Oracle's Net Profit Reaches $17 Billion with $48 Billion in Concurrent Financing
Oracle achieved a net profit of $17 billion in the last fiscal year, while raising $48 billion through $43 billion in bond issuance and $5 billion in equity financing, aimed at building AI cloud infrastructure for clients such as OpenAI, Meta, and xAI.
As of May 31, the company reported revenues of $67.4 billion, with a backlog of performance obligations rising to $638 billion, approximately 9.5 times the annual revenue; in the fourth fiscal quarter alone, $85 billion was added, mostly from large-scale AI contracts. The management team expects to recognize about 12% of this, approximately $77 billion, in the next 12 months, with another 34% expected over the following two years.
Contracts can be signed in advance, but costs are recognized gradually over several years. GPUs, data centers, and power must be in place before services begin. Capital expenditures for the same fiscal year rose to $55.7 billion, a 162% increase from $21.2 billion the previous year, with free cash flow at negative $23.7 billion. The $48 billion financing bridges the gap between construction expenses and future cloud service fees.
Some large contracts have adopted a new payment structure: clients prepay funds for Oracle to procure GPUs, or purchase chips themselves to provide to Oracle. Prepayments and client-provided hardware have totaled $75 billion, which the company claims reduces the capital it needs to raise. In the fourth fiscal quarter, new AI infrastructure contracts amounted to $67 billion, most of which belong to the aforementioned two categories.
The company plans to refinance approximately $40 billion in the next fiscal year, including a previously announced $20 billion at market price, and expects net cash capital expenditures of about $70 billion; client prepayments and timing differences are estimated at $20 to $25 billion, with on-book capital expenditures likely higher. During the same period, cloud infrastructure revenue grew by 93%, with over 1.2 gigawatts of capacity delivered to clients throughout the year.
In market dynamics, buyers are companies and platforms needing models for training and inference clusters, while the seller is Oracle, transitioning from database software to data center operations; this round of expansion is driven by signed multi-year computing power contracts. Funds flow from bond investors, equity subscribers, and clients prepaying for GPUs into land, power, and cabinets. Beneficiaries are large clients securing capacity and construction chains, while those under pressure must first pay interest and construction costs, then wait for contracts to convert into revenue.
Construction is on schedule, and future cloud service fees may gradually cover debts; delivery delays mean interest and vacant data centers will accumulate before revenue. The company currently aims to turn the signed $638 billion contracts into operational data centers.
Source: Public Information
ABAB AI Insight
Larry Ellison has transformed Oracle into a high-margin licensing software company, subsequently using the same balance sheet to acquire NetSuite and Cerner, integrating the application layer into the database empire. This round of bond issuance is no longer tied to software copyrights but to unpowered concrete and transformers. The OCI led by Clay Magouirk has shifted the company from 'selling licenses' to 'fronting costs and then leasing computing power', with the historical path being license → cloud subscription → utility-like capital expenditures, shifting risk from product cycles to construction timelines and electricity prices.
Funds are mobilized in three layers: the company issues bonds and mandatory convertible preferred shares, a $20 billion market-priced issuance, and $75 billion in client prepayments or self-supplied GPUs. Additionally, there are data center loans at the project company level, with $16.3 billion financing for a single Michigan site, and debts from partners in Texas, Wisconsin, and New Mexico moving some construction off the main balance sheet. The motivation is not to supplement working capital but to convert signed performance obligations into billable cabinets; the strategy is to use investment-grade credit to front physical infrastructure for model labs that cannot yet build their own power grids.
Similar structures can be seen in power companies financing based on installed capacity, U.S. telecoms during fiber cycles, and intermediaries like CoreWeave that specialize in GPU leasing. Microsoft and Google cover concurrent expansions with operating cash flow, while Oracle relies on the bond market. The industry position has shifted from the rental phase of software licenses to a heavy asset expansion phase that must first control land, substations, and cooling systems; the order book size now outpaces its own delivery organization.
Essentially, this is a reconstruction of the supply chain. Training demand has shifted the scarce link from software interfaces to electricity and construction permits; whoever can sign executable multi-year contracts between chips and electricity can turn client prepayments and bonds into data centers. The mechanism is: revenue recognition lags behind capital expenditures, and the contracts themselves become collateral for refinancing; if construction timelines or a single large client's payment capacity misalign, interest will appear on the income statement before cloud service fees.
ABAB News · Cognitive Laws
Orders are not cash; construction timelines are the balance sheet
Client prepaying for chips turns the buyer into a co-investor
Once a software company starts building power stations, the valuation logic shifts to interest rates