AI Blueprints
Agentic AI Suite
AirgapAI Products
Iternal AI Blueprint Builder
Enterprise AI initiative decision framework.
Evaluate AI opportunities across value, feasibility, cost, governance, risk, adoption, and implementation readiness.
Regulated & Public Sector
Commercial & Operations
AirgapAI Chat
Local, air-gapped LLM chat for classified environments.
Runs entirely on your own hardware for CUI, HIPAA, and other regulated workloads — no data leaves your network.
Explore AirgapAIGuides
AirgapAI Comparisons
Best-Of Guides
Benchmarks & Indexes
AI Strategy Blueprint
Free Chapter 1 (32 pages, PDF) — the framework behind 4 #1 Amazon best sellers.
Get your free chapterTraining by Role
Training by Industry
Learn & Research
On edu.iternal.ai
Iternal AI Academy
AI fluency for every role in your company.
912+ courses · 10-minute lessons · role-based paths.
Company
Partnerships
Contact
Work with Iternal
See the platform in a live demo.
Product walkthroughs for AirgapAI, Blockify, the AI Blueprint Builder, and the AI Academy.
Book a demo
AI Blueprints
AI Blueprint Builder AI Blueprint Books AI Strategy Guide (Hub)Strategy & Advisory
AI Strategy Consulting Fractional Chief AI Officer AI Executive Education AI Team Training AI AcademyAI Consulting Services
AI Consulting AI Implementation Services Machine Learning Consulting AI Data & Analytics Consulting Digital Transformation Consulting AI Agent Development Services AI Governance Consulting Generative AI ConsultingAgentic AI Suite
Turnkey SI Ultramemory Agent Skills Library AgentAuth Ultracache Iternal Workforce Waypoint RFP Blockify PRISM Nebulous IdeaFORGEAirgapAI Products
AirgapAI Chat AirgapAI Transcribe AirgapAI Code All PricingRegulated & Public Sector
Federal Contractors (CUI/CMMC) Defense & Aerospace State & Local (SLED) Healthcare & HIPAA Legal & Law FirmsCommercial & Operations
Financial Services Manufacturing Energy & Utilities Edge Deployment All IndustriesBy Role
CISOs & Security IT Leaders Compliance Officers Enterprise ArchitectsBooks & Partners
AI Partner Blueprint AI Strategy Blueprint Book AI Strategy Framework Free Partner Chapter AI Partner ProgramTools
AI ROI Calculators AI Readiness Assessment Browse All Assessments LLM Pricing CalculatorGuides
Guides by Job AI Cost & Budget Guide LLM Selection & Benchmarks Best Local AI Tools (Enterprise) On-Prem Hardware Sizing How to Deploy LLMs On-Premise Local LLM Deployment Guide Private LLM Guide AI Agent Security Checklist Agentic AI AI Agent Orchestration What Is an MCP Server? Token Usage & Cost Projections Reducing AI Token Costs How to Write AI PromptsLibrary
Case Studies Webinars Book a Demo Events What is Iternal? What is Blockify? Release NotesAirgapAI Comparisons
vs Microsoft Copilot vs ChatGPT Enterprise vs Google Gemini vs Amazon Q vs Azure OpenAIBlockify Comparisons
All Blockify Comparisons Best Vector Databases Best RAG Frameworks Best Data Ingestion Tools Best AI GovernanceBest-Of Guides
Best Copilot Alternatives Best ChatGPT Alternatives Best Local AI Tools Best AI Books Best AI Consulting Firms Best AI Strategy Frameworks Top AI Companies in Austin Best Secure AI Transcription Best Private AI Coding Assistants Best Private AI Appliances Best Secure AI TranslationBenchmarks & Indexes
LLM Benchmark Repository All ComparisonsGet Started
Overview & Pricing Plans & Pricing AI-Fluency Quiz (2 min) AI Training for Teams Enterprise AI Training Guide Student Free AccessTraining by Role
For Marketing Teams For Sales Teams For Executives For Finance Teams For HR Teams For Legal Teams For OperationsTraining by Industry
For Manufacturing For Government For HealthcareLearn & Research
Best AI Training Courses AI Training for Employees How to Write AI Prompts AI Skills Gap AI Training ROIOn edu.iternal.ai
Browse All Courses Verify CertificateCompany
About Iternal Leadership CareersPartnerships
Partners AI Partner BlueprintContact
Contact Iternal Book a DemoFollow
LinkedIn TwitterDatacenter GPU sourcing
Iternal sources NVIDIA HGX B300, GB300 NVL72 and Vera Rubin systems directly from Dell Technologies, HPE and Lenovo, and runs each purchase end to end through executive-level relationships at the OEMs. You contract with and pay the OEM. Iternal’s fee is one disclosed line on the OEM’s quote, owed only if an order is placed.
Interview · Dell, NVIDIA and the datacenter GPU market · runtime 64:44
Three OEM routes · two silicon partners
Iternal sources through three OEMs: Dell Technologies is the first route for an order, and HPE and Lenovo are the other OEM routes. NVIDIA and Intel are silicon partners whose teams Iternal brings in alongside the OEM on platform selection and supply.
Lead OEM path
A Leading ISV Partner of Dell
Iternal’s OEM partnership with Dell has run close to seven years. Iternal brings in Dell’s technical design, facility and deployment specialists as the order needs them.
From Dell’s AI Factory with NVIDIA lineup
Source: Dell Technologies, Mar 2026
OEM route
HPE is the second OEM route. Its US online store lists the NVIDIA GB300 NVL72 by HPE as request-a-quote, with no posted price, and HPE Financial Services offers financing in certain countries, subject to credit approval (HPE, Nov 2026).
OEM route
Lenovo is the third OEM route. Lenovo states it is shipping NVIDIA Vera Rubin NVL72 globally and offers financing and consumption models through Lenovo Global Financial Services (Lenovo Press, Oct 2026).
Silicon partner
NVIDIA Inception member
NVIDIA sells its datacenter platforms through select NVIDIA Partner Network partners and points buyers to OEM partner servers, DGX systems and cloud providers (NVIDIA, Nov 2026). For a multi-node cluster, that makes an OEM order the practical path, and Iternal brings NVIDIA’s team into it on platform selection and supply.
Silicon partner
Intel’s #1 AI ISV Partner & Prestige Alliance Member
Iternal engages Intel on the Xeon host processors in a cluster and on Intel accelerators where a workload calls for them.
Six steps · an OEM team at each
Iternal runs a white-glove purchase and coordinates each step with the OEM’s own specialists, who design, build, ship and install the systems.
Step 01
OEM teamFacility and cooling specialists
Iternal starts with the platform (HGX B300, GB300 NVL72, Vera Rubin NVL72 or HGX Rubin NVL8), the node count, networking, storage, power and cooling, and brings in the OEM’s facility and cooling specialists to check the site.
Cooling differs by generation. NVIDIA’s DGX B300, an eight-GPU Blackwell Ultra system, is air-cooled and rated at 15 kW maximum (NVIDIA DGX B300 User Guide, Nov 2026). Lenovo’s GB300 NVL72 relies primarily on warm-water liquid cooling at 18 to 45 °C (Lenovo Press, Mar 2026). NVIDIA designs Vera Rubin NVL72 racks for 45 °C liquid-cooling inlet operation and advises facilities teams to validate that design early (NVIDIA Technical Blog, Aug 2026). Validate the site for the systems you plan to run next as well as the ones on order.
For model and workload sizing, read the hardware sizing guide.
Step 02
OEM teamTechnical design team
Iternal works with the OEM’s technical design team to reach one buildable configuration and a full-stack order: compute, networking, storage and OEM services together. With the whole stack on one order, the OEM schedules one complete build.
Step 03
OEM teamOEM executives and order management
OEMs confirm delivery dates once an order is in their system. From that point, Iternal escalates with OEM executives. A final configuration, finished compliance paperwork and a ready site shorten the path, because the OEM can schedule the order without waiting on open configuration, paperwork or site questions.
Step 04
OEM teamSales leadership and the OEM’s financing arm
Iternal negotiates OEM pricing at the executive level. Payment runs on the OEM’s terms after its credit review, instead of a large prepayment months before delivery. Dell’s standard US commercial terms, for example, set payment at 30 days after invoice and keep quotes valid for 30 days (Dell Technologies, US Commercial Terms of Sale, Sep 2026).
OEM financing is subject to credit approval: Dell Financial Services on a Dell order, HPE Financial Services in certain countries (HPE, Nov 2026) and Lenovo Global Financial Services (Lenovo Press, Oct 2026). Iternal’s fee is one disclosed, success-based line on the OEM’s quote; if no order is placed, nothing is owed.
Step 05
OEM teamTrade-compliance team
Iternal assembles the OEM’s know-your-customer questionnaire and export-control package with you and coordinates it with the OEM’s trade-compliance team, so the paperwork runs alongside configuration instead of after it. Export-control obligations stay with the buyer.
Step 06
OEM teamLogistics and deployment services
Iternal coordinates delivery timing with the OEM and connects you with the OEM’s deployment services for installation, rack integration and burn-in. Under Dell’s standard US terms, title passes when Dell hands the hardware to the carrier (Dell Technologies, US Commercial Terms of Sale, Sep 2026).
Contract, fee, payment terms, supply and warranty
A traditional reseller, such as an AI server reseller or value-added reseller (VAR), buys systems from the OEM or a distributor and resells them, with its margin inside the price. A GPU broker locates units or capacity from third parties, often on the secondary market. Iternal is neither, because the OEM sells to you directly and Iternal runs the purchase.
Iternal never takes title to the hardware, so no resale margin sits inside the OEM’s price.
Iternal’s fee is its own line on the OEM’s quote, where you can compare it against any other option.
If no order is placed, nothing is owed to Iternal.
Iternal negotiates the quote with OEM executives on your behalf.
You pay the OEM on its terms after credit review, or finance through the OEM, instead of prepaying an intermediary.
Warranty and support are contracted directly between you and the OEM.
| Topic | Iternal, OEM-direct | Traditional reseller or VAR | GPU broker or secondary market |
|---|---|---|---|
| Who you contract with and pay | The OEM: Dell Technologies, HPE or Lenovo | The reseller, which buys from the OEM or a distributor | The broker or the seller it represents; varies by seller |
| How the intermediary is paid | One disclosed, success-based fee line on the OEM’s quote; no hardware markup | A margin inside the resale price, not itemized on the quote | A commission or spread; disclosure varies by seller |
| Pricing | The OEM’s quote, negotiated with OEM executives, with the fee on its own line | Deal pricing the reseller obtains from the OEM, plus its margin | Market price for the units or capacity on hand |
| Payment terms and financing | OEM terms after OEM credit review; OEM financing subject to credit approval | Reseller credit terms and credit lines, which suit buyers that prefer not to go through OEM credit review | Varies by seller; get deposit, refund and title terms in writing |
| Supply and allocation | The order goes into the OEM’s own system, and Iternal escalates with OEM executives | Through the reseller’s OEM relationship and registered deal | Limited to the inventory or capacity on hand |
| Warranty and support | OEM warranty and support, contracted directly with the OEM | OEM warranty passes through an authorized reseller, which may add its own services | Varies by seller; the buyer confirms support entitlement with the OEM |
| Compliance and chain of custody | KYC and export paperwork completed with the OEM’s trade-compliance team; the OEM is the seller of record | Paperwork routed through the reseller to the OEM | Varies by seller; the buyer verifies chain of custody and export compliance |
| Facility, bundle and deployment expertise | Iternal brings in the OEM’s facility, design and deployment specialists | The reseller’s own integration and services teams, where it offers them | Varies by seller |
Resellers bring real strengths: credit lines, multi-brand carts and their own integration services. OEM-direct sourcing fits large orders that need OEM pricing, OEM payment terms and one accountable team at the OEM.
2026 supply and pricing data
Datacenter GPU orders in 2026 compete for constrained supply, at prices that move with memory costs, during the platform transition from GB300 to Vera Rubin.
record $95 billion
AI server backlog Dell reported for its fiscal second quarter of 2027
about 270%
Rise TrendForce expects in server DRAM contract prices during 2026
roughly twice
Average selling price TrendForce expects for Vera Rubin systems, relative to GB300
over $30 billion
NVIDIA sovereign AI business in fiscal 2026, more than triple the year before
NVIDIA Q4 FY2026 earnings call, via The Motley Fool, Feb 2026
On Dell’s May 2026 earnings call, its chief operating officer said demand “continues to exceed supply with memory as the primary constraint.” Dell is prioritizing large strategic customers, and its largest customers are locking in supply through three- to five-year deals (Supply Chain Dive, Jun 2026).
OEMs confirm delivery dates once an order is in their system. A buyer still settling the configuration or the paperwork has no order for the OEM to schedule.
Dell’s standard US commercial terms keep a quote valid for 30 days, subject to change for shortages, cost increases, currency moves or tariffs (Dell Technologies, US Commercial Terms of Sale, Sep 2026). In early 2026 Dell’s chief operating officer said Dell’s quotes “are valid for the shortest period of time they’ve ever been” (CIO Dive, Mar 2026).
TrendForce expects server DRAM contract prices to rise about 270% in 2026 (TrendForce, Aug 2026), so a quote that lapses during a long internal review may return at a different price.
NVIDIA said in May 2026 that Vera Rubin production shipments would begin in fall 2026 (NVIDIA, May 2026). TrendForce expects GB300 rack systems to lead NVIDIA’s 2026 shipments with demand through the first half of 2027, and Vera Rubin racks to take the largest share in 2027 (TrendForce, Aug 2026).
Buyers planning a 2027 deployment can ask Iternal to have the OEM quote both generations, then decide on delivery date, facility and budget.
NVIDIA said its sovereign AI business more than tripled year over year, to over $30 billion, in fiscal 2026 (NVIDIA Q4 FY2026 earnings call, via The Motley Fool, Feb 2026). The CNAS sovereign AI index tracks 184 projects across 67 countries and the EU, and infrastructure made up 80% of new projects in the first half of 2026 (CNAS, Aug 2026).
In July 2026 the EuroHPC Joint Undertaking opened a call for tenders for up to seven AI Gigafactories, with more than €20 billion in private investment expected (EuroHPC Joint Undertaking, Jul 2026).
Three steps sit in the buyer’s control: a final configuration, a finished KYC and export package, and a site validated for power and cooling. With all three done, the order can go in as soon as the OEM confirms supply, and Iternal starts escalating with OEM executives from that point.
US export rules and OEM KYC
Datacenter GPU systems are export-controlled, and OEMs ask for know-your-customer and export documentation before they ship. Iternal assembles that package with you and coordinates it with the OEM’s trade-compliance team, while your organization remains responsible for compliance.
ECCN 3A090.a / 4A090.a
US export rules control advanced AI chips under ECCN 3A090.a. Computers and assemblies that contain them fall under ECCN 4A090.a or corresponding .z entries, such as servers classified as 5A992.z (eCFR, 15 CFR Part 774, Nov 2026).
Country Group D:5
Exporting these items to an entity headquartered in, or whose ultimate parent is headquartered in, a Country Group D:5 country requires a BIS license for any destination outside the United States, unless a listed license exception applies (US Bureau of Industry and Security, May 2026). The ownership structure up to the ultimate parent therefore belongs in the package from the start.
BIS red-flag guidance
BIS guidance asks suppliers to obtain an end-user certification that identifies every party and the end use, plus a written attestation from the data center. The same guidance lists a data center that cannot affirm its power, cooling and space as a red flag (US Bureau of Industry and Security, May 2025), so a validated site supports the compliance file as well as the deployment plan.
BIS license review
BIS processed about 30,500 export license applications in 2025, with an average processing time of 62 days (BIS FY2025 Annual Report, Jul 2026). Orders that need a license carry that review in their schedule.
OEM terms of sale
Dell’s US commercial terms make the buyer responsible for compliance with export-control and sanctions laws (Dell Technologies, US Commercial Terms of Sale, Sep 2026). Buying direct from the OEM does not remove that obligation or transfer it to Iternal.
Chain of custody
In December 2025 the US Justice Department announced it had shut down a smuggling network that, according to court documents, exported or attempted to export at least $160 million in export-controlled datacenter GPUs between October 2024 and May 2025 (US Department of Justice, Dec 2025). An OEM-direct order keeps a documented chain of custody between the OEM and your organization, because Iternal never takes title to the hardware.
Buyer groups and readiness
Iternal sources multi-node and rack-scale orders of datacenter GPU systems. There is no fixed minimum order; Iternal confirms fit from the budget and server count in your request.
Organizations that run their own training and inference clusters in an owned data center or a colocation facility.
Operators adding owned capacity that they rent or resell as compute to their own customers.
National and regional programs building sovereign AI infrastructure for public research, services and industry.
Research computing centers expanding shared GPU clusters for faculty and research groups.
Teams that train and serve their own models on hardware they own.
Lead times, pricing and paperwork
NVIDIA sells its datacenter platforms through select NVIDIA Partner Network partners and points buyers to cloud providers, OEM partner servers and DGX systems (NVIDIA, Nov 2026). For an organization that wants to own HGX B300, GB300 NVL72 or Vera Rubin systems, the practical path is an order with a server OEM; NVIDIA names Dell Technologies, HPE and Lenovo among its top system builders (NVIDIA, May 2026).
Iternal runs that OEM-direct order end to end. Your organization contracts with and pays the OEM.
A reseller buys systems from the OEM or a distributor and resells them, with its margin inside the price. A GPU broker locates units or capacity from third parties, often on the secondary market. With Iternal, the OEM sells to you directly: you contract with and pay the OEM, and Iternal’s fee is one disclosed line on the OEM’s quote that you can compare against any other option. Iternal adds no hardware markup, because it never takes title to the hardware.
Authorized resellers bring real strengths, including credit lines, multi-brand carts and integration services, and they pass through the OEM warranty. The difference lies in who holds the contract and whether the intermediary’s pay is visible.
You contract with and pay the OEM, on the OEM’s terms after its credit review. Dell’s standard US commercial terms, for example, set payment at 30 days after invoice and pass title when Dell hands the hardware to the carrier (Dell Technologies, US Commercial Terms of Sale, Sep 2026). Warranty and support come from the OEM under that contract.
Iternal’s fee is one disclosed, success-based line on the OEM’s quote: if no order is placed, nothing is owed. Iternal takes no markup on hardware and never takes title to it.
OEM financing is subject to the OEM’s credit approval. HPE offers financing through HPE Financial Services in certain countries (HPE, Nov 2026), Lenovo offers financing and consumption models through Lenovo Global Financial Services (Lenovo Press, Oct 2026), and Dell’s finance arm is Dell Financial Services.
There is no published lead time for these systems. OEMs confirm delivery dates once an order is in their system, and the date depends on the configuration, the OEM’s supply and the readiness of the site.
NVIDIA said in May 2026 that Vera Rubin production shipments would begin in fall 2026 (NVIDIA, May 2026), and Lenovo states that it is shipping Vera Rubin NVL72 globally (Lenovo Press, Oct 2026).
Three things shorten the path, and all three sit with the buyer: a final configuration, a finished KYC and export package, and a site validated for power and cooling. Iternal coordinates all three with you and the OEM’s design, trade-compliance and facility teams before the order goes in, then escalates with OEM executives once the order is in the OEM’s system.
OEMs price these systems on quotes. HPE’s US online store, for example, lists the NVIDIA GB300 NVL72 by HPE as request-a-quote, with no posted price (HPE, Nov 2026).
The total depends on GPU count and configuration, networking, storage, OEM services, the support term and memory, where TrendForce expects server DRAM contract prices to rise about 270% in 2026 (TrendForce, Aug 2026). Quotes also expire: Dell’s standard terms keep a quote valid for 30 days, subject to change (Dell Technologies, US Commercial Terms of Sale, Sep 2026), and Dell’s chief operating officer said in early 2026 that Dell’s quotes “are valid for the shortest period of time they’ve ever been” (CIO Dive, Mar 2026).
For the next generation, TrendForce expects Vera Rubin systems to carry roughly twice the average selling price of GB300 (TrendForce, Aug 2026). Iternal’s fee appears as its own line on the OEM’s quote, so you can compare it with the hardware price.
US export rules control advanced AI chips under ECCN 3A090.a, and computers that contain them fall under ECCN 4A090.a or corresponding .z entries (eCFR, 15 CFR Part 774, Nov 2026). Exports to an entity headquartered in, or whose ultimate parent is headquartered in, a Country Group D:5 country require a BIS license for any destination outside the United States, unless a listed license exception applies (US Bureau of Industry and Security, May 2026).
Expect an end-user certification that identifies every party and the end use, plus a written attestation from the data center (US Bureau of Industry and Security, May 2025). When a license is needed, BIS reported an average processing time of 62 days for the applications it handled in 2025 (BIS FY2025 Annual Report, Jul 2026).
The buyer stays responsible: Dell’s US terms, for example, make the buyer responsible for compliance with export-control and sanctions laws (Dell Technologies, US Commercial Terms of Sale, Sep 2026). Iternal assembles the package with you and coordinates it with the OEM’s trade-compliance team, and Iternal provides no legal advice.
The answer turns on your deployment date, facility and budget. TrendForce expects GB300 rack systems to lead NVIDIA’s 2026 shipments with demand through the first half of 2027, Vera Rubin racks to take the largest share in 2027, and Vera Rubin systems to carry roughly twice the average selling price of GB300 (TrendForce, Aug 2026).
The facilities differ as well. Lenovo’s GB300 NVL72 datasheet describes a hybrid design that relies primarily on warm-water liquid cooling at 18 to 45 °C, with supplemental airflow (Lenovo Press, Mar 2026). NVIDIA states that Vera Rubin NVL72 racks are designed for 45 °C liquid-cooling inlet operation and advises facilities teams to validate that design early in planning (NVIDIA Technical Blog, Aug 2026).
Iternal can have the OEM quote both generations and bring in the OEM’s facility specialists, so you can compare price, delivery and site work side by side.
Work email required
What happens next
Iternal has your request and will reply by email. To shorten the OEM’s KYC review, start gathering the items listed under export-control and KYC readiness above.
NVIDIA, May 2026. NVIDIA said in May 2026 that production shipments of Vera Rubin would begin in fall 2026, naming Dell Technologies, HPE and Lenovo among its top system builders. nvidianews.nvidia.com
Dell Technologies, Mar 2026. Dell's AI Factory with NVIDIA lineup includes the PowerEdge XE9812 for Vera Rubin NVL72 and HGX Rubin NVL8 servers (XE9880L, XE9885L, XE9882L) in a 100% direct-liquid-cooled rack architecture. dell.com
Lenovo Press, Oct 2026. Lenovo states it is shipping NVIDIA Vera Rubin NVL72 globally and offers financing and consumption models through Lenovo Global Financial Services. lenovopress.lenovo.com
HPE, Nov 2026. HPE's US online store lists the NVIDIA GB300 NVL72 by HPE as request-a-quote, with no posted price, and offers financing through HPE Financial Services in certain countries, subject to credit approval. buy.hpe.com
Dell Technologies, US Commercial Terms of Sale, Sep 2026. Dell's standard US commercial terms set payment at 30 days after invoice, pass title when Dell hands the hardware to the carrier, and keep quotes valid for 30 days, subject to change for shortages, cost increases, currency moves or tariffs. i.dell.com
Dell Technologies, US Commercial Terms of Sale, Sep 2026. Dell's US terms make the buyer responsible for compliance with export-control and sanctions laws. i.dell.com
TrendForce, Aug 2026. TrendForce expects GB300 rack systems to lead NVIDIA's 2026 shipments with demand through the first half of 2027, Vera Rubin racks to take the largest share in 2027, and Vera Rubin systems to carry roughly twice the average selling price of GB300. trendforce.com
TrendForce, Aug 2026. TrendForce estimates server DRAM contract prices rose 64% in the second half of 2025 and expects a rise of about 270% in 2026. trendforce.com
CIO Dive, Mar 2026. Dell’s chief operating officer said in early 2026 that Dell’s quotes “are valid for the shortest period of time they’ve ever been.” ciodive.com
Supply Chain Dive, Jun 2026. On Dell’s May 2026 earnings call, its chief operating officer said demand “continues to exceed supply with memory as the primary constraint”; Dell is prioritizing large strategic customers, and its largest customers are locking in supply through three- to five-year deals. supplychaindive.com
StorageNewsletter (Dell fiscal Q2 2027 results), Sep 2026. Dell reported a record $95 billion AI server backlog and $60.9 billion in AI server orders in its fiscal second quarter of 2027. storagenewsletter.com
NVIDIA Q4 FY2026 earnings call, via The Motley Fool, Feb 2026. NVIDIA said its sovereign AI business more than tripled year over year, to over $30 billion, in fiscal 2026. fool.com
CNAS, Aug 2026. The CNAS sovereign AI index tracks 184 projects across 67 countries and the EU, and infrastructure made up 80% of new projects in the first half of 2026. cnas.org
EuroHPC Joint Undertaking, Jul 2026. The EuroHPC Joint Undertaking opened a call for tenders in July 2026 for up to seven AI Gigafactories, expected in at least seven EU Member States, with more than €20 billion in private investment expected. eurohpc-ju.europa.eu
eCFR, 15 CFR Part 774, Nov 2026. US export rules control advanced AI chips under ECCN 3A090.a; computers and assemblies that contain them fall under ECCN 4A090.a or corresponding .z entries, such as servers classified as 5A992.z. ecfr.gov
US Bureau of Industry and Security, May 2026. Exporting these items to an entity headquartered in, or whose ultimate parent is headquartered in, a Country Group D:5 country requires a BIS license for any destination outside the United States, unless a listed license exception applies. bis.gov
US Bureau of Industry and Security, May 2025. BIS guidance asks suppliers to obtain an end-user certification that identifies every party and the end use, plus a written attestation from the data center, and lists a data center that cannot affirm its power, cooling and space as a red flag. bis.gov
BIS FY2025 Annual Report, Jul 2026. BIS processed about 30,500 export license applications in 2025, with an average processing time of 62 days. bis.gov
US Department of Justice, Dec 2025. In December 2025 the US Justice Department announced it had shut down a smuggling network that, according to court documents, exported or attempted to export at least $160 million in export-controlled datacenter GPUs between October 2024 and May 2025. justice.gov
NVIDIA, Nov 2026. NVIDIA sells its datacenter platforms through select NVIDIA Partner Network partners and points buyers to cloud providers, OEM partner servers and DGX systems. nvidia.com
NVIDIA DGX B300 User Guide, Nov 2026. NVIDIA’s DGX B300, an eight-GPU Blackwell Ultra system, is air-cooled and rated at 15 kW maximum. docs.nvidia.com
Lenovo Press, Mar 2026. Lenovo’s GB300 NVL72 datasheet describes a hybrid design that relies primarily on warm-water liquid cooling at 18 to 45 °C, with supplemental airflow, removing most of the heat without chilled water. lenovopress.lenovo.com
NVIDIA Technical Blog, Aug 2026. NVIDIA states that Vera Rubin NVL72 racks are designed for 45 °C liquid-cooling inlet operation and advises facilities teams to validate that design early in planning. developer.nvidia.com