“Losses despite rising revenue, obsolescence if investment stops”: AI companies’ dilemma of cutthroat competition—will the semiconductor industry’s game of chicken repeat itself?

OpenAI and Anthropic see profitability slide amid fierce investment and price competition
Big Tech companies operating AI businesses also suffer worsening cash flow as the “game of chicken” to secure market leadership continues
Memory market reshaped after cutthroat competition, with financial strength and technology determining winners and losers

Profitability is deteriorating at OpenAI and Anthropic. Despite sustained, strong revenue growth, profitability remains elusive as investment in computing infrastructure and price competition among artificial intelligence (AI) models intensify. Industry observers suggest that, just as the memory semiconductor market consolidated into an oligopoly dominated by a handful of companies after prolonged cutthroat competition, AI companies with financial strength and technological capabilities could strengthen their market dominance through a “game of chicken.”

OpenAI and Anthropic maintain revenue growth

According to Bloomberg and other foreign media reports on the 12th (local time), OpenAI posted revenue of USD 10 billion in the third quarter (July–September) of this year. That was approximately 50% higher than the previous quarter’s USD 6.7 billion and about 2.6 times the figure a year earlier. Growth was attributed to increased demand for the latest AI models, including GPT-5.6 and Astra, and the resulting expansion in application programming interface (API) revenue. Enterprise AI services and advertising also supported the improvement. OpenAI said its enterprise business doubled in size in the third quarter, accounting for more than half of total revenue.

Anthropic also recorded significant revenue growth. According to an amended registration statement for its initial public offering (IPO) recently filed with the U.S. Securities and Exchange Commission (SEC), Anthropic’s third-quarter revenue reached USD 16.96 billion, up approximately 47% from USD 11.6 billion in the previous quarter. Although quarterly revenue growth slowed from 144% in the second quarter to 47% in the third, increased usage of its AI coding tool Claude Code and enterprise API services suggests that its revenue growth trajectory remains intact.

Both companies mired in losses

The problem is that both companies are struggling to achieve profitability. OpenAI’s adjusted operating loss in the third quarter widened by approximately 8% to USD 10.9 billion from USD 10.1 billion in the previous quarter. Its adjusted operating margin improved from -150% to -109%, but losses still exceeded revenue. Anthropic also posted a third-quarter operating loss of USD 567 million, swinging from an operating profit of USD 974 million in the previous quarter. The burden of computing costs needed to meet surging AI demand is cited as a major reason for deteriorating profitability. Indeed, Anthropic procured USD 5 billion worth of computing resources in the third quarter alone through relatively expensive short-term contracts. Its cumulative computing and infrastructure costs for January–September this year surged 329% year on year to USD 18.808 billion.

Price competition among AI models has also emerged as a key factor. A push by Chinese companies offering low-cost models and intensifying competition among global AI developers for enterprise customers have accelerated fee reductions. In July, OpenAI cut API pricing for GPT-5.6 Luna and Terra by 80% and 20%, respectively, and in August it lowered pricing for GPT-5.6 Sol by more than 20%. Anthropic also said in its amended registration statement that price cuts for new models and contract discounts for enterprise customers offset USD 600 million of its revenue increase from the second to the third quarter. Fable 5.1, which Anthropic released in September, reduced the cost of typical tasks by approximately 25% and complex coding tasks by up to 45% compared with its predecessor. Opus 5.5, unveiled that same month, also reduced costs per task by approximately 40%.

Table 1. OpenAI and Anthropic third-quarter results

An industry structure that fuels competition

The structural characteristics of the AI industry explain why these companies continue such reckless competition. Companies in the industry must continually secure large-scale computing resources and data center infrastructure to develop high-performance models and expand services, and the financial strength to support those efforts determines their technological competitiveness and market position. In particular, as AI model performance has improved rapidly and new product release cycles have shortened recently, established leaders have been forced to make additional investments immediately whenever competitors narrow the technology gap. This is why companies prioritize expanding market share and securing customers over improving short-term profitability.

Market observers liken this competitive dynamic to the so-called “prisoner’s dilemma.” Companies could ease their financial burdens by jointly moderating the pace of investment, but any individual company that cuts investment while rivals continue spending faces a substantial risk of losing its technological competitiveness and customer base. Commenting on the situation, one market expert said, “AI companies are engaged in a game of chicken, continuing to invest while accepting enormous losses until the other side gives up first.” The expert added, “With the AI market’s technical standards and long-term competitive hierarchy still undecided, companies appear more concerned about being left behind in the future market than about immediate returns on investment.”

Big Tech also weighed down by cutthroat competition

Big Tech companies that have entered the AI industry are also participating in this cutthroat competition. Major technology companies, including Microsoft (MS), Alphabet, Amazon and Meta, are pouring enormous sums into building data centers and purchasing advanced semiconductors to secure market leadership. These four companies’ combined capital expenditure approached USD 201.9 billion in the third quarter of this year. That was more than double the year-earlier figure and exceeded the USD 175.5 billion in cash generated from operating activities during the same period.

The strain is also clearly evident in individual companies’ financial statements. Alphabet posted negative free cash flow of USD 9.1 billion in the third quarter of this year, while Meta also recorded a cash outflow of USD 1.2 billion. Amazon’s free cash flow over the latest 12 months swung to a deficit of USD 42.4 billion from a surplus of USD 14.8 billion in the corresponding year-earlier period. Only Microsoft maintained a relatively stable financial position, generating USD 18.6 billion in free cash flow in the third quarter.

A similar precedent in the memory industry

Some observers say the current situation recalls the game of chicken previously seen in the memory semiconductor industry. In 2007, expanded production by Taiwanese semiconductor companies triggered oversupply and a collapse in prices in the DRAM market. The global financial crisis the following year then compounded the situation, sharply eroding profitability across the industry. Unable to withstand the intense cutthroat competition, Germany’s Qimonda went bankrupt in 2009, while Taiwanese companies also suffered massive losses. In 2010, increased production by Japanese and Taiwanese companies sparked a second game of chicken. Even Elpida, which had stayed afloat with public funding from the Japanese government, went bankrupt in 2012 and was subsequently acquired by U.S.-based Micron.

After these two rounds of chicken, the global DRAM market, once contested by about 20 companies, became firmly dominated by three players: Samsung Electronics, SK Hynix and Micron. Samsung Electronics’ share of the global DRAM market, in particular, rose from approximately 27% in 2007 to around 45% in 2015, while the top three companies’ combined share reached approximately 95% as of 2015. As competitors exited, oversupply eased and the market consolidated into an oligopoly, improving the surviving companies’ pricing power and profitability.