iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Nvidia’s major AI-infrastructure announcements are best understood as multi-year development plans, not as gigawatts of data-center capacity already operating. They combine processor and networking commitments with land, buildings, electricity supply, grid work and financing. The headline figures show the scale of potential buildout, while schedules, permits, equipment availability and power connections determine what is ultimately delivered.
The deals also sit inside a rapidly growing data-center electricity market. The International Energy Agency (IEA) reported that data-center electricity demand increased 17% in 2025, compared with 3% growth in global electricity demand. That sector-wide increase cannot be assigned to Nvidia’s agreements alone, and the local effect on grids or household bills will depend on where projects are built and how their electricity is supplied.
What Nvidia’s announcements actually represent
Nvidia is helping turn expected demand for AI computing into physical infrastructure programs. A typical arrangement can cover high-performance systems, networking, a data-center campus, a customer commitment, construction responsibility, electricity generation and grid upgrades. That is materially different from announcing a shipment of servers or reporting a facility that is already energized.
Three distinctions are essential when reading the numbers:
- Announced versus operating: letters of intent, investment intentions and options describe future activity. They do not prove that the stated capacity is online.
- IT capacity versus generation capacity: an IT-gigawatt figure describes the power available to computing equipment. A generation-gigawatt figure describes generating capability that may serve the facility and other system needs. They cannot be added as if they were the same measure.
- Pipeline versus commissioned capacity: a company can have land, a power reservation or a construction plan without having completed the buildings, grid connection or equipment installation.
Because the OpenAI and PORTS-Pike announcements are related and the projects use different definitions and schedules, their gigawatt figures should not be summed. The available announcements also do not establish how much electricity use or operating capacity is attributable to Nvidia’s deals specifically.
The announced projects and their status
| Announcement | Capacity description | Timing and status | Power, ownership or investment terms |
|---|---|---|---|
| OpenAI partnership, September 22, 2025 | At least 10 GW of Nvidia systems for OpenAI’s next-generation infrastructure | Letter of intent; first phase targeted for the second half of 2026 on the Vera Rubin platform | Nvidia said it intended to invest up to $100 billion progressively as systems were deployed |
| PORTS-Pike, Pike County, Ohio, August 17, 2026 | Initial 4.25 IT-GW, with an option for a further 3.75 IT-GW; the announcement describes 8 IT-GW of AI-factory capacity in total | Planned phases begin in 2028; announced land, power and shell capacity rather than verified operating capacity | At least 10 GW of new generation to support 8 IT-GW; at least $4.2 billion in regional grid infrastructure; Nvidia investment of $1.5 billion in SB Energy. SB Energy is to build, own and operate the site for OpenAI under a 20-year lease |
| IREN partnership, May 7, 2026 | Up to 5 GW of DSX-aligned infrastructure across IREN’s global pipeline, with future deployments expected to focus on the 2-GW Sweetwater campus in Texas | Prospective support across a pipeline; not a verified operating total | The announcement describes intended support for future deployments rather than a completed facility |
The OpenAI announcement is the broadest headline commitment: at least 10 GW of Nvidia systems and a potential $100 billion investment tied to deployment. It is still a letter of intent, and its release does not verify how much equipment was operating when the announcement was made.
PORTS-Pike illustrates how these projects are assembled. The site combines a customer (OpenAI), a developer and operator (SB Energy), Nvidia systems, a long-term lease, dedicated generation and regional grid construction. Its 4.25 IT-GW initial phase and 3.75 IT-GW option describe computing load, while the “at least 10 GW” generation figure describes a different layer of the power plan. The 2028 start for planned phases leaves several years for design, approvals, procurement and construction.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →The IREN agreement applies to up to 5 GW across a global pipeline rather than one operating campus. Sweetwater, a 2-GW campus in Texas, is identified as a focus for future deployments. That wording makes the 5-GW number a development opportunity, not a measured total of energized Nvidia capacity.
Why the gigawatt scale matters
A gigawatt is one billion watts of instantaneous power. For an AI campus, an IT-GW figure can represent a very large collection of accelerated servers, networking equipment, cooling systems and related electrical infrastructure. It is a power rating, not an annual energy total: annual consumption depends on how continuously the equipment runs and on the efficiency of the facility.
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
Generation capacity is also not identical to delivered IT power. A project needs substations, transformers, transmission or distribution connections, backup arrangements and internal electrical systems between a generator and the servers. Some generation may support the wider grid, provide reserve capacity or operate at a different time from the computing load.
That is why PORTS-Pike’s description of at least 10 GW of new generation for 8 IT-GW of AI-factory capacity should not be interpreted as 18 GW of data-center capacity. It indicates a proposed supply arrangement around an 8-IT-GW buildout.
Free tools Windows power users keep installed
One-click scans. No signup required.
How much electricity data centers use
Current global level
The IEA estimated that data centers consumed 415 terawatt-hours (TWh) of electricity in 2024, about 1.5% of global electricity use. TWh measures energy consumed over time, unlike GW, which measures instantaneous capacity. Because facilities are concentrated in particular regions, a 1.5% global share can still create a much larger local impact on generation, transmission and distribution systems.
Observed 2025 growth
In its April 16, 2026 update, the IEA reported that data-center electricity demand rose 17% in 2025, while total global electricity demand rose 3%. The comparison shows that data centers were growing much faster than electricity demand overall, but it describes the entire sector. It does not isolate Nvidia-powered facilities or any one corporate deal.
Forecasts through 2030
The IEA’s 2025 Energy and AI analysis projected global data-center electricity consumption at around 945 TWh in 2030 in its base case. It also said consumption is set to double by 2030 and that electricity use by AI-focused data centers could triple. These are forecasts based on assumptions about AI adoption, efficiency, utilization and infrastructure availability, not settled outcomes.
Rank #3
- Video/Sound Cards
- Passive Cooling
Accelerated servers, mainly associated with AI workloads, were projected to account for almost half of net growth in global data-center electricity consumption through 2030. The IEA also models alternatives in which efficiency gains restrain demand or faster AI adoption and more flexible infrastructure increase it. A single number therefore cannot capture the range of plausible outcomes.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallWhere the electricity may come from
Data-center growth does not imply one universal fuel mix. The IEA’s 2025 supply analysis identifies renewables as the fastest-growing source for data-center electricity. It projects renewables to meet nearly half of demand growth from 2024 to 2030, while natural gas and coal together meet more than 40% of additional demand over that period. Nuclear power is expected to make a larger contribution after 2030.
Those are global scenario results. A facility in Texas, Ohio or another region will reflect that area’s existing generators, transmission constraints, market rules and new-build economics. A company may contract for renewable energy while drawing physical power through a grid that also contains gas, coal, nuclear and hydroelectric generation. Contractual procurement and the electricity physically flowing to a server are related but not identical concepts.
Can the grid keep up with AI campuses?
The main risk is not simply whether enough fuel exists. The IEA’s 2026 update identifies tightened supply chains for gas turbines, transformers, advanced chips and other IT components. Planning and regulatory systems are also under strain, delaying grid connections and approvals. A project can therefore have a customer and a site while waiting for equipment, permits or a deliverable interconnection.
Grid connection and approvals
Large campuses can require new substations, transmission work and generation. Interconnection queues, environmental reviews, land-use decisions and construction permits can take years. The PORTS-Pike announcement’s planned 2028 phases show why an announced gigawatt total should not be treated as immediately available capacity.
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
Equipment and construction bottlenecks
Transformers and gas turbines are specialized products with long lead times. Advanced chips, servers, cooling systems and electrical components must arrive in a coordinated sequence. Shortage of any one item can leave a partially built site unable to operate at its planned load.
Fast-changing AI demand
AI data centers can experience rapid and large swings in demand. Reliable supply must handle those changes without destabilizing the local system. The IEA says onsite natural-gas projects remain early-stage in many cases and face technical and financial hurdles, so an announced onsite plant is not automatically a near-term solution.
Storage and flexible workloads
Onsite batteries can help manage peaks, provide short-duration backup and reduce the speed of changes seen by the grid. In September 2026, Nvidia, Google and Emerald AI announced the AI Energy Management Alliance. Nvidia’s description says a flexible facility could shift computing workloads, discharge storage, use paired generation or respond to system contingencies. Those are intended operating strategies, not evidence that flexibility eliminates the need for new generation, transmission or grid upgrades.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What the deals could mean for local electricity systems
A new AI campus can increase demand in one utility territory far more than the global percentage suggests. The local outcome depends on the project’s size, utilization, cooling design, connection voltage, contracted supply and the timing of other industrial loads.
Utilities may recover the cost of substations, transmission, generation or reserve capacity through negotiated contracts and rate structures. In some jurisdictions, a large customer pays for dedicated facilities; in others, some costs can be allocated across a broader customer base. The announcements listed here do not establish the tariff treatment or household-bill effect for every community, so no general claim about bills is justified.
PORTS-Pike’s stated commitment to at least $4.2 billion in regional grid infrastructure shows that the physical network can be a major part of the project. It does not reveal how those costs will be allocated among Nvidia, SB Energy, OpenAI, utilities, public agencies or other customers.
How to evaluate a new Nvidia infrastructure announcement
- Identify the date and document type. A letter of intent, memorandum, option or company intention is different from a construction notice, an interconnection agreement or a commissioning report.
- Read the capacity definition. Check whether the number is IT load, total facility load, generation capability, a campus maximum or a pipeline estimate.
- Separate phases and options. Record the initial phase, optional expansion and target start date independently. An option is not an operating asset.
- Locate the project and grid connection. Geography determines available generation, transmission constraints, permitting requirements and likely construction sequence.
- Assign responsibilities. Note who supplies the chips and systems, who builds and operates the facility, who owns generation, who is the customer and who funds grid work.
- Check commissioning evidence. Look for energization, installed equipment, commercial operation or measured load before calling capacity operational.
- Keep sector forecasts separate. IEA totals and scenarios describe worldwide data centers. They provide context for scale but do not measure electricity caused by a particular Nvidia agreement.
What to watch before treating planned capacity as real
- Final permits and an executed grid-interconnection agreement.
- Orders and delivery schedules for transformers, turbines, servers and cooling equipment.
- Construction milestones for substations, generation and the data-center shell.
- Evidence that the first phase has been energized and is carrying a computing load.
- Updates that distinguish an exercised expansion option from a reserved one.
- Utility or regulatory filings showing how infrastructure costs and electricity rates are assigned.
Nvidia’s agreements show how AI infrastructure is becoming a coordinated buildout of compute, land, buildings and power rather than a standalone hardware purchase. Their scale is consequential, but the numbers remain plans until projects clear the physical and regulatory steps between announcement and operation. The IEA’s sector data confirms fast-growing electricity demand and a wide range of possible 2030 outcomes; it does not support attributing that growth to Nvidia alone.
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.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problems
