Free tools Windows power users keep installed
One-click scans. No signup required.
Nscale’s October 15, 2025 announcement was a planned, multi-site Microsoft infrastructure contract covering approximately 200,000 NVIDIA GB300 GPUs. It was not necessarily a single $14 billion cash purchase of chips, and the public announcement did not show that all of the GPUs had already been delivered or put into operation.
The arrangement covered AI-compute capacity in Texas, Portugal, England and Norway. Later 2026 announcements added substantially larger Rubin GPU projects, including a letter of intent for 1.35 GW of capacity in West Virginia. Those later commitments should be treated as follow-on developments, not folded into the original GB300 figure.
As an Amazon Associate I earn from qualifying purchases.
The short version
- Announcement: October 15, 2025
- Customer: Microsoft
- Infrastructure provider: Nscale, including its Nscale-Aker joint venture for the Norway project
- Hardware: Approximately 200,000 NVIDIA GB300 GPUs
- Locations: Texas, Portugal, England and Norway
- Reported value: Up to approximately $14 billion, according to secondary coverage
- Commercial structure: A staged infrastructure and AI-compute-capacity arrangement, rather than a clearly disclosed chip-only purchase
Nscale’s announcement describes a vertically integrated role involving data-center development, GPU clusters, power, networking, orchestration and AI-compute services. Dell Technologies was also identified as part of the infrastructure rollout.
Where the original GPUs were planned to go
| Site | Approximate GPUs | Infrastructure details | Planned timing |
|---|---|---|---|
| Texas | 104,000 | Approximately 240 MW initially, with a longer-term target toward 1.2 GW | Expected over 12–18 months from the October 2025 announcement |
| Sines, Portugal | 12,600 | Start Campus site | Planned from the first quarter of 2026 |
| Loughton, England | 23,000 | Nscale campus | Planned from 2027 |
| Narvik, Norway | 52,000 | Nscale-Aker hyperscale AI campus | Staged deployment |
These listed allocations total approximately 191,600 GPUs. That is consistent with the announcement’s rounded “approximately 200,000” figure; it should not be treated as a contradictory exact inventory count.
What Microsoft is actually securing
The public material supports describing Microsoft as securing access to hosted AI infrastructure built around NVIDIA systems. Nscale is supplying more than standalone processors: the arrangement involves facilities, electricity, cooling, networking, cluster management and operating capacity.
That distinction matters financially. A conventional hardware purchase would imply Microsoft simply bought and took delivery of 200,000 chips. The announced structure instead points to a multi-year capacity and infrastructure relationship. The public announcements do not disclose the full pricing, payment schedule, utilization guarantees, take-or-pay provisions or termination rights.
Why the deal matters to Microsoft
Microsoft needs large amounts of computing capacity for AI training, fine-tuning and inference across its own products and cloud customers. Working with a specialist infrastructure provider can supplement Azure’s internal construction program and give Microsoft access to capacity in multiple regions.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →The announcements do not establish that the GPUs are dedicated to a particular Microsoft product, OpenAI model or named customer. The defensible conclusion is broader: Microsoft is expanding access to advanced AI capacity while diversifying where that capacity is built.
Why the deal matters to Nscale and Aker
For Nscale, Microsoft provides the kind of anchor demand that can support financing, equipment procurement and the construction of large AI campuses. It also gives a relatively young infrastructure company a path toward hyperscale operations across Europe and the United States.
Rank #2
- Part number 900-53651-2500-000 and model: P3651
- This is the 2 slot version for when there is no empty slots between 2 slot cards. If you have one or more empty slots between the cards or the cards are 3 slot this NVLink will not work. See the attached images showing the card layout.
- NVLink 3.0 for any brand of RTX Ampere model graphics cards: 3090, A30, A40, A100 / H100 (Requires three NVLinks), A800, A4500, A5000, A5500, A6000
- This is the same as PNY part number: NVLAMP-2SLOT-BSP and RTXA6000NVLINK-KIT
- This is the same as Dell part number: 0RWJ7Y
But contracted capacity is not the same as recognized revenue, profit or completed infrastructure. Each project still depends on power availability, grid connections, financing, construction, GPU delivery, networking, cooling and reliable operations.
Aker’s role should also be separated carefully. Aker is an investor in Nscale and participates through the Nscale-Aker joint venture associated with the Norway campus. That does not make Aker the customer: Microsoft is the capacity buyer, while Nscale and its partners provide and operate the infrastructure.
Recommended Free Tools
What does “up to $14 billion” mean?
Secondary coverage reported the original arrangement as worth up to approximately $14 billion. That wording should not be converted into an exact purchase price or assumed to mean Microsoft paid $14 billion upfront. The figure may encompass multi-year capacity reservations, equipment, facility construction and operating services.
There is also no basis for saying Nscale will earn $14 billion as revenue or profit. Contract value, bookings, recognized revenue, cash received and profit are different measures. Without the underlying contract, those details remain undisclosed.
Why the Texas numbers need context
Texas was the largest original component, with approximately 104,000 GB300 GPUs and an initial capacity of about 240 MW. The project was described as having a longer-term expansion path toward 1.2 GW. Reporting also described an option for Microsoft to add another 700 MW of AI capacity from 2027.
Rank #3
- Video/Sound Cards
- Passive Cooling
Those are different stages of development. The 240 MW figure refers to the initial stated capacity; 1.2 GW is a longer-term target, and the additional 700 MW is an option or planned expansion—not capacity already delivered.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
GB300 is not the same as Rubin
The original October 2025 arrangement concerned NVIDIA GB300 systems. Later announcements shifted toward NVIDIA Vera Rubin NVL72 systems and future technologies.
GPU counts from the two generations should not be added together or compared as if they represented identical computing power. System configuration, networking, power requirements and platform performance all matter. A later Rubin announcement is not retroactive evidence about the October 2025 GB300 deployment.
What happened after the original announcement?
March 2026: West Virginia Monarch
In March 2026, Nscale announced plans to acquire American Intelligence & Power Corporation, owner of the Monarch Compute Campus, and separately announced a letter of intent with Microsoft covering approximately 1.35 GW of AI-compute capacity and about 430,000 NVIDIA Vera Rubin GPUs.
The campus was described as spanning 2,250 acres, with development targets of approximately 2 GW by the first half of 2028 and expansion toward 8 GW by 2031. Those are targets, not completed operating capacity. See the Euronext announcement and the RNS filing.
Rank #4
- CUDA Cores: 4608 / NVIDIA Tensor Cores: 576 / NVIDIA RT Cores: 72
- GPU Memory: 24 GB GDDR6 with ECC / Bandwidth: 624 GB/Sec
- System Interface: PCI Express 3.0 x16
- Four DisplayPort 1.4 Connectors
- 3D Stereo Support with Stereo Connector
April 2026: Norway expansion
On April 14, 2026, Nscale announced an expansion at its 230 MW Narvik project involving more than 30,000 NVIDIA Rubin GPUs for Microsoft, with deployment planned for 2027. This was an expansion of the Norway relationship and should not be silently substituted for the original 52,000-GB300 allocation. Nscale’s announcement provides the company’s account of the project.
May 2026: Portugal expansion
In May 2026, Nscale announced a later Sines expansion involving more than 66,000 Rubin GPUs at the Start Campus site, beginning in late 2027. That is separate from the original 12,600-GPU Portugal allocation. The announcement was distributed through PR Newswire.
Financing developments
Nscale announced $790 million in financing for its Norway buildout in May 2026 and a $900 million revolving credit facility in July 2026. Financing supports the company’s ability to develop capacity, but it does not by itself prove that every announced GPU deployment is complete or operating.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.The main risks and open questions
- Execution: Construction, power connections, cooling, networking and equipment delivery can delay deployment.
- Financing: Large campuses require substantial debt and equity, and development targets can change.
- Power: Announced megawatts may describe planned or ultimate capacity rather than immediately available IT load.
- Technology transitions: New Rubin systems may arrive while earlier GB300 facilities are still being built.
- Customer concentration: Microsoft is a powerful anchor customer, but dependence on one buyer creates commercial exposure.
- Contract uncertainty: Pricing, utilization, payment timing and termination protections have not been fully disclosed.
- GPU-count ambiguity: Public announcements use “GPUs” as their unit, but a GPU count alone does not describe complete system performance.
- Environmental impact: Renewable electricity claims do not eliminate construction, transmission, cooling, water-use or local grid impacts.
- Regulation: Advanced-chip exports, licensing and cross-border logistics can affect international deployments.
Why investors and business readers should care
The announcement illustrates the economics of the AI infrastructure buildout. Demand from a major cloud and software company can help a specialist provider finance facilities and commit to expensive GPU systems. In return, Microsoft can supplement its own cloud capacity without owning every site or managing every construction project directly.
However, the financial opportunity comes with infrastructure risk. The value of an AI campus depends not only on the number of processors ordered, but also on when power becomes available, whether the equipment is delivered, how efficiently the clusters operate and whether customers use the reserved capacity.
Bottom line
Nscale’s October 15, 2025 announcement was a major commitment to Microsoft-accessible AI infrastructure: approximately 200,000 GB300 GPUs planned across four sites in the United States and Europe. The headline should be understood as a staged, infrastructure-and-capacity arrangement, not proof of a completed $14 billion chip purchase.
By August 2026, the relationship had expanded through much larger Rubin-related announcements in West Virginia, Norway and Portugal. Those developments strengthen the strategic importance of the relationship, but they remain distinct from the original 200,000-GB300 announcement and include a later letter of intent whose legal status differs from a completed contract.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
