Oracle Expands AI Cloud with Nvidia, Google Deals

Oracle (ORCL) deepened its AI infrastructure push on Tuesday, expanding its Nvidia chip commitment to "multi-gigawatt scale" while simultaneously adding Google TPUs and inference-focused chips from Cerebras Systems to its cloud catalogue.

The moves signal that Oracle is repositioning its cloud unit as a hardware-agnostic AI platform - a strategy that could broaden its addressable customer base and ease reliance on a single chip supplier ahead of the company's investor day later this week.

Key Takeaways

• Oracle scales Nvidia Vera Rubin server capacity to multi-gigawatt levels.
• Google TPUs and Cerebras inference chips added to Oracle Cloud for first time.
• Oracle's own Acceleron networking chips will replace some Nvidia networking silicon.

Market Context & Deal Scale

The Nvidia (NVDA) relationship remains Oracle's anchor. Nvidia CEO Jensen Huang has previously said Vera Rubin systems can generate up to $40 billion in revenue per gigawatt for the chipmaker, giving the multi-gigawatt commitment significant implied value - though neither company disclosed contract pricing 1.

For context, Oracle said during its most recent earnings call that it is maintaining capital expenditure guidance of roughly $90 billion to $95 billion for fiscal year 2027, with customers increasingly co-financing chip purchases. That customer-funded model could limit near-term balance-sheet pressure even as the infrastructure buildout accelerates. Investors tracking the broader AI infrastructure financing landscape may find parallels in Blackstone's $63 billion AI infrastructure deal, which similarly relies on blended financing structures.

Detailed Analysis: The Diversification Play

The headline expansion of Nvidia capacity is paired with a deliberate diversification of the chip stack. Oracle said it will offer Google's tensor processing units (TPUs) through its cloud platform for the first time, giving enterprise customers access to Alphabet's (GOOGL) proprietary AI accelerators without leaving Oracle's infrastructure 1.

Beyond Google, Oracle is adding chips from Cerebras Systems and Intel-backed SambaNova Systems - both specialists in AI inference, the computationally distinct workload that occurs when a trained model responds to user queries. Nvidia dominates AI training, but inference is a more fragmented and price-sensitive segment where challengers like Cerebras have found traction.

Notably, even the Nvidia-based deployments will incorporate a degree of Oracle independence: the company said its in-house Acceleron networking chips will substitute for some of Nvidia's own networking silicon within those clusters. That decision suggests Oracle is engineering margin and differentiation into its infrastructure layer, a dynamic that could become increasingly significant as Oracle Cloud Infrastructure scales its AI compute footprint.

Management Quotes & Outlook

"Different AI workloads have different compute needs. That choice gives [customers] the flexibility to choose the right infrastructure for the performance, throughput and economics of each AI workload."

- Mahesh Thiagarajan, President, Oracle Cloud Infrastructure 1

Cerebras CEO Andrew Feldman framed the Oracle partnership as a pivotal distribution milestone for his company. "That's a massive expansion of our customer base of tier-one customers," Feldman said in an interview with Reuters. "We're being deployed in other people's data centers at large scale." 1

Oracle did not specify how it intends to finance the expanded chip purchases beyond reiterating that customers are increasingly funding their own deployments. The company is expected to provide more detail at its investor day event this week, which may clarify capital allocation priorities for shareholders monitoring the $90 billion-plus capex programme.

Conclusion

Oracle's multi-chip strategy reflects a broader industry recognition that no single semiconductor architecture will serve every AI use case. By pairing Nvidia's leading training hardware with Google TPUs and inference-optimised chips from Cerebras and SambaNova - while inserting its own Acceleron networking silicon - Oracle is attempting to control more of the AI stack and reduce vendor concentration risk.

Whether the strategy translates to improved cloud revenue growth and margin expansion will likely be the central question at this week's investor day. The financial terms of Tuesday's announcements remain undisclosed, leaving valuation impact largely a function of execution.

Not investment advice. For informational purposes only.