US-China Big Tech "Battle of Titans" Heats Up... Who Are the Hidden Winners?

US Launches Aggressive Token Price Cuts as China Expands Investment
"Increased Infrastructure Spending Makes Semiconductors and Electricity the Ultimate Beneficiaries"

The global artificial intelligence (AI) industry has recently entered a full-scale "chicken game" phase. As major American AI companies lower their rates to defend market traffic, Chinese big tech firms are responding by ramping up investments. In this process, there is growing analysis that the ultimate beneficiaries will be infrastructure sectors such as semiconductors and electricity, which directly absorb massive investment flows.

According to Meritz Securities, as the performance of large language models (LLMs) from China, such as "Kimi K3," has risen to the frontier level of the United States, companies like OpenAI and Anthropic have started cutting model prices as a means of countering competition. These price cuts by U.S. AI firms are not just short-term promotions but indicate a structural shift toward normalized price levels for frontier models, marking the onset of an intense structural chicken game.

In fact, on July 30, OpenAI slashed the price of GPT-5.6 Luna by 80%, lowering it to $0.20 per 1M input tokens, and Anthropic decided to apply a permanent price cut to Claude Sonnet 5. Thanks to such aggressive price competition, global average daily token usage in August rose by 36.8% month-over-month, reaching approximately 11 trillion tokens, with token usage by American companies increasing by 47.9%, delivering short-term traffic defense effects.

However, this has not overturned the absolute cost-competitiveness of Chinese firms, which is grounded in factors like lower labor costs. Users responded more sensitively to the absolute price per token than to the rate of price cuts, as the absolute price of leading Chinese models such as DeepSeek remains at just a quarter of similar American models. As a result, despite a price hike, DeepSeek saw its token usage soar by 99.8%, and Chinese models continue to hold a majority share with 56% of total token volume.

Chinese big tech companies are rapidly accelerating their capital expenditures. According to second-quarter earnings results, Tencent’s capital expenditures soared 176% year-on-year to 52.8 billion yuan, Alibaba’s increased by 75% to 67.7 billion yuan, and Baidu’s rose by 200% to 11.4 billion yuan. The combined capital expenditure forecast for China’s four leading big tech companies this year has been revised up by 26.8%, from 473 billion yuan to 600 billion yuan. Assessments suggest that, beyond simple investment expansion, the rate of increase itself has now entered an accelerated phase.

As the speed of AI capital expenditure intensifies, Chinese companies are also accelerating capital raising. On the 23rd, Alibaba announced it would deploy all 80 billion Hong Kong dollars raised through a rights offering entirely toward expanding AI infrastructure. Major frontier LLM firms such as Moonshot AI, DeepSeek, and Zhifu are also stepping up preparations for their respective initial public offerings.

American big tech companies, too, are continuing their investments despite negative free cash flow (FCF). According to KB Securities, it is highly unlikely that big tech leaders—who have previously become winners through long-term investments despite losses, as seen with Amazon and Meta—will halt their investment activities on their own. Additionally, much like how WorldCom in 1999 could not stop capital spending despite negative FCF until it went bankrupt, the current focus should be on whether companies continue to invest, rather than numerical figures for cash flows at this point.

There is little doubt in the market about the AI industry’s overall growth potential. Token price reductions are being offset by increased usage, resulting in higher revenue, and major U.S. software firms are also improving operating margins by lowering labor costs, among other measures. However, amid the fierce chicken game among LLM developers, consensus is growing that the tangible results of large-scale investments will be concentrated in the AI infrastructure value chain—where continuous equipment supply is essential—rather than at the platform company level.

Choi Seolhwa, researcher at Meritz Securities, said, "The current AI competition between the U.S. and China still very much takes the form of a nationalist race to secure computing power, and it is highly likely that the funds raised by Chinese AI firms will ultimately flow into computing infrastructure, including Chinese semiconductors. As a result, the greatest beneficiaries will be in AI infrastructure sectors such as semiconductors and electricity."

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