On November 20, a single remark from Bridgewater’s Ray Dalio—“AI is forming a bubble”—stretched global capital markets’ nerves to the limit. Before the words faded, NVIDIA fell 10% in a single month, and the UK’s star AI company Robin AI directly went bankrupt and was put up for sale. Within ten days, short sellers, analyst reports, and plunges came one after another—the bubble thesis gained its first “tangible” feel. From Wall Street shorting waves to star-company bankruptcies, from top investors’ warnings to index pullbacks, the AI industry is going through an unprecedented crisis of confidence. Has this AI carnival that has burned through tens of billions of dollars already reached the threshold of bursting?
I. Capital Carnival: The “Unbridled” Path of AI Valuations
In October 2025, OpenAI announced a $6.6 billion financing and set a record for the largest single fundraising by a startup globally, with its valuation breaking through $157 billion, leaping to the world’s third-largest unicorn. The news quickly ignited the global tech and investment circles and again pushed the “AI bubble” controversy to a climax. From ChatGPT’s sudden emergence in 2022, to Sora stunning the world in 2024, and to the AI Agent concept sweeping the globe in 2025, artificial intelligence has undergone a rapid leap from concept to capital frenzy in just three years. However, behind this seemingly prosperous feast, the doubts about whether AI has fallen into a bubble have grown louder. This article, drawing on multiple media reports and industry data, will objectively dissect the real picture of today’s AI industry and explore whether this technological revolution has morphed into a capital game.
OpenAI’s fundraising record is not an isolated case. According to Caijing magazine, global AI financing totaled $125 billion in 2024, up 42% year over year—far outpacing overall global venture capital growth. The US and China alone accounted for nearly 80% of AI startup funding. In China, there were 1,023 financing events in AI in 2024, totaling more than $45 billion, up 38% from 2023. Top players such as Moonshot AI, MiniMax, and Zhipu AI saw their valuations double within half a year; some firms even joined the “unicorn” ranks before their products were commercialized.
The speed of these valuation surges has far outstripped the growth logic of traditional tech companies. Take Moonshot AI as an example: when it completed a $1 billion round in March 2024, it was valued at $2.5 billion; by October 2024, its valuation had skyrocketed to $5 billion, doubling in just seven months. This “lightspeed valuation” phenomenon has sparked broad concern. In early 2025, GSR Ventures partner Zhu Xiaohu publicly stated: “Over 90% of AI companies are overvalued and will face brutal consolidation.” He noted that many AI projects can fetch hundreds of millions of dollars in valuation based on nothing more than a demo or a concept—eerily similar to the scenes of the 2000 dot-com bubble, driven by capital fervor detached from fundamentals.
Even more alarming is the severe disconnect between revenues and valuations at some AI firms. According to Yicai, in 2024 fewer than 5% of Chinese AI startups posted annual revenue exceeding RMB 100 million (about $14 million), yet there were over 30 companies valued at more than $1 billion. This structural imbalance of “high valuation, low revenue” has become a key piece of evidence for the AI-bubble thesis.
II. The Technology Illusion: The Gap from “Disruptive” to “Middling”
AI’s actual ability to land in real scenarios has become the key yardstick for testing whether there is a bubble. Although large models have shown astonishing capabilities in text generation and image recognition, their limitations have become increasingly prominent in commercialization. For AI coding assistants, for example, a 2024 GitHub survey showed that only 17% of developers believed AI tools “significantly improved efficiency,” while more than 60% said AI-generated code required extensive manual correction and in fact increased workload. Microsoft’s 2025 internal assessments similarly found that the code adoption rate of its AI-assisted programming tools in real projects was under 30%, far below market expectations.
In healthcare, the accuracy of AI diagnosis faces challenges as well. A 2024 study in Nature showed that AI’s misdiagnosis rate for rare diseases was as high as 34%, far higher than human experts’ 12%. In China, after a Class-A tertiary hospital introduced an AI-assisted diagnosis system, medical disputes due to misdiagnosis rose 22% within a year, and the hospital ultimately had to suspend some of the system’s functions.
More awkward still, some AI products have fallen into a loop of “dazzling demos, weak deployment.” An “AI salesperson” released by a leading AI company in 2024 could converse smoothly with customers onstage at the launch, but in actual rollout, it was heavily complained about for failing to understand dialects and handle complex contexts, and was taken offline just three months after launch. This huge gap between “technology illusion” and business reality is draining market trust in AI.
III. The Profitability Dilemma: An Unbridgeable Chasm Between Burn Rate and Returns
The ferocity of AI’s cash burn has become jaw-dropping. According to OpenAI’s 2025 financials, its annual loss reached $5 billion, mainly due to compute costs and R&D. The daily operating cost of ChatGPT alone exceeded $7 million, while annual revenue was only $3.7 billion, leaving a huge deficit. To keep running, OpenAI had to rely on financing “blood transfusions”; of its $6.6 billion round in 2025, explicit use-of-proceeds was earmarked for “compute procurement over the next three years.”
Chinese AI firms are in similarly dire straits. At one leading large-model company, R&D investment in 2024 reached RMB 1.5 billion (about $208 million), but annual revenue was only RMB 80 million (about $11.1 million), meaning R&D was 18.75× revenue. In an internal letter, the CEO admitted: “If we cannot achieve a commercialization breakthrough within the next two years, the company will face capital-chain risks.” This business model of “high input, low return” has trapped the AI industry in a vicious circle of “the more advanced, the more loss-making.”
More brutally, the “Matthew effect” is intensifying. CB Insights data shows that in 2024, 90% of global AI financing went to the top 10% of companies, leaving the remaining thousand-plus AI startups to split just 10% of the funds. Many small and mid-sized AI firms, unable to secure follow-on financing, have already seen waves of layoffs and bankruptcies. In the first half of 2025, more than 200 AI startups in China deregistered, up 150% year over year.
IV. Historical Mirror: The Cycle from the Dot-Com Bubble to the AI Frenzy
Today’s AI shows striking similarities to the 2000 dot-com bubble. Between 1999 and 2000, the Nasdaq surged from 1,500 to 5,048—an increase of over 230%—with internet firms taking half the stage. Back then, any company with “.com” in its name could see its valuation multiply. Pets.com, founded in 1999, saw its stock soar 70% on listing day despite never turning a profit, with market cap once hitting $300 million, but it declared bankruptcy just 268 days later.
History rhymes. Since 2024, global AI concept stocks’ average P/E has exceeded 200×, far higher than the 15–20× average of traditional sectors. In China, one AI-chip firm posted net profit of just RMB 50 million (about $6.9 million) in 2024, yet its market cap once broke RMB 100 billion (about $13.9 billion), implying a 2,000× P/E. Such valuations detached from fundamentals mirror the “price-to-dream” ratios of the dot-com era.
Equally noteworthy is that today’s “concept speculation” playbook in AI resembles that period: from 2023’s “Year One of LLMs,” to 2024’s “AI Agents explosion,” and to 2025’s “AI+Everything,” every year a new concept is hyped by capital. As one VC partner revealed: “When investing in AI projects now, we don’t look at the tech or revenue; we look at whether the story sounds compelling, whether it can tell a ‘disrupt-the-industry’ narrative.” This story-driven logic is a textbook sign of a bubble.
V. The Road Ahead: Finding Hard-Core Value Amid the Foam
Despite conspicuous bubble signs, AI is not value-less. Unlike the 2000 dot-com bubble, today’s AI already has real application scenarios and commercial cases. For instance, AI’s substitution rate in customer service has reached 30%, significantly reducing labor costs; in film and TV production, AI-assisted generation has cut VFX costs by 40% and shortened production cycles by 50%. McKinsey forecasts that by 2030, AI will contribute $13 trillion to global GDP, with a 42% compound annual growth rate.
The key is how to sift out companies with real technical moats and commercial flywheels amid the foam. History shows that after a bubble bursts, 90% of companies disappear, but the remaining 10% grow into the next generation of giants. Amazon and Google emerged precisely after the 2000 collapse. The current AI industry likewise needs a “defrothing” process to weed out firms that live by concept alone and keep the players that truly create value.
For investors, it is time to return to business fundamentals and focus on hard indicators such as revenue structure, customer retention, and technical barriers, rather than being seduced by “disruptive innovation” narratives. For entrepreneurs, abandon “valuation-driven” thinking, focus on real pain points, and build sustainable business models. As one AI-unicorn founder said: “Only when AI is no longer an ‘AI company,’ but becomes infrastructure like water and electricity, will this industry truly mature.”
Does an AI bubble exist? The answer is likely yes. But “bubble” is not a pejorative—it is the growing pain of every technological revolution. From railway manias to the internet bubble, history repeatedly proves: after the bubble bursts, real value emerges. Today, the AI industry stands at a crossroads—leftward lies the abyss of capital revelry; rightward, the long march of technology landing. The ultimate winners will be the few who can stay clear-headed amid the clamor and hold to essentials within the foam.
[Disclaimer]: The above content reflects analysis of publicly available information, expert insights, and BCC research. It does not constitute investment advice. BCC is not responsible for any losses resulting from reliance on the views expressed herein. Investors should exercise caution.