Oracle (NYSE: ORCL) stunned the markets on September 10, 2025: after results published the day before, the group reported $14.9 billion in revenue, a booming cloud, and an order book at $455 billion. Larry Ellison and Safra Catz promise giant AI contracts and an acceleration of data centers. The financial, competitive, and ecological costs remain to be measured.
Results and RPO: the number that changes everything
On September 9, 2025, Oracle published solid results: $14.9 billion in revenue and an adjusted EPS of $1.47. [EPS = earnings per share, an indicator that relates net income to the number of shares outstanding]
But the number that disrupts expectations is elsewhere: an order book raised to $455 billion. The group mentions multi-year contracts concluded with three major AI clients, driven by increased demand for computing power.
These amounts, rare in enterprise tech, do not yet indicate how quickly they will turn into revenue. Oracle promises a rapid ramp-up of its cloud infrastructure by 2030, a timeline dependent on chips, energy, and local authorizations.
Larry Ellison and Safra Catz, the architecture of power
The group remains shaped by its leading duo: Larry Ellison, co-founder and chairman of the board (also CTO), strategist of the AI shift, and Safra Catz, CEO known for her financial discipline and execution. Ellison sets the vision "all for AI infrastructure," Catz closes contracts and directs investments. Their complementarity, proven for over twenty years, illuminates Oracle’s reorientation: from a database champion to a provider of AI computing capacity.
Ellison cultivates a provocative style, technological bets when Safra Catz, a former investment banker, is known for being austere and obsessed with costs. Two temperaments, one obsession: execute quickly.
"We should be interconnected with everyone." Larry Ellison, about multi-cloud.
"AI is a driver of cloud adoption." Safra Catz.
"Mega-contracts" still opaque: caution
The centerpiece is called RPO: these are firm revenue commitments spread over several years. Oracle indicates having closed four giant contracts with three players in the sector, all eager for GPUs and energy. Amid rumors, some media talk of a massive agreement with OpenAI. This agreement would reach several hundred billion dollars over a five-year period. But these spectacular figures have never been publicly confirmed, caution is advised. No official confirmation is given by Oracle, the group merely highlights the unprecedented size of the commitments and their multi-cloud nature. This caution is necessary: the scale of these amounts would defy historical references in enterprise cloud and concentrates, by itself, part of the risk of over-interpretation.
The multi-cloud bet to accelerate
Rather than exclusivity, Oracle multiplies bridges with its rivals. The goal: to allow companies already with AWS, Microsoft, or Google to add Oracle services without migrating everything. This flexibility reduces the fear of lock-in and accelerates the signing of AI-specific contracts. In these contracts, available capacity is crucial.
The bet is aggressive but understandable: selling a place in the clients’ architecture, even if it means coexisting with better-established competitors.
AWS, Microsoft, Google: the response
The historical leaders retain the bulk of the market share. But Oracle’s offensive on deliverable AI capacity puts pressure: the battle is no longer just about the list price, but about access to GPUs, electricity, and contractual guarantees. All elements that, for clients, now weigh as much as technical performance.
Oracle on the Stock Exchange: an unprecedented leap in 30 years
On September 10, 2025, the Oracle (NYSE: ORCL) stock price gained more than 35% in the session, one of its strongest increases since 1992, bringing the market capitalization close to $1 trillion. Larry Ellison’s fortune briefly surpassed or matched that of Elon Musk, rankings that are volatile by nature.
The energy cost of a race for GPUs
Powering computing farms requires megawatts and a lot of water for cooling. Location decisions, the energy mix, and heat reuse become criteria of cost as much as acceptability. Oracle promises optimizations, but the equation remains physical.
Regulation: interoperability and "reversibility"
In Europe, the Data Act imposes rules for changing providers ("reversibility"), limits certain exit fees, and promotes interoperability of cloud services. In the United Kingdom, the competition authority has extended its review of the cloud market, pointing out barriers to multi-cloud. This regulatory undercurrent aims to rebalance the power dynamics between providers and clients. For Oracle, which positions itself as a multi-cloud artisan, this can be a competitive advantage provided it guarantees data portability and contractual protection for clients over time.
Geopolitics of data centers
Regions offering decarbonized electricity, land, and reliable supply chains will attract new campuses. Elsewhere, projects risk slowing down or becoming more expensive. Europe is betting on data portability and contract reversibility, a framework that can benefit those who truly practice multi-cloud.
International angle: China, Japan, and the geopolitics of capacity
China. The market relies on Alibaba Cloud and Tencent Cloud; U.S. controls on AI chips limit access to advanced GPUs. American-style mega-contracts remain unlikely in the short term.
Japan. Tokyo subsidizes new data centers and secures GPUs, local demand intensifies around generative AI. With its Tokyo and Osaka regions, OCI can capture this growth via multi-cloud.
LightOn, the other tempo of European AI
LightOn (Euronext Growth Paris) moves at a different pace. On September 9, 2025, the company announced a half-year revenue of €710,000 (+15% year-on-year) and an ARR of €1.23 million at the end of June. However, the management has revised down its 2025 target to €3–4 million, citing long sales cycles for its Paradigm platform. The stock lost about 12% the next day.
The snapshot says something about the market: while hyperscalers are snapping up multi-year contracts for computing capacity, medium-sized European publishers are building, more slowly, recurring revenue, client by client. A two-speed scenario that questions these players’ ability to finance infrastructure and connect to major ecosystems.
What could still go wrong
The amounts and beneficiaries of the "mega-contracts" remain unknown: Oracle does not publish the details, and the figures circulating have not been confirmed. Caution is advised.
The pace will depend on material elements: availability of GPUs, access to electricity, and obtaining local authorizations. These will set, beyond promises, the real pace of the projects.
The quality of the order book remains to be clarified: what part is legally committed and executable, what is the average duration, and what is the sensitivity to exit clauses?
What actors must anticipate
For companies, the challenge is to demand reversibility clauses and readable SLAs, while designing portable architectures.
For public authorities, it is about ensuring interoperability and an effective multi-cloud, arbitrating between hosting data centers and environmental constraints.
For investors, the key question remains the conversion of RPO into cash flow, investment discipline, and the sustainability of energy costs.
Beyond the session, the real test
The stock market euphoria does not tell the whole story. The challenge will be industrial: project timelines, available megawatts, and securing GPUs. If Oracle keeps its promise, the entire digital economy will have to contend with an energy scarcity that has become a strategic parameter.