'AI War' Google, Amazon, Microsoft... Now It's a Contest of Enduring Costs [Weekend Money]

Big Tech Capital Expenditure to Reach $7.6 Trillion by 2031
FCF Deterioration Concerns Mount After Second Quarter Results
Interest Rates and Inflation Become Key Variables in the AI Rally

Expectations that artificial intelligence (AI) will change the world remain strong. The problem, however, lies in the cost required to make this a reality. Until the first half of the year, the market viewed AI investments as building up for future returns. The prevailing belief was that companies securing more chips, more data centers, and more computing resources would become the winners. However, entering the second half of the year, the key question is changing. Now, rather than who can accumulate the most, it has become more important to determine who can ultimately bear the cost until the end.

On August 14, Meritz Securities analyzed in a recent report that, starting from the second half of this year, true winners will be sorted out among hyperscalers such as Google, Amazon, Microsoft, and Meta. This semiconductor cycle has been driven by demand rather than supply, making the financial strength of hyperscalers—the main players driving this demand—especially important.

It is hard to say that the fundamentals of AI investment have collapsed. Most hyperscalers increased their capital expenditure (CapEx) plans in second quarter earnings reports, and both Microsoft and Amazon delivered strong results. Expectations for positive earnings by semiconductor companies, as well as further rises in chip prices, also remain intact.

Yet the market has begun to scrutinize AI investments more closely. While falling token prices were previously viewed as positive for AI adoption, if the decline is too rapid, it takes on a different meaning. When competition overheats, as in the era of internet portals or social networking services, the technology itself may become widespread, but expectations for direct revenue generation from that technology may diminish.

The China factor is also a concern. The performance of Chinese large language models (LLMs) is narrowing the gap with the United States more rapidly than anticipated, and token usage for open-weight models in China is also rising fast. Although this is positive for the proliferation of AI, it heightens uncertainty about future profitability compared to the massive levels of investment being made.

The biggest concern is cash flow. According to estimates by Goldman Sachs, capital expenditure by U.S. hyperscalers is expected to reach 800 billion dollars (approximately 1,134.56 trillion won) this year, increase to 1 trillion dollars next year, and total 7.6 trillion dollars by 2031. If free cash flow (FCF) surges starting in 2028, such investments may be validated, but until then, companies must endure a lean period lasting one to two years.

That lean period may be steeper than expected. After first quarter results, it was predicted that the FCF for hyperscalers would turn negative around the first half of next year. However, after second quarter earnings, new forecasts indicate that FCF could turn negative as early as the fourth quarter of this year, and that combined deficits could exceed 100 billion dollars by next year.

If capital expenditure exceeds operating cash flow, raising external funding becomes unavoidable. There are also concerns that some long-term data center contracts are being recorded as off-balance sheet liabilities. According to analysis by the Bank for International Settlements (BIS), U.S. hyperscaler liabilities may amount to 1.6 trillion dollars—not the 500 billion dollars shown on balance sheets.

Ultimately, interest rates are the wild card. As the amount of debt increases, any rise in interest rates only heightens the burden. The continuing war between the United States and Iran has pushed West Texas Intermediate (WTI) crude prices above 80 dollars per barrel, and renewed concerns over the Federal Reserve's tightening stance are also weighing on sentiment. Yeosam Yoon, a researcher at Meritz Securities, explained, "If interest rates rise further in response to inflation, the first to be hit will likely be companies investing heavily in AI, given the enormous capital requirements. This is not about doubting the fundamentals of AI investment itself, but rather highlighting the vulnerability of the macroeconomic environment."

This content was produced with the assistance of AI translation services.