NAVER and Brookfield sign deal for up to $9 billion in AI infrastructure

NAVER and Brookfield signed an October 27 agreement for up to $9 billion in AI-computing infrastructure investment, replacing a nonbinding term sheet. A November 9 account from Global Data Center Hub details the financing structure: Brookfield would own the hardware, NAVER Cloud would sell its computing services, and NAVER would guarantee the operating company’s usage fees.

The agreement is signed, but the financing and deployment are not complete. On November 1, KB Kookmin Bank announced a mandate to arrange about KRW 10.7 trillion in loans for a KRW 12.6 trillion project. Final lending commitments have not been disclosed, and none of the capacity under this structure is operating yet.

One company owns the chips; another sells the computing

Brookfield plans to create a company tentatively called Asset Co. to buy graphics processing units, or GPUs, and data-center assets. A Brookfield investment vehicle would provide all its equity capital. NAVER Cloud would operate the AI factory, serve customers and pay Asset Co. for using the infrastructure. That division lets NAVER retain the operating role and customer relationships without buying the hardware upfront.

NAVER says the usage fees would be linked to NAVER Cloud’s business revenue, easing financial pressure during the early stage. For lenders, the repayment source is the expected long-term revenue from selling AI computing. The structure therefore depends on the operating business, not simply on ownership of expensive equipment.

The guarantee backs fees, not the bank loan

NAVER guarantees NAVER Cloud’s fee obligations for up to eight years. Global Data Center Hub, citing The Korea Herald, puts the nominal guarantee at KRW 18.74 trillion. Crucially, that support does not guarantee the syndicated bank loan itself. Banks would lend to the Brookfield side, which receives NAVER Cloud’s payments; NAVER stands behind those payments rather than directly behind the borrowing.

The fee formula remains undisclosed. Public details do not establish whether there is a minimum payment, how fees change when sales slow, or what happens to the GPUs at the end of the arrangement. Those terms would determine how much payment flexibility the revenue link provides. The uncertainty also matters because GPU resale values are difficult to predict, according to The Korea Herald’s reporting cited in the account.

Outside ownership would not necessarily keep the project outside NAVER’s accounts. NAVER expects to consolidate Asset Co. as a subsidiary because it would control the company under Korean accounting standards, despite Brookfield supplying all the equity. The assets and associated borrowing are therefore expected to appear in NAVER’s group accounts. External funding reduces the upfront hardware bill; it does not make the project disappear from financial reporting.

A staged buildout still needs paying customers

The deployment plan starts at NAVER’s GAK Sejong data center. Its capacity targets, measured in megawatts, are staged across two years rather than arriving at once:

• First half of 2027: 55 megawatts.
• End of 2027: 100 megawatts.
• 2028: 200 megawatts.

NAVER’s stated targets are enterprise and government customers in Asia, Europe and the Middle East. But it has not named customers for this capacity, and no purchase commitments have been disclosed. The wider ambition is a one-gigawatt-scale global AI factory; the financing described here covers the period through 2028, with no public funding or schedule for the remainder.