SemiconductorsBRIEF

NVIDIA and OpenAI Expand AI Mega Data Center Push to 10 GW Power Scale

林紀旭 James LinEditor-in-Chief
Published · Updated
OpenAI, SB Energy, and NVIDIA are building an 8 IT-GW AI campus in Ohio, backed by NVIDIA's $1.5 billion SB Energy investment and a prior $100 billion commitment to OpenAI, while SB Energy and SoftBank plan at least 10 GW of new power generation and $4.2 billion in grid infrastructure to support it.

How Are NVIDIA and OpenAI Teaming Up, and Which Partners Are Involved?

OpenAI, SB Energy, and NVIDIA are jointly building roughly 8 IT-GW of AI data center capacity at the PORTS-Pike Technology Campus in OhioCITE:E1. Within this structure, NVIDIA is providing credit support for the campus's initial 4.25 IT-GW of Local Power System (LPS) build-out and holds an option to take on the remaining 3.75 IT-GW of capacity, while OpenAI has signed a 20-year lease with SB EnergyCITE:E4. This Ohio project builds on an earlier, broader partnership: in September 2025, NVIDIA announced a $100 billion (roughly NT$3 trillion) investment to build OpenAI's next-generation AI data centersCITE:E10, with the first gigawatt-scale system slated to go live on NVIDIA's Vera Rubin platform in the second half of 2026CITE:E12.

How Big Is the Mega Data Center, and What Are the GPU and Technical Specs?

The PORTS-Pike campus is sized at approximately 8 IT-GW, while SB Energy and SoftBank are planning at least 10 GW of new generation capacity to power itCITE:E1CITE:E2. This follows NVIDIA's September 2025 announcement that it would build OpenAI a next-generation AI data center with total power demand of 10 GW, equipped with millions of GPUs to support advanced AI model developmentCITE:E11. NVIDIA further estimates that each generation of its deployed AI Factory systems corresponds to about 1.5 million NVIDIA GPUsCITE:E7.

MetricValueSource
PORTS-Pike campus capacity8 IT-GWCITE:E1
SB Energy/SoftBank planned new generationat least 10 GWCITE:E2
Initial LPS build-out (NVIDIA credit support)4.25 IT-GWCITE:E4
NVIDIA's optional additional capacity3.75 IT-GWCITE:E4
September 2025 OpenAI data center power demand10 GWCITE:E11
GPUs per AI Factory generation1.5 millionCITE:E7

How Much Money Is NVIDIA Committing, and What Is Its Financing Role?

NVIDIA is investing $1.5 billion directly in SB Energy as part of its expanding role in AI infrastructureCITE:E3. Beyond that direct equity stake, NVIDIA is backstopping the PORTS-Pike build-out by providing credit support for the initial 4.25 IT-GW phase and holding an option on the remaining 3.75 IT-GW, while OpenAI's commitment takes the form of a 20-year lease with SB EnergyCITE:E4. This financing structure sits on top of NVIDIA's earlier and larger commitment: the $100 billion (about NT$3 trillion) investment announced in September 2025 to build OpenAI's next-generation AI data centers, which NVIDIA said would be deployed in a phased investment modelCITE:E10CITE:E12.

How Much Power Infrastructure Is Needed, and What Is the Build Timeline?

SB Energy and SoftBank plan to add at least 10 GW of generation capacity, with the first roughly 800 MW coming online in 2028CITE:E2CITE:E5. That initial phase will mainly draw on AEP's existing power infrastructure, while subsequent expansion will require new power plants and transmission facilities, with natural gas explicitly named as one power sourceCITE:E5. To support the broader build-out, SB Energy and SoftBank also plan to invest at least $4.2 billion in regional grid infrastructureCITE:E6.

Power MilestoneFigureSource
First capacity tranche online800 MW by 2028CITE:E5
Planned new generation capacityat least 10 GWCITE:E2
Regional grid infrastructure investmentat least $4.2 billionCITE:E6

How Big Is the Commercial Opportunity for NVIDIA and OpenAI?

NVIDIA estimates that each generation of AI Factory systems, at about 1.5 million GPUs, could generate roughly $150 billion to $200 billion in revenueCITE:E7. Looking further out, NVIDIA says OpenAI's existing and planned NVIDIA Compute capacity is set to reach about 12 GW by 2030, and if the PORTS-Pike partnership is expanded beyond its initial 4.25 GW scope, that total opportunity could grow to about 16 GW, corresponding to roughly $600 billion in potential NVIDIA Compute business by 2030CITE:E8.

Commercial MetricFigureSource
Revenue per AI Factory generation$150–200 billionCITE:E7
OpenAI's NVIDIA Compute by 2030 (current plan)~12 GWCITE:E8
OpenAI's NVIDIA Compute by 2030 (expanded PORTS-Pike)~16 GWCITE:E8
Potential NVIDIA Compute business by 2030~$600 billionCITE:E8

Where Does This Fit in the Global AI Infrastructure Race, and How Is It Viewed?

The PORTS-Pike campus joins a cluster of US AI infrastructure projects now scaling toward 10 GW and beyond, including OpenAI's own Stargate project, which originally set a 10 GW build target, Fermi's Project Matador, planned at up to 17 GW, and Creekstone Energy's Delta Gigasite in Utah, planned at over 10 GWCITE:E9. On the technology side, NVIDIA's first gigawatt-scale system is due to go live on its Vera Rubin platform in the second half of 2026CITE:E12. Commenting on the NVIDIA-OpenAI partnership, Microsoft Taiwan General Manager Bian Zhixiang (卞志祥) said NVIDIA's investment in OpenAI would form a positive cycle, in which OpenAI develops models that are then delivered to industry through Microsoft's platform to advance general-purpose AICITE:E13. Bian also said Microsoft remains OpenAI's primary compute supplier, though OpenAI may seek other resources when its compute needs exceed what Microsoft can supplyCITE:E14. He added that Microsoft's Taiwan data center has activated more than 250 features synchronized with its global data centers, meeting multiple international regulatory standards for data complianceCITE:E15.

What This Means

Taken together, the figures show a partnership operating on two tracks: a $100 billion, 10 GW commitment announced in September 2025CITE:E10CITE:E11, and a more granular Ohio build-out disclosed in August 2026 that breaks that scale into an 8 IT-GW campus, a 4.25/3.75 IT-GW financing split, and a 2028 first-power dateCITE:E1CITE:E4CITE:E5. NVIDIA's dual role as both investor ($1.5 billion in SB Energy, $100 billion tied to OpenAI's data centers) and infrastructure financier (credit support and an option on additional capacity) sits alongside its own projection that the resulting compute business could be worth up to $600 billion by 2030CITE:E3CITE:E10CITE:E8. That projection depends on power delivery timelines — an 800 MW first tranche in 2028, scaling toward a 10 GW-plus build — set against a landscape where Stargate, Project Matador, and Delta Gigasite are pursuing comparable or larger gigawatt-scale targetsCITE:E5CITE:E2CITE:E9.

📊 Evidence

📎 Sources

  1. technews.tw
  2. money.udn.com
Author's Take林紀旭 James Lin

The numbers here describe two overlapping financing layers rather than one deal: a $100 billion, 10 GW headline commitment from September 2025, and a more granular August 2026 structure in which NVIDIA backstops 4.25 IT-GW of build-out directly and holds an option on another 3.75 IT-GW. That option is the detail worth tracking — it lets NVIDIA scale its exposure to the Ohio campus in step with actual demand rather than committing to the full 8 IT-GW upfront. The real test of whether the $600 billion 2030 compute opportunity NVIDIA cites is credible will be whether the first 800 MW tranche actually lands in 2028 on schedule, using AEP's existing grid, before the harder second phase — new power plants, new transmission, and the $4.2 billion grid build SB Energy and SoftBank have committed to — has to deliver the remaining gigawatts.

林紀旭 James LinEditor-in-Chief

Related

BRIEF

NVIDIA Partners Bring Physical AI and Smart Automation to Automation Taipei 2026

NVIDIA's partner network — including Advantech, FANUC, Solomon, Techman Robot, ITRI, Delta, Universal Robots, and Spingence — is showing physical AI and edge robotics at Automation Taipei 2026, held at Taipei Nangang Exhibition Center Halls 1 and 2. Jetson Thor, Jetson AGX Thor, Jetson Orin, Isaac Sim, Isaac Lab, and Omniverse are powering humanoid robots, AMRs, quadrupeds, and digital-twin factories across the show floor.

林紀旭 James Lin ·
BRIEF

NVIDIA's AI Moat Shifts From Chips to Capital as Huang Backs OpenAI With Up to $105 Billion

NVIDIA (輝達) has pledged up to $105 billion in funding for OpenAI's Ohio data center and is separately pursuing a $500 billion chip-financing agreement with Wall Street firms. CNBC frames the moves as NVIDIA's AI moat shifting from chips to capital, a reading backed by NVIDIA's own numbers: quarterly free cash flow up 18-fold to $48.5 billion and marketable equity holdings up from $12.9 billion to $30.2 billion.

EffectStory 編輯部 ·
BRIEF

Nanya Technology Says It Is Working With Regulators to Evaluate New Fab Sites Amid Yunlin 12-Inch Plant Rumors

Nanya Technology (南亞科) has not confirmed reports of new 12-inch fabs in Yunlin and Pingtung, saying only that it continues to work with regulators on potential fab-site evaluation and will announce any concrete decision per disclosure rules. The rumor coincided with a record intraday stock price of NT$542 and a separate trade-secret indictment involving a company engineer.

EffectStory 編輯部 ·