Oracle has announced plans to secure between $45bn and $50bn in funding during 2026 to expand the capacity of its cloud infrastructure business.

The funding will support additional infrastructure capacity needed to meet contracted demand from cloud clients such as AMD, Nvidia, Meta, OpenAI, TikTok, and xAI.

Oracle aims to raise these funds through a combination of equity and debt financing, splitting the total roughly in half between the two approaches.

For the equity portion, Oracle intends to utilise a mix of equity-linked and common equity issuances. This will include an initial offering of mandatory convertible preferred securities and an at-the-market equity programme authorised for up to $20bn.

The company plans to issue shares under this programme in line with market prices and capital requirements.

On the debt side, Oracle expects to complete a single issuance of investment-grade senior unsecured bonds early in 2026.

The company does not plan any additional bond offerings for the year beyond this transaction.

Goldman Sachs & Co. will lead the senior unsecured bond offering, while Citigroup will manage both the at-the-market equity issuance and the convertible preferred equity offering.

Oracle’s board of directors has approved all transactions related to this plan.

The company stated that this approach aims to preserve its investment-grade rating and maintain balance sheet strength as it continues expanding its cloud services.

Recently, Oracle senior vice president Josh Pitcock, in the company’s official blog, said Oracle expects 2026 to be a pivotal year for advancing AI in the US and is building AI infrastructure to support medical research, scientific breakthroughs and economic growth.

He said Oracle is developing AI data centre campuses with OpenAI in Texas and at sites in New Mexico, Wisconsin and Michigan, and pledged the facilities will benefit local communities.

Pitcock said Oracle will fund on-site power generation or grid upgrades, plus transmission, storage and substations, to protect ratepayers. He added the centres use closed-loop cooling to limit water use and include setbacks, screening, road upgrades and local hiring.

In October 2025, Oracle and Nvidia collaborated to build an AI supercomputer at Argonne National Laboratory for the US Department of Energy (DOE).

The primary system, named Solstice, deployed 100,000 Nvidia Blackwell graphics processing units (GPUs) and used Nvidia networking to achieve a combined output of 2,200 exaflops of AI performance when paired with a second system, Equinox, which utilised 10,000 Nvidia Blackwell GPUs.

Both systems aimed to expand the DOE’s AI research capabilities across scientific, security, and energy fields.