Sharon AI locks $356M GPU-backed debt at 9.95% to fund AI Factory buildout

SharonAI Holdings said Thursday it entered a US$356 million committed senior secured, GPU-backed SPV debt facility priced at a fixed 9.95% (excluding fees), with proceeds earmarked for compute deployments tied to customer contracts.

The AI buildout is no longer only an equity story. Companies that rent AI chips are borrowing against those chips and the contracts the chips will serve. Sharon AI’s $356 million GPU-backed facility at a fixed 9.95% is another sign that private credit will fund the racks if that contract book is thick enough — the company says the book already tops $8.8 billion in total contract value, while it aims past 68,000 NVIDIA GPUs by mid-2027.

On Thursday, 1 October 2026, SharonAI Holdings Inc. said it entered a US$356 million committed senior secured, GPU-backed SPV debt facility. The company trades on Nasdaq as SHAZ. It calls itself Sharon AI, and it calls itself a leading Australian Neocloud. A Neocloud, in that sentence, is a company that rents out AI computing, the chips and the rooms that run them, rather than only selling software. PR Newswire carries the release. The page stamps Oct 01, 2026, 07:00 ET, which is 7:00 a.m. Eastern. The dateline is New York. The source line is SharonAI Holdings Inc. The line under the headline says proceeds will support deployment of compute infrastructure dedicated to customer contracts. Those lines are the release.

What kind of loan this is, as the release states it. Committed means the lenders have agreed to provide up to US$356 million. Three hundred fifty-six million U.S. dollars is that ceiling. The release does not say how much has been drawn, the amount actually borrowed so far. Senior means these lenders stand ahead of creditors who rank behind them, if the borrower cannot pay. Secured means named assets stand behind the loan. GPU-backed means those assets are graphics processing units, the chips that do the AI math, and the cash flows tied to them. An SPV, a special-purpose vehicle, is a separate company set up to hold a particular deal. Here the release says the debt is senior, secured, and backed by the GPUs and the cash flows tied to them. The rate is fixed at 9.95%, excluding fees. Fixed means the rate does not move with the market. Excluding fees means charges on top of that rate are not inside the 9.95%. 9.95 percent is just under 10 percent. If the full US$356 million were drawn, that rate is about US$35.4 million a year before fees, because 0.0995 times 356 million is about 35.4 million. That dollar figure is arithmetic. The release does not print it, it does not say the full amount is drawn, and it does not print the fees. Those terms are Sharon AI’s.

What the money is for, and what stands behind it. The facility will be secured against the GPUs and associated cash flows. Proceeds support deployment of compute infrastructure dedicated to customer contracts. The release calls that a contract-backed funding structure, and it says the structure reflects the security and delivery discipline of the platform. It also calls the close a further step toward an ambition of gigawatt-scale AI compute across Australia, New Zealand, and the broader Asia-Pacific. A gigawatt is a billion watts, about the electricity of a large power plant. Gigawatt-scale, here, is how large the company says it wants the computing to get. It is not a statement that a site of that size is already running. The release does not name a city for a new hall, and it does not name a GPU model as the collateral. It says GPUs. Those lines are Sharon AI’s.

Who is in the deal. The release says the facility includes marquee Australian, Asian, and global investors, including Goldman Sachs and select large private credit funds. Marquee, in that sentence, means the names the company wants a reader to notice. Private credit is loans from investment funds, rather than a deposit-taking bank’s ordinary loan book. Including is the release’s word. It does not say Goldman Sachs led the facility, and it does not say how much any one investor put up. Jarden Australia acted as sole financial advisor and arranger. Sole means Jarden was the only firm in that role. An arranger is the advisor that brings the lenders and the borrower together. Those lines are the release.

The build-out this loan is supposed to begin. Sharon AI calls the facility the first in an expected series of GPU financings. The series, the release says, supports a scheduled build-out of over 68,000 NVIDIA GPUs deployed by mid-2027. NVIDIA is the company that designs the chips. Over 68,000 is more than 68,000. Mid-2027 is the middle of that year. Scheduled and expected are the company’s words. The release lists further GPU financings among the statements it calls forward-looking, meaning plans, not a count of chips already in racks. It does not say this US$356 million pays for all 68,000 chips. It does not name the model. Those lines are Sharon AI’s.

The larger capital total, and the contract book. With the closing of this transaction, the company says it will have secured over US$2.6 billion of institutional debt and equity over the past 10 months. Institutional means the money came from large investors. Debt is borrowed. Equity is money taken in for a share of the company. Over US$2.6 billion is the total the company states after this close. It is not this facility by itself. The release does not list the earlier deals inside that total. Ten months before 1 October 2026 is 1 December 2025. The release does not print that calendar day. It says the past 10 months. In the same announcement, James Manning, co-founder and chief executive, says the book of customer offtake now stands at a total contract value, which he shortens to TCV, of over US$8.8 billion. Offtake means customers contracted to take the compute. Total contract value is the full stated value of those contracts. It is not cash already collected, and it is not one year of revenue. Over US$8.8 billion divided by this US$356 million commitment is about 25. That ratio is arithmetic on the two company figures. It is not a claim that every contract dollar has been borrowed against. Those figures are Sharon AI’s.

What Manning said the loan is for. He said demand for sovereign and secure, trusted AI infrastructure keeps outrunning supply, globally and particularly across Australia, New Zealand, and the broader Asia-Pacific, and that access to debt that can scale is an important part of the company’s growth. Sovereign, in that sentence, means computing a customer can keep under its own rules, rather than only in someone else’s public cloud. He said the facility shows how Sharon AI expects to use debt markets to fund GPU deployments, using that offtake book. He said the structure is designed to enhance return on equity and, in time, long-term shareholder value. Return on equity is profit compared with the shareholders’ own money. Designed and expect are his words. He also said that with a strong balance sheet, a growing pipeline of contracted capacity, and a disciplined approach to where the capital goes, the company believes it is in a good position to keep scaling its AI platform across the Asia-Pacific. The release says the money supports a growing customer base of hyperscale, AI-native, government, enterprise, and research organizations. Hyperscale means the very large cloud operators. AI-native means a company built around AI. The release names the categories. It does not name a customer. Those lines are the release.

Who the about box says the company is. Sharon AI is a leading Australian Neocloud delivering trusted sovereign AI infrastructure. Through its AI Factory platform, and through technology and co-location partners, it says it sells the computing organizations need to build, train, and run AI, from training a model through inference and agentic AI. Co-location is renting space and power in someone else’s data center. Inference is running a trained model so it answers. Agentic AI, here, is software that takes a next step on a task, not only a chat reply. The about box says the company serves customers globally. The site it prints is www.sharonai.com. The media address is media@sharonai.com. The investor address is investors@sharonai.com. The release also says the company uses its investor-relations page, sharonai.com/investors, to put out material news. Those lines are the release.

The picture is a frame of the Sharon AI website. A lime bar across the top states the facility: Sharon AI enters into a US$356 million GPU-backed debt facility, a senior secured SPV facility, a fixed 9.95%, Goldman Sachs, and NASDAQ: SHAZ, AI Factory, Australia. Under the bar, the page shows the company logo and the site’s line about Australian sovereign AI infrastructure. The bar does not print a calendar date. It is a site frame with that financing caption. It is not a photograph of a data hall, and it does not show a named chip.

In plain terms, Sharon AI said on Thursday that it entered a committed US$356 million loan, senior and secured, backed by GPUs and the cash flows tied to them, at a fixed 9.95% before fees. Goldman Sachs and large private-credit funds are among the investors the company names. Jarden Australia arranged it. The company says this is the first of a series of chip financings for a schedule of more than 68,000 NVIDIA GPUs by mid-2027, that the close takes institutional debt and equity over the past 10 months above US$2.6 billion, and that customer contracts total more than US$8.8 billion. Those build-out, capital, and contract figures are Sharon AI’s. The release does not say how much of the US$356 million has been drawn, does not name a GPU model as collateral, and does not name a customer.