Larry Ellison, chairman of Oracle Corporation, speaking with President Trump, SoftBank Group CEO Masayoshi Son, and OpenAI CEO Sam Altman, announcing Stargate at the White House in January (Jabin Botsford/Getty Images)

OpenAI, Oracle, and SoftBank announce five new AI data center sites, putting Stargate ahead of schedule

Two of the sites will be in Texas, one in New Mexico, one in Ohio, and one in the Midwest.

David Crowther

9/24/25 4:38AM

Since Oracle’s $300 billion deal with OpenAI grabbed the headlines a few weeks ago — sending Oracle’s stock soaring and briefly making its chairman the world’s richest person — some of the biggest questions involved what the cloud computing company would do to deliver all the capacity it had promised, and how that would be financed.

Now, we’re getting more specifics on these fronts: OpenAI, Oracle, and SoftBank announced late last night that they would be expanding Stargate with five new AI data center sites. Separately, Bloomberg is reporting that Oracle is looking to tap the US corporate bond market for $15 billion today, which includes a 40-year offering.

The data center build-out plans put the much-hyped Stargate initiative, announced by President Trump and the companies’ leaders back in January, ahead of schedule. Per a press release from OpenAI:

“The combined capacity from these five new sites — along with our flagship site in Abilene, Texas, and ongoing projects with CoreWeave — brings Stargate to nearly 7 gigawatts of planned capacity and over $400 billion in investment over the next three years. This puts us on a clear path to securing the full $500 billion, 10-gigawatt commitment we announced in January by the end of 2025, ahead of schedule.”

Three of the sites appear to fall under the $300 billion OpenAI and Oracle deal signed in July. Those three new sites are to be located in Shackelford County, Texas; Doña Ana County, New Mexico; and an unannounced site somewhere in the Midwest. The Stargate flagship site, in Abilene, Texas, is also being mooted for a potential expansion of 600 megawatts of capacity. Together, the Abilene upgrade plus the three new sites can deliver over 5.5 gigawatts of capacity, according to OpenAI, and will create over 25,000 on-site jobs.

The other two newly announced sites will be located in Lordstown, Ohio, and Milam County, Texas, and are being developed through a partnership between SoftBank and OpenAI, with the sites able to scale up to 1.5 gigawatts of capacity, per OpenAI’s statement.

The data center in Abilene, Texas, is reportedly already up and running on Oracle’s Cloud Infrastructure, and the first Nvidia GB200 racks were delivered to the site in June.

As of 5 a.m. ET on Wednesday, Oracle’s stock was 0.4% higher. Nvidia, which jumped on Monday as investors cheered news of its bumper ~$100 billion OpenAI investment but dipped yesterday, has also been trading 0.7% higher in Wednesday’s premarket, as the company deepens its vertical integration across the AI supply chain.

Go Deeper: Bank of America explains why Nvidia almost has to invest in OpenAI and Intel

Jon Keegan6h

OpenAI shares how it will charge for ChatGPT ads

Last week, OpenAI announced that ads were going to be rolling out in ChatGPT in the coming weeks.

Now we have more details about what OpenAI is telling advertisers. According to a report from The Information, the company has reached out to “dozens” of advertisers, and will charge based on ad views.

Advertisers are still waiting for further details, but OpenAI is asking for less than $1 million each in ad spending while it tests out the new system, per the report.

Ads are supposed to begin in February, and will only appear for free ChatGPT and ChatGPT Go users.

OpenAI Lines Up Advertisers, Reveals Key Details Ahead of Ads Launch

Advertisers are still waiting for further details, but OpenAI is asking for less than $1 million each in ad spending while it tests out the new system, per the report.

Ads are supposed to begin in February, and will only appear for free ChatGPT and ChatGPT Go users.

Rani Molla7h

Apple is reportedly working on a wearable AI pin

Move over OpenAI, Apple is reportedly also developing a mysterious AI-powered wearable device: a pin that looks like a “thin, flat, circular disc with an aluminum-and-glass shell.”

The Information reports that the device is the size of an Apple AirTag and has two cameras, a speaker, three microphones, and wireless charging. It could be available by early 2027.

Apple, which has lagged its peers in AI and recently teamed up with Google to support its upcoming Siri revamp, is hoping to keep up with ChatGPT and Google, which, like Apple, has an AI smartphone. Meta and Google are both also pushing into smart AI glasses.

It’s not to be mistaken with OpenAI’s secretive wearable AI device, which is being made in conjunction with former Apple designer Jony Ive and expected to debut in late 2026. The latest rumors suggest the unnamed device, meant to eventually compete with smartphones, might be earbuds.

Apple Developing AI Wearable Pin

Jon Keegan

10h

As public pushback against data centers grows, AI companies promise they’ll be good neighbors

Communities concerned with energy prices, pollution, and water usage are starting to push back on Big Tech’s AI data center building spree. The companies are trying to assure the public that they will pay their way.

Rani Molla12h

Morgan Stanley expects Tesla to have 1,000 Robotaxis by the end of 2026. Musk had predicted 1,500 by the end of 2025

Ahead of Tesla’s earnings report next week, Morgan Stanley has released a note estimating that the company will scale its Robotaxi fleet much more slowly than CEO Elon Musk has said. The firm thinks the automaker will have 1,000 vehicles in its Robotaxi service by the end of 2026 — 500 fewer than Musk estimated a few months ago Tesla would have by the end of 2025.

More key to Tesla’s success, however, will be removing the safety monitors from those rides, which Morgan Stanley says will be a “precursor to personal unsupervised FSD [Full Self-Driving] rollout.” Musk, of course, had also promised to remove safety drivers in Austin by the end of 2025, but driverless rides are still in the testing stage.

Rani Molla14h

Meta says it’s delivered new AI models internally this month and they’re “very good”

Meta’s last AI model release, Llama 4, was marred by delays and accusations of rigged benchmarks, but the company says the latest models built by its Superintelligence Labs team look promising. CTO Andrew Bosworth told reporters at the World Economic Forum that the team delivered new models internally in January and they’re “very good.”

Bosworth didn’t specify what the models are, though The Wall Street Journal has reported that Meta is working on a large language model and an AI image and video model code-named Avocado and Mango, respectively.