Valuations Reset After Overheating
Behind Rapid Growth: Rising Investment and Costs
Profit-Sacrifice Strategy Raises Sustainability Concerns

Sam Altman, CEO of OpenAI, the developer of ChatGPT, has described the surge of AI investment as a “bubble,” warning of overheated market sentiment. While he pointed out that the valuations of some companies have risen abnormally, he also stressed that AI is as significant as the commercialization of the internet. But with OpenAI expected to remain unprofitable for years to come, his optimism has struggled to gain traction.

Market Reacts Immediately to Insider Assessment

On the 21st, industry sources said Altman told U.S. tech media outlet The Verge in a recent interview: “Investors have been getting overly excited about AI lately.” As global big tech firms and startups pour massive sums into AI, valuations have ballooned in a short period. “When bubbles form, people often get overly excited about a small piece of truth,” Altman said. “Some company valuations are insane.”

Even as he voiced concern over the current frenzy, Altman underscored the lasting significance of AI. Citing the internet’s commercialization as the greatest transformation in modern life, he said, “If you ask me, ‘Are investors in an overall phase of overexcitement about AI?’ my answer is yes. But if you ask, ‘Is AI the most important development in a long time?’ my answer is also yes.”

Despite Altman’s defense of AI’s importance, investors rushed to sell. On the 19th, the day the interview was published, Nvidia, the largest company by market capitalization, closed down 3.5 percent on the New York Stock Exchange. AMD fell 5.4 percent, Broadcom 3.6 percent, Super Micro Computer 5.7 percent, and CoreWeave 4.04 percent, among others. The market interpreted Altman’s remarks not as exaggeration but as the sober assessment of an insider directly experiencing investor overexuberance.

Long Road Before Investors See Returns

Although OpenAI’s revenues are soaring, turning a profit remains a distant prospect. Bloomberg reported in March, citing internal sources, that OpenAI’s revenue this year is expected to reach $12.7 billion, more than triple last year’s $3.7 billion. The growth is extraordinary, but industry projections put OpenAI’s earliest profitability no sooner than 2029.

The reason lies in the immense costs of expansion, from maintaining servers and staffing to securing data. The computing resources required to train and run massive models are pushing costs higher at a pace that outstrips revenue growth, according to industry analysts. Additional expenses from cloud rental fees and competition for GPUs are further weighing on margins. This explains why, despite multiplying its subscriber base and revenues each year, OpenAI’s breakeven point has been pushed back by more than four years.

Experts describe the situation as an “anomalous case of rapid growth coupled with high-cost structures.” Corporate valuations are skyrocketing, but in terms of actual cash flow, expenses consistently exceed revenue. OpenAI’s willingness to endure losses is seen as a strategy to preserve market leadership, but in the long run, it could test the patience of investors.

Finding Balance Between Growth and Profitability

Even in developing its latest model, GPT-5, OpenAI acknowledged that the burden of computing resources and power consumption was a key constraint on performance tuning. As the number of parameters and context length expanded, GPU usage and memory requirements grew exponentially, driving up costs in both training and inference. Additional expenditures also arose from deploying failover systems and efficient scheduling to maintain stable training. As a result, the company scaled back some functions and opted for conservative settings in the final release.

This reality ties directly into OpenAI’s corporate strategy. In an interview with CNBC earlier this month, Altman stated, “For the company’s growth, we must prioritize investment in training and compute infrastructure, and that inevitably delays profitability.” He formalized the company’s deficit-tolerant strategy, underscoring that market dominance outweighs immediate profits. “Since we are not yet a public company, we are free from the pressures of public markets and will continue to pour money into AI training and compute,” he said.

Market participants, however, question the sustainability of this approach. Stable service quality at slightly lower performance may allow for short-term revenue gains, but as demand for large-scale inference becomes routine, infrastructure costs will inevitably strain the company’s finances. As a result, alternatives under discussion include reducing per-inference costs, diversifying pricing plans, securing long-term enterprise contracts, model optimization, and automating data pipelines. With the market closely watching whether these strategies can preserve growth while improving profitability, failure to strike the right balance could erode confidence in OpenAI’s long-term future.