
Oracle is seeking $20 billion through share and bond sales as it expands its AI and cloud infrastructure. Quarterly disclosures show that customer prepayments brought in substantial cash in the three months through August, while capital spending exceeded operating cash flow.
The fundraising arrives after a difficult month for AI stocks. It also coincides with Marvell’s agreement to buy Lumentum. Together, the announcements show how infrastructure suppliers are using capital markets to fund expansion and acquire technology.
What Oracle is asking investors to fund
Oracle’s preliminary share-offering prospectus, dated November 2, describes $10 billion of common stock alongside a separate $10 billion senior-note offering. Each can close independently. The document allows proceeds to support general corporate purposes, including capital expenditure and debt repayment, giving Oracle flexibility over how it uses the money.
The filings put into motion a plan outlined at Oracle’s October 28 Investor Day. Management expects capital spending to peak in fiscal 2027 and 2028, with company-wide free cash flow turning positive in the second half of fiscal 2028. It targets at least $30 billion of free cash flow in fiscal 2029. These are management forecasts.
Oracle’s stated $20 billion limit on additional capital-market funding through fiscal 2030 excludes refinancing existing debt. It also follows an earlier $20 billion share sale completed in the first quarter of fiscal 2027. The company expects vendor financing, customer-supplied hardware and customer prepayments to reduce its upfront cash requirements.
The mix matters. Equity funding spreads ownership across more shares, while debt adds repayment and interest obligations. The preliminary bond prospectus lists maturities from 2029 to 2046, with final rates still blank. The eventual pricing will help determine the cost of financing the expansion.
Contracts, cash receipts and free cash flow
Oracle’s quarterly filing for the period ended August 31 shows why the timing of payments deserves attention.
| Oracle, first quarter of fiscal 2027 | Amount |
| Cash from operating activities | $23.10 billion |
| Capital expenditure | $28.50 billion |
| Free cash flow | −$5.40 billion |
| Customer prepayments with a significant financing component, included in operating cash flow | $11.4 billion |
The prepayments brought in cash before Oracle delivered the corresponding services. Despite that support, investment spending exceeded operating cash generation. Oracle also reported approximately $125.34 billion in notes payable and other borrowings, combining its current and non-current balances.
Its remaining performance obligations reached $664 billion. Oracle expected to recognize about 13 percent of that amount as revenue over the following twelve months. That percentage describes revenue recognition, not the timing of customer payments.
Oracle’s annual report explains the distinction: remaining performance obligations include deferred revenue, amounts already invoiced but not yet collected and recognized, and amounts to be invoiced in future periods. Revenue recognition and billing can occur at different times.
For investors, this makes payment schedules a separate question from contract growth. An expanding backlog establishes future work to deliver. Assessing whether that work can finance more capacity requires examining when cash arrives, when equipment must be paid for and what margin remains after providing the service.
Marvell buys a larger role in AI connectivity
Marvell’s agreement to acquire Lumentum uses shares to change ownership of an infrastructure supplier. The announced transaction carries an implied enterprise value of $115 billion, including debt. Lumentum shareholders would receive 4.5 Marvell shares for each share they own and hold roughly 33 percent of the combined company.
With a fixed exchange ratio, the value Lumentum shareholders receive will move with Marvell’s share price.
The industrial rationale centers on moving data between computing systems. Marvell brings chip and connectivity technology; Lumentum adds optical components. The companies expect their combination to improve the integration of electrical and optical connections used in AI infrastructure.
For cloud operators, those connections are part of the cost of turning installed processors into useful computing capacity. Faster chips still need networks able to move data between them. The engineering challenge helps explain why optical suppliers attract strategic interest alongside processor designers.
From demonstrations to deployed infrastructure
Marvell and Lumentum were already working together before the acquisition announcement. In March, they announced a joint optical-circuit-switching demonstration for the OFC industry conference. It paired Marvell connectivity products with Lumentum’s switching platform.
Optical circuit switching creates direct paths for light signals between endpoints. The companies presented their approach as a way to reduce intermediate signal conversions and improve network efficiency. Commercial deployment depends on compatible hardware, customer qualification and reliable production.
Useful capacity is also central to Oracle and Nvidia’s work on infrastructure utilization. Financing a server installation and earning an adequate return from it are separate milestones. For infrastructure buyers, delivery dates, usable performance and integration with existing systems are more informative than the headline size of a financing or acquisition.
Marvell and Lumentum expect the transaction to close in the first half of calendar 2028, subject to shareholder and regulatory approvals. Oracle’s nearer-term milestone is the pricing and completion of its financing. Those events will provide concrete terms against which to assess the companies’ expansion plans.
Popular news

DeepSeek and Moonshot Allegedly Used OpenAI Answers to Train Their Own AI

AI Stocks Are Plunging Despite Rising Revenue: What’s Going On?

OpenAI and Foxconn in Talks Over New AI Data Center Partnership

Keeping work and personal life separate in ChatGPT just got easier

DeepSeek’s AI Agent Prototype Takes Coding Tasks From Team Chats

