The Financial Times (FT) reported on the 14th (local time) that the total amount promised for purchase by five major hyperscalers (large AI data center operators) and Nvidia exceeds 1.5 trillion dollars (approximately 2.12 trillion won).

The FT explained that this figure was obtained after reviewing the quarterly reports submitted by Alphabet, Microsoft (MS), Amazon, Meta Platforms, Oracle, and Nvidia.

Their purchase commitments increased from about 1 trillion dollars at the end of the first quarter to more than 1.5 trillion dollars at the end of the second quarter.

The FT explained that the 1.5 trillion dollar figure is based on the premise that Nvidia, which has not yet announced its second-quarter earnings, had its purchase commitments at the end of the second quarter frozen at the first-quarter-end level, adding that such a scenario is highly unlikely.

The expectation is that the figure tallied after Nvidia's earnings announcement will be even higher.

Alphabet saw the largest surge.

Alphabet's purchase commitments jumped from 332.4 billion dollars at the end of the first quarter to 811 billion dollars at the end of the second quarter.

Alphabet specified in its quarterly report that "as of the end of the second quarter, we have significant purchase commitments and other contractual obligations amounting to 811 billion dollars, of which 207 billion dollars are short-term obligations."

Alphabet explained, "These purchase commitments primarily consist of technology infrastructure and inventory-related costs under long-term supply contracts and outstanding purchase orders. Additional contractual obligations include content licensing commitments and energy take-or-pay contracts."

Analyzing the details, the FT stated, "Most of the purchase commitments appear to consist of commitments to secure chips and power necessary for the data centers the companies are building."

Meta, the operator of Facebook and Instagram, also saw a significant increase in its purchase commitments, rising from 238 billion dollars at the end of the first quarter to 349.3 billion dollars at the end of the second quarter.

Meta stated in its quarterly report that "as of the end of the second quarter, we have entered into non-cancellable contracts totaling 349.3 billion dollars, including short-term and long-term contracts, the majority of which relate to third-party cloud capacity agreements and other technology infrastructure investments."

The FT stated that the purchase commitment figures it identified match those previously calculated by Morgan Stanley.

According to a Morgan Stanley report last month, purchase commitments by the five companies—Alphabet, MS, Amazon, Nvidia, and Oracle—reached 982 billion dollars (1.391 trillion won) at the end of the first quarter, which generally represents contractual obligations to purchase AI chips, computing resources, power required to operate data centers, and other equipment.

Previously, Goldman Sachs analysts explained, based on the quarterly reports of hyperscalers, that they had entered into 1.5 trillion dollars in lease agreements, of which 1 trillion dollars has not yet been reflected as general liabilities on financial statements because the contract periods have not yet commenced.

Purchase contracts and lease agreements are categorized as "hidden liabilities" that do not appear on financial statements.

Financial experts warn that off-balance-sheet liabilities can create an "illusion" that AI companies' financial health is robust, particularly for general investors who primarily look at financial statements.

Because lease and purchase commitments vary widely in their disclosure formats across companies and often do not transparently reveal the details of the agreements, they can boomerang into liquidity risks when AI profitability deteriorates or market volatility increases.

Since lease and purchase contracts are contractual promises, they are not reflected as general liabilities on financial statements and are only recorded in the footnotes until actual transactions, such as leasing or delivery of goods, commence.

(Photo: Getty Images)