counter easy hit

Oracle brings ChatGPT and Codex into its apps and opens cloud to new AI chips

Oracle brings ChatGPT and Codex into its apps and opens cloud to new AI chips
5

Oracle is embedding OpenAI’s ChatGPT and Codex across its application stack while adding Google’s TPUs, OpenAI’s Jalapeño chip and Cerebras hardware to Oracle Cloud Infrastructure. The announcements land at Oracle AI World in Las Vegas, where OpenAI CEO Sam Altman is set to share a keynote stage with Oracle co-CEO Clay Magouyrk.

Oracle on Tuesday folded OpenAI’s ChatGPT and Codex into its application stack and said its cloud will run Google’s tensor processing units, OpenAI’s Jalapeño chip and Cerebras hardware. The company made both announcements at Oracle AI World in Las Vegas.

The timing carries weight for OpenAI. Anthropic filed its S-1 the day before, putting the frontier lab’s chief rival on a path toward an initial public offering. OpenAI’s Oracle deal reads as its own move to reach corporate buyers who already run Oracle databases in healthcare and finance.

Mahesh Thiagarajan, Oracle’s president of Cloud Infrastructure, told CNBC at the conference that ChatGPT “allows anybody to build an application, deploy it, manage it.” Greg Brockman, OpenAI’s president and co-founder, said in the release that “more businesses are choosing OpenAI as we make it easier to put AI to work through the tools and systems they already use.”

Altman joins Magouyrk on the keynote stage Tuesday at 12:40 p.m. ET.

Thiagarajan said his team watched OpenAI’s GPT-6 Astra model train at the Stargate campus in Abilene, Texas, an Oracle-backed data center cluster, and saw the results feed back in real time. Oracle runs the databases that log and secure records for hospitals, banks and insurers, which puts its systems next to the models companies deploy.

The hardware news cuts deeper. Oracle Cloud Infrastructure will add Google Cloud TPUs, chips Google designed for its own AI workloads, alongside OpenAI’s Jalapeño processor and Cerebras accelerators. Google’s custom silicon has never run on another hyperscaler’s cloud before, so the arrangement marks a first for Google as well as Oracle.

Chip makers and cloud providers both want their hardware in the hands of frontier labs. Google gets its TPUs in front of customers who buy compute from Oracle instead of Google Cloud. OpenAI gets silicon it controls into a major provider’s data centers. Cerebras gets distribution beyond the deals it already holds.

The Cerebras agreement is material to the company, though a Cerebras spokesperson said the firm will not disclose the scale because it sits in a quiet period. Cerebras shares climbed about 8% on the news. Oracle shares rose nearly 5%.

Seven months ago, Amazon Web Services, the largest cloud provider, said it would bring Cerebras hardware into its data centers. The Oracle deal extends that reach.

No exit from Nvidia

Thiagarajan said customers still face a shortage of chips, and that giving buyers more accelerators to choose from helps them work around it. He rejected the idea that Oracle plans to move away from Nvidia GPUs, calling Nvidia’s parts “the best, most fungible” option and pointing to the size of the AI market.

That framing matters for investors. Nvidia remains the default choice for training large models, and Oracle’s new lineup supplements that supply rather than replaces it. Customers mixing chip types also gain leverage when negotiating prices with any single vendor.

Project Jupiter and local opposition

Thiagarajan also addressed Project Oracle’s multibillion-dollar New Mexico data center in Sunland Park, where the project has broken ground despite pushback from residents. He described Jupiter as one of 30 to 40 data center projects the company develops each quarter.

“We’re actually not too worried about what we’re doing because I think we’re building global infrastructure,” he said.

A hearing on air quality permitting is scheduled for Dec. 7 in Sunland Park. The political resistance led Oracle to send a force majeure notice last month to backer Blue Owl Capital. Oracle says the project stays on schedule.

What investors want next

Attention now shifts to Oracle’s investor day. Buyers will look for evidence that the company can hit its fiscal 2030 revenue targets and for a read on how rising capital costs have changed the margins on its cloud contracts. Bank of America analysts told clients they expect updates on the biggest data center projects among the topics management will address.

Oracle shares are down more than 30% this year, a decline tied in part to questions about how the company finances its data center pipeline. The OpenAI expansion and the new chip lineup answer a product question. The investor day answers the funding question.

Leave A Reply

Your email address will not be published.