Oracle’s $50 Billion Data Center Push: How Fuel Cells and AI Agents Aim to Sustain an 84% Cloud Surge

Oracle builds a massive AI data center campus with Bloom Energy fuel cells, bypassing grid constraints, as cloud revenue surges 84% and capex hits $50B.

Oracle’s infrastructure ambitions are taking a decidedly industrial turn. The software giant is building a massive AI data center campus in New Mexico under the code name “Project Jupiter,” and it plans to power the facility with Bloom Energy fuel cells arranged in a dedicated microgrid. The move is designed to bypass traditional grid constraints — a growing bottleneck for hyperscale cloud deployments — and reflects the lengths Oracle is willing to go to capture the surging demand for its cloud computing capacity.

The urgency is underscored by a stunning quarterly update. In the third fiscal quarter, Oracle’s cloud infrastructure revenue surged 84% year over year, pushing total sales to $17.2 billion, a 22% gain that comfortably beat analyst estimates. Management disclosed that demand for its proprietary cloud infrastructure now runs at twice the available supply, leaving little room for error as the company races to build out capacity. That effort is backed by a record order book: remaining performance obligations stood at $553 billion at the end of the period, providing a multi-year runway of contracted revenue.

Yet the stock has taken a breather after a torrid rally. In Frankfurt, Oracle shares closed at €164.00 on Monday, down 1.3% on the day, but that masks a 38% monthly gain. The recent six-day winning streak had pushed the stock into expensive territory: the forward price-to-earnings multiple now sits near 35, well above its long-term average. Chief among the concerns is the sheer scale of capital required. Oracle has earmarked roughly $50 billion in capital expenditures for fiscal 2026, more than double the prior year, and plans to raise $45 billion to $50 billion through capital markets to help fund the buildout. Long-term debt already stands at $125 billion, and free cash flow turned negative last quarter due to the investment cycle.

That has prompted a measured response from large investors. Some asset managers have been raising their stakes, but insider sales have accumulated to around $2.6 million in recent months, suggesting that even executives are taking some profits off the table. The company’s high valuation leaves little margin for error: the current share price assumes that breakneck growth can be sustained, and any stutter could invite a sharp re-rating.

The biggest wild card remains OpenAI. Reports that the AI pioneer has missed both user and revenue targets recently knocked Oracle shares down 4.1% in a single week, highlighting the market’s sensitivity to the health of key cloud tenants. The partnership between the two companies is a complex web of capacity reservations, financing arrangements, and forward demand commitments, and any sign of weakness from OpenAI raises questions about the stability of that revenue pipeline. Analysts at firms such as Wedbush and Guggenheim remain constructive, noting opportunities in the multicloud and GPU-as-a-service markets, but the specter of a single large customer faltering weighs on sentiment.

To diversify both its use cases and its energy sources, Oracle is also pushing AI deeper into its own software. The company recently integrated Eightfold AI’s “agentic interviewing” features into Oracle Fusion Cloud Recruiting, allowing automated agents to handle initial candidate screenings. It is a small but telling example of where Oracle sees its enterprise software heading: away from form-based workflows and toward autonomous, AI-driven processes that can command higher subscription fees over time.

Oracle reports fiscal fourth-quarter results in June, with Wall Street projecting earnings of roughly $1.96 to $2.00 per share. The consensus price target stands at around $259, implying roughly 50% upside from current levels — but that depends on the company’s ability to convert its massive backlog into free cash flow. If the multi-year contract commitments begin translating into cash faster than expected, the AI infrastructure thesis gains real substance. If the negative free cash flow persists, the funding gap will remain the most stubborn obstacle to the growth story.