43][13]
  • NVIDIA and analysts project 4 million Vera CPU shipments in FY2027 and two-thirds of the x86 server CPU market [8], while AMD argues publicly that EPYC architecture favors agentic AI workloads [54] — two irreconcilable market trajectory assessments now being tested by live hyperscaler procurement decisions. [8][55][54]

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 hand-delivers first 1.2 TB/s Vera CPUs to OpenAI, Anthropic ... — reactive:nvidia-vera-computex-launch
  5. [5] 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
  6. [6] NVIDIA just dropped $81.6B in Q1 revenue up 85% YoY 🤯 — reactive:nvidia-vera-computex-launch (2026-05-21)
  7. [7] "Demand has gone parabolic. The reason is simple: Agentic AI has arrived." — reactive:nvidia-vera-computex-launch (2026-05-21)
  8. [8] 'Nvidia is already on track' to deliver 4 million Vera CPUs in FY2027 — reactive:nvidia-vera-computex-launch
  9. [9] NVIDIA Vera Rubin NVL72 wins Computex 2026 awards for AI ... — reactive:nvidia-vera-computex-launch
  10. [10] NVIDIA Vera CPU Is ‘Packing a Heavy-Hitting Punch’ Against Competition — NVIDIA Blog (2026-05-26)
  11. [11] NVIDIA DGX Rubin NVL8 Supports Intel Xeon 6 as Host CPU Option for x86-Based AI Inference - StorageReview.com — reactive:nvidia-vera-computex-launch
  12. [12] Nvidia-backed UK AI firm Nscale raises $1.1 billion funding round — reactive:nvidia-vera-computex-launch
  13. [13] UK AI Infrastructure Startup Nscale Receives $674 Million (£500 ... — reactive:nvidia-vera-computex-launch
  14. [14] 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
  15. [15] Nscale to Deliver 66,000+ NVIDIA Rubin GPUs to Microsoft at Start Campus' Site in Portugal — reactive:nvidia-vera-computex-launch
  16. [16] 66,000+ NVIDIA Rubin GPUs. 200MW of AI infrastructure. One ... — reactive:nvidia-vera-computex-launch
  17. [17] Nscale acquires 8GW Monarch Compute Campus, Microsoft signs on for 1.35GW of compute - DCD — reactive:nvidia-vera-computex-launch
  18. [18] 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
  19. [19] Supermicro Reveals DCBBS® with New NVIDIA Vera Rubin NVL72 ... — reactive:nvidia-vera-computex-launch
  20. [20] Pioneering the next era of gigascale AI with NVIDIA Vera Rubin ... — reactive:nvidia-vera-computex-launch
  21. [21] PEGATRON Unveils Next Generation AI Platforms Powered by NVIDIA Vera Rubin NVL72 and NVIDIA HGX Rubin NVL8 at GTC 2026 | PEGATRON SVR — reactive:nvidia-vera-computex-launch
  22. [22] PEGATRON Unveils Next Generation AI Platforms Powered by ... — reactive:nvidia-vera-computex-launch
  23. [23] Compal Introduces High-Density NVIDIA HGX™ Rubin NVL8 Integrated Solution at GTC 2026 — reactive:nvidia-vera-computex-launch
  24. [24] KR5288 Rubin | 5U NVIDIA HGX™ Rubin NVL8 Server - Aivres — reactive:nvidia-vera-computex-launch
  25. [25] Samsung sells out of 2026 HBM4 supply as memory resurgence ... — reactive:aws-garman-a100-demand
  26. [26] SK Hynix Secures 70% of Nvidia's HBM4 Orders - Semicon — reactive:nvidia-vera-computex-launch
  27. [27] Nvidia's memory costs soar 485%, latest AI systems now cost $7.8 ... — reactive:nvidia-vera-computex-launch
  28. [28] 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
  29. [29] SK Hynix Surges 15% to New High: HBM Shortage Until 2028, How Much Longer Can AI Memory King Rise? — reactive:nvidia-vera-computex-launch
  30. [30] The Data Center Isn't Ready. NVIDIA's Vera Rubin platform ships in… — reactive:nvidia-vera-computex-launch
  31. [31] NVIDIA Vera Rubin: 600kW Racks by 2027 | Introl Blog — reactive:nvidia-vera-computex-launch
  32. [32] Micron in High-Volume Production of HBM4 Designed for NVIDIA ... — reactive:nvidia-vera-computex-launch
  33. [33] Micron Is Locked Out of HBM4 in NVIDIA's Vera Rubin Systems — reactive:nvidia-vera-computex-launch
  34. [34] NVIDIA to Use SK hynix and Samsung HBM4 for "Vera Rubin" Without Micron | TechPowerUp — reactive:nvidia-vera-computex-launch
  35. [35] Why Nvidia Snubbed Micron For Samsung, SK Hynix - Dailymotion — reactive:hbm-memory-supply-squeeze
  36. [36] NVIDIA CEO Jensen Huang at Dell Technologies World: ‘Demand Is Going Parabolic, Utterly Parabolic’ — NVIDIA Blog (2026-05-18)
  37. [37] 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)
  38. [38] Meta Builds AI Infrastructure With NVIDIA — reactive:nvidia-vera-computex-launch
  39. [39] NVIDIA GTC 2026: Google Cloud Deepens Partnership for AI ... — reactive:nvidia-vera-computex-launch
  40. [40] Microsoft's strategic AI datacenter planning enables seamless, large ... — reactive:nvidia-vera-computex-launch
  41. [41] Nvidia-Backed Nscale Plans Huge Data Center Cluster in West ... — reactive:nvidia-vera-computex-launch
  42. [42] NVIDIA Vera Rubin Deployment in Europe Announced — reactive:nvidia-vera-computex-launch
  43. [43] Nscale acquisition includes plan to build AI facility in Mason County — reactive:nvidia-vera-computex-launch
  44. [44] Nscale raises $433m in pre-Series C funding, backed by Nvidia and others | Kai Nicol-Schwarz posted on the topic | LinkedIn — reactive:nvidia-vera-computex-launch
  45. [45] Michael Dell, Jensen Huang: Boldest Statements From Dell Technologies World 2026 — reactive:nvidia-vera-computex-launch
  46. [46] Supermicro Announces Support for Upcoming NVIDIA Vera Rubin ... — reactive:nvidia-vera-computex-launch
  47. [47] Google Cloud GTC 2026: Vera Rubin NVL72 Systems Guide — reactive:nvidia-vera-computex-launch
  48. [48] NVIDIA and Nebius Partner to Scale Full-Stack AI Cloud - SEC.gov — reactive:nvidia-vera-computex-launch
  49. [49] Nebius to offer NVIDIA Vera Rubin NVL72 in US and Europe from H2 2026 | Corporate - EQS News — reactive:nvidia-vera-computex-launch
  50. [50] Meta scales AI data centers with NVIDIA - eeNews Europe — reactive:nvidia-vera-computex-launch
  51. [51] Micron Singapore - Facebook — reactive:nvidia-vera-computex-launch
  52. [52] Nvidia's Vera Rubin GPU: Redesigning Data Centres for 600kW Racks — reactive:nvidia-vera-computex-launch
  53. [53] 600kW racks in 2027 | Maciek Szadkowski — reactive:nvidia-vera-computex-launch
  54. [54] Agentic AI Changes the CPU/GPU Equation - AMD — reactive:agentic-compute-cpu-gpu
  55. [55] 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
  56. [56] HBM4 Mass Production Delayed to End of 1Q26 By Spec Upgrades ... — reactive:nvidia-vera-computex-launch
  57. [57] HBM4 Mass Production Delayed According to TrendForce ... — reactive:nvidia-vera-computex-launch
  58. [58] SK Hynix set to ship HBM4 for Nvidia's Vera Rubin this month — reactive:nvidia-vera-computex-launch
  59. [59] 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
  60. [60] Vera Arrives: NVIDIA’s First CPU Built for Agents Lands at Top AI Labs — NVIDIA Blog (2026-05-18)
  61. [61] NVIDIA AI - Jensen Huang Says “Buy Dell” - LinkedIn — reactive:nvidia-vera-computex-launch
  62. [62] NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI — NVIDIA Blog (2026-05-21)
  63. [63] NVIDIA GTC Taipei at COMPUTEX 2026 | June 1-5 — reactive:nvidia-vera-computex-launch