Nvidia Invests $1.5 Billion, Offers Up to $105 Billion in Credit Support for Ohio AI Center

01Nvidia invests $1.5 billion in SB Energy and offers up to $105 billion in credit support for Ohio AI data center

Nvidia said Monday it will invest $1.5 billion in SB Energy, the data center and power developer behind Ports-Pike, an OpenAI-related project near Cincinnati, Ohio. The arrangement also makes Nvidia the sole supplier of the project’s compute infrastructure—the systems used to run AI workloads—and commits the chipmaker to provide as much as $105 billion in credit support for construction.

Ports-Pike is planned for land owned by the U.S. Department of Energy that was previously used to enrich uranium for the U.S. nuclear arsenal and Navy submarines. SB Energy’s existing investors include SoftBank and OpenAI, giving Nvidia a direct financial and supply role in a large project already tied to two major participants in the AI industry.

The three parts of Nvidia’s involvement reinforce one another. Its $1.5 billion investment gives SB Energy new capital, the credit support could help the developer finance construction, and the sole-supplier agreement reserves the data center’s computing demand for Nvidia. The source does not disclose the credit terms or explain how much of the maximum support will ultimately be used.

According to documents Nvidia filed with the SEC, Ports-Pike is expected to begin at 4.25 gigawatts and could expand to 8 gigawatts. The eventual scale remains uncertain because no completion date or final expansion commitment was reported.

Supplying power on that scale requires a separate energy build-out. SB Energy plans to construct a 9.2-gigawatt natural gas power plant at the site, with an expected cost of $33 billion. That generating capacity would exceed the data center’s stated maximum scale, though the source does not detail how the balance would be used.

Construction costs are already a significant pressure. BloombergNEF data cited in the report shows that the cost of building natural gas power plants has risen 66% over the past two years. Once this plant and other projects are completed, they could also compete with export markets for gas, a combination that could triple natural gas prices in some parts of the country. That is a forecasted risk, not a reported price increase caused by Ports-Pike.

The project’s financing terms, completion schedule, final size and effect on local natural gas prices remain undisclosed.

02Expert for 3M Used ChatGPT to Draft Explosion Report, Exposing 350 Pages of Chats in Discovery

An expert witness hired by 3M used ChatGPT to draft significant portions of a report defending the company in litigation over a deadly Houston factory explosion. The January 2020 blast at Watson Grinding killed three people and destroyed roughly 200 homes; a federal investigation attributed it to flammable gas leaking from a degraded, poorly crimped rubber welding hose.

Dozens of homeowners sued 3M and Watson Grinding, alleging that 3M failed to service the factory’s gas-detection system properly and made other errors that contributed to the explosion. 3M retained Josh Autenrieth of Knighthawk Engineering to prepare an expert report addressing its conduct.

The extent of Autenrieth’s AI use emerged through discovery, the pretrial process in which opposing parties exchange relevant evidence. A plaintiffs’ lawyer, Will Moye, found a five-page “Citation Overlay” document that appeared to have been generated by AI and demanded all prompts Autenrieth had used. His deposition was paused for three hours while the material was assembled. The plaintiffs then received 350 pages of ChatGPT conversations, including public links to the full chats.

Those conversations showed Autenrieth asking ChatGPT to create an “exceptional expert witness report” defending 3M’s standard of care. After uploading hundreds of court records, he instructed the model to counter another witness’s claims and “show how 3M is 0% at fault” for the explosion.

ChatGPT produced a roughly 30-page draft stating that 3M was “0% responsible.” That sentence was removed after Autenrieth asked the model to review the report from opposing counsel’s perspective. ChatGPT warned that the absolute claim was an easy target and could make him appear to be an advocate instead of an independent expert.

The final court submission nevertheless retained the structure of ChatGPT’s initial draft, and large sections matched the model’s original output or revisions it later suggested. Autenrieth also asked ChatGPT to grade the report, identify likely attacks and help formulate defenses.

Moye alleges that Autenrieth relied on ChatGPT to form his opinions and write the report, rather than using it merely as an aid. The disclosed prompts now underpin a dispute over the report’s credibility, but the source does not say how the court ultimately evaluated the report or determine 3M’s liability.

03Tracker Leads Rare-Book Order to Amazon AI Facility, Where Employees Say Bindings Are Cut Off for Scanning

Booksellers have suspected for about a year that AI companies are buying rare books in bulk, scanning them and destroying the physical copies. Such books can contain text that is difficult to find online and potentially valuable for training frontier AI models, whose developers generally keep their training data secret. Until now, however, sellers had struggled to identify the buyers behind those orders.

A 404 Media investigation has now linked one rare-book shipment to Amazon. Working with a bookseller, the outlet placed an AirTag—a small location-tracking device—inside a rare book that was part of a bulk order and expected to be acquired by an AI company. The tracker followed the shipment across the country to an Amazon warehouse in Las Vegas, Nevada. The evidence establishes Amazon as the destination of this order, but does not show that every suspected bulk purchase of rare books came from the company. 404 Media

Employees at the Las Vegas site told 404 Media that their work consists of receiving large shipments of printed books, removing the bindings and scanning the pages. Cutting off the bindings makes scanning faster but destroys the physical books in the process.

404 Media described the warehouse as an Amazon AI training facility. The team operating there is called VGT3, and its logo depicts a dinosaur holding a book. Together, the tracked shipment, the workers’ account of large-scale destructive scanning and the facility’s identification provide the investigation’s evidence that the books are being processed for AI training.

Amazon did not specifically address the investigation’s findings. In a statement, the company confirmed that it buys books through commercial channels “to help develop and improve” products and services used by customers, but it did not explicitly say that the books are used to train AI. The available sources also do not identify which Amazon models or products might use the scanned text, how many books the operation has processed, or when the practice began. Ars Technica