이미지: TechCrunch AI
Summary
- NVIDIA is investing $1.5 billion in SB Energy, a data center developer linked to SoftBank and OpenAI
- The investment secures NVIDIA's position as the sole computing infrastructure supplier for OpenAI's data center near Cincinnati, Ohio
- SB Energy plans to build a 9.2-gigawatt natural gas power plant on the site at a cost of $33 billion
- 엔비디아 투자액
- 15억달러
- 엔비디아 신용공여 한도
- 최대 1050억달러
- 데이터센터 규모
- 초기 4.25기가와트 → 최대 8기가와트
- 천연가스 발전소 규모
- 9.2기가와트
- 발전소 예상 비용
- 330억달러
- 천연가스 발전소 건설비 상승률
- 최근 2년간 66% (블룸버그NEF)
- 소프트뱅크의 엔비디아 주식 매각액
- 58억달러 (지난해 11월)
What $1.5 billion bought: exclusive supply rights
NVIDIA announced on the 17th (local time) that it will invest $1.5 billion in SB Energy, a data center developer connected to SoftBank and OpenAI. What stands out isn't the size of the investment itself, but what NVIDIA got in return. According to NVIDIA's announcement, the deal makes NVIDIA the sole supplier of computing infrastructure for OpenAI's "Ports-Pike" data center being built near Cincinnati, Ohio. On top of that, NVIDIA agreed to provide a credit facility of up to $105 billion to help fund construction of the facility.
The Ports-Pike data center is an extension of the story METAL LAB reported in early August: NVIDIA guarantees Ohio site for 20 years, commits 4.25GW to OpenAI. At the time, NVIDIA pledged to supply 4.25 gigawatts of computing infrastructure to the site over 20 years. This latest investment in SB Energy can be seen as the step that locks in the funding and equity structure needed to make good on that pledge. According to SEC filings, the facility could expand from an initial 4.25 gigawatts to as much as 8 gigawatts.
What is SB Energy
SB Energy is a data center and power developer whose existing investors include SoftBank and OpenAI. SoftBank previously held $5.8 billion worth of NVIDIA stock but sold its entire stake last November to raise funds for other AI investments. In effect, the cash raised from selling NVIDIA shares is now flowing back into a data center developer in which NVIDIA itself is taking a stake.
For OpenAI, this is another example of the company continuing to lease computing resources from outside partners — Microsoft, Amazon, Google, and NVIDIA among them. OpenAI builds the models itself, but the data centers and chip supply chains that run those models are increasingly built through equity and credit arrangements with partner companies, as seen here.
A gas power plant on a former uranium enrichment site
The land where the facility will be built is owned by the U.S. Department of Energy (DOE) and was previously used to enrich uranium for U.S. nuclear weapons and naval submarines. SB Energy plans to build a 9.2-gigawatt natural gas power plant on the site to supply power to the data center, at an estimated construction cost of $33 billion.
Part of why that cost is so high is a broader surge in natural gas power plant construction costs. According to BloombergNEF, construction costs for natural gas plants have jumped 66% over the past two years. On top of that, by the time SB Energy's plant and several others come online, they will be competing with export markets for natural gas — a combination some forecasts suggest could push natural gas prices up to three times higher in certain regions.
Editor's take
In one sentence: NVIDIA is now funding the power supply chain of its own customers. A company once described as merely "selling picks and shovels" without joining the chatbot race has moved on to building the mine itself. The gap between the $1.5 billion investment and the $105 billion credit facility alone makes clear that the real goal of this deal isn't equity returns — it's locking in exclusive supply rights. Wherever OpenAI sources its computing power, this structure ensures that NVIDIA's chips and NVIDIA's money are both baked into the deal.
The shift becomes clearer when you compare it to just a few years ago. Back then, AI infrastructure investment was a simple picture: OpenAI renting Microsoft's cloud. Now, chipmakers are investing directly in data center developers and even co-building natural gas power plants to make up for power shortages. The competition among AI models has spilled over into competition over power plants, transmission lines, and land.
For domestic companies, the lesson here is clear: focusing only on the race to secure GPUs means falling behind. Bottlenecks in power and land — and regulation — are already surfacing even in the U.S., and the fact that a former uranium enrichment site had to be repurposed shows just how scarce suitable land has become. This signals that the next constraint on data center expansion will be the power grid. Teams preparing AI infrastructure domestically need to build out power procurement plans alongside their GPU procurement plans.
More announcements with a similar structure are likely to follow in the coming weeks. This isn't the first time NVIDIA has taken a direct equity stake in a data center developer, and as long as the power bottleneck remains unresolved, these kinds of deals — chip companies investing directly in power infrastructure — are only likely to increase.
