《The OpenAI Podcast》談 2026 AI 泡沫：別看股價，API 才是真實指標

In the latest episode of “OpenAI Podcast,” OpenAI CFO Sarah Friar and Vinod Khosla, founder of Silicon Valley venture capital firm Khosla Ventures, directly address the most frequently asked question in the market: “Is AI currently a bubble?” Both believe that instead of just focusing on stock prices, valuations, or fundraising amounts, we should return to the most fundamental indicators—how many people are actually using it and how many APIs are being called in practice. These real usage metrics are the key to judging whether AI is overheating.

Bubbles should not be judged by stock prices, but by whether people are actually using it

Khosla straightforwardly states that when the outside world talks about bubbles, the focus is often on stock prices, valuations, or investment enthusiasm. However, these are essentially reflections of investor sentiment—oscillations of fear and greed—and do not necessarily correspond to real-world demand. He believes that the true measure of whether there is a “bubble” is not price, but usage volume.

In the AI world, the most direct indicators are API call counts, such as:

• How many companies are truly integrating AI into their systems?
• How many products are calling models daily?
• Are API call volumes growing or declining?

As long as these numbers continue to increase, it indicates genuine demand, not just market sentiment artificially propping things up.

Looking back at the internet bubble, although stock prices collapsed, usage was unaffected

Khosla also compares this to the internet boom of the 1990s. He points out that, in hindsight, people often say “the internet bubble burst,” but in reality, it was a “stock price bubble” burst, not a demand bubble. If at the time, instead of focusing on tech stock prices, we had looked at internet traffic, user numbers, or data transfer volumes, we would have seen that despite wild stock price fluctuations, actual internet usage did not experience a so-called “bubble collapse.” People continued browsing, emailing, consuming content, and using services. The only change was investors’ perceptions of prices, not users’ reliance on the technology.

Khosla believes AI is on a similar path. Market sentiment might cause enthusiasm for certain AI companies today and skepticism tomorrow, but these fluctuations are not the same as the reality of whether developers are calling APIs daily or whether companies are integrating AI into their workflows.

Stock prices failing to reflect technological value will not show bubble-like collapses on APIs

Khosla even states confidently that he is almost certain:

“You won’t see a bubble burst like a sharp decline in API call volumes.”

Even if market sentiment fluctuates and media discuss bubbles daily, as long as demand and usage persist, API call volumes won’t suddenly vanish. In his view, API call volume represents:

• Genuine demand
• Actual usage
• Real value being generated

As for stock prices, he believes they are part of the capital market’s cycle and may not reflect the true value of the technology in the real world.

It’s not a lack of demand, but a mismatch in computing power supply

OpenAI CFO Friar also echoes Khosla’s perspective from a company management standpoint. She points out that the reality inside OpenAI is not “market cooling off,” but rather “computing power supply can’t keep up with demand.” She says that over the past few years, OpenAI’s computing power and revenue have grown in high sync:

• More computing power enables more products and features
• More products lead to higher actual usage, which in turn increases revenue

Friar admits that currently, the real limiting factor is not market interest but insufficient computing resources. If more computing power were available now, OpenAI could train more models, launch more features, and explore more multimodal applications. But the reality is, we are already constrained by computing capacity.

This also means that, in her view, talking about a bubble now is somewhat misplaced, because in most cases, demand leads and supply follows.

From Wall Street back to the technical front, APIs are the true test of a bubble

In this discussion, both try to shift the term “bubble” from the financial market to the technical realm. Instead of judging heat based on valuation levels or stock price fluctuations, they focus on a more tangible question:

“How many people are actually using it? Not who has the highest valuation or whose stock price is the most feverish, but how many APIs are being called and used daily in practice.”

(AI bubble theory is harshly challenged by reality: the leading seven tech giants’ wealth doubles, stock prices soar )

This article, “The OpenAI Podcast,” discusses the 2026 AI bubble: Don’t look at stock prices, APIs are the real indicator. Originally published on Chain News ABMedia.