Trending
DCD Studio: The physical constraints of AI scaling with Sean James, Nvidia Oracle Cloud to Offer Google TPUs and New AI Chips Tata Communications to invest $450m into 10,000km of AI-ready fiber to support data center connectivity in India Mike Sievert steps down from T-Mobile US board CenTrak User Conference 2026: Staff Safety and BLE are Key for RTLS Adoption Vera Health Raises $28.5M for Clinical AI Search Engine: Surpasses 800,000 Global Physician Users La Caisse snaps up 49.9% stake in Bell Canada’s tower unit KBRA assigns preliminary ratings to up to $525m in US Signal notes, secured against ten data centers Oracle will offer Google’s TPU AI chips via OCI in multi-gigawatt deployment A taxonomy of American data center opposition: The neighbors US Signal acquires data center in Hillsboro, Oregon Patient CRM Modernization: Four Decisions Technology Can’t Make for You Neocloud General Compute raises $100m in Series A funding round Readiness checklist & project plan workbook UK budget: Grid connection incentives for data centers under new “covenant”

Oracle AI World takeaways: making Nvidia envious, and storage… sexy?

Who could have predicted that the AI cloud sphere would more and more resemble a combination of various chips? Following a day at Oracle AI World’s Las Vegas event that mainly revolved around applications, day two brought an influx of chip news. Although Nvidia remains a dominant force, Oracle is broadening its horizons by working with other chip companies and moving beyond GPU architecture.

The most thrilling aspect for us is Oracle’s plan to offer access to Google’s custom TPUs (tensor processing units), making it the first hyperscaler to provide these chips outside of Google Cloud Platform. This is exciting because It’s somewhat peculiar, considering the situation became strange when Google decided to stop using TPUs as their hidden advantage and decided to provide them to a chosen group of clients.

They later revealed a partnership with Blackstone to establish another cloud platform powered by TPUs. It’s reasonable to believe that a) there’s an enormous demand for computational services (which we are aware of), and b) Google has an excess of TPUs that they don’t have the facilities or power capacity prepared for, preventing a valuable resource from remaining unused.

This information is dated October 2026. DCD discussed the partnership with Oracle, where Karan Batta, SVP of Oracle Cloud Infrastructure, stated: “This is not an experimental endeavor.” Oracle will make a significant cloud investment at a gigawatt scale over time. It’s best to view it as Google having our new TPU available from the start when they launch it.” In terms of commitment to deployment, this experiment is not significant, particularly when you agree to buy gigawatts of chips over multiple years.

However, there is a level of experimentation that contributes to the overall strategy. Batta states that in the present scenario, Oracle functions in a “supply-constrained environment, while costs are escalating, including memory and SSDs for storage.” Consequently, the company believes that providing people with greater flexibility and options leads to better decision-making.

Considering this, Oracle is in discussions about the manner in which significant clients procure capacity in regard to energy, committing to 100MW or potentially 1GW of capacity, several years ahead. “However, in three years, we cannot predict what the supply situation will be like.” It remains uncertain how the market will behave concerning pricing.

 

Join the conversation

Your email address will not be published. Required fields are marked *