$222 billion continues to "bleed" the AI track, capital is giving rise to a batch of unprofitable zombie companies

As a large amount of capital flows into the artificial intelligence sector, the market is facing an increasingly severe structural risk: huge funds are supporting a group of startups that, although fully functional, have lost their commercial viability. These companies, unable to turn a profit or repay debts, continue to survive thanks to ongoing financing "blood transfusions," and are referred to in the industry as "zombie companies." According to predictions from the National Venture Capital Association, by 2025, the venture capital transaction volume in the AI and machine learning fields will reach $222 billion, accounting for more than 65% of all venture capital funding in the United States. This proportion represents a significant increase from 47% in 2024 and 10% in 2015. However, despite ample funding, the speed of product delivery has struggled to match market hype, leading to a significant amount of capital flowing to companies that are, in reality, no longer competitive. This capital misallocation has raised widespread concerns among investors and economists. Joseph Favorito, founder of Landmark Wealth Management, pointed out that long-term support for debt-financed or insolvent AI companies could slow innovation and cause widespread economic ripple effects. This "blood transfusion" behavior is crowding out limited funds that could be used in more productive areas, leading to resources and talent being misallocated to underperforming companies. Market analysts warn that the current boom is reminiscent of the internet bubble era. With the end of the subsidy era and increasing industry consolidation, companies that cannot generate actual profits will face bankruptcy or be sold off at a loss, while those that can foresee this trend and possess real value-generating capabilities will emerge as the ultimate winners. Capital Frenzy and Zombie Risk In the AI startup sector, signs of "zombie companies" are becoming increasingly evident. Economists define such companies as those unable to repay debts, cover operating costs, or generate sufficient revenue, yet continue to survive through continuous injections of new capital, debt restructuring, or investors unwilling to accept losses. Typically, underperforming companies would quickly go bankrupt due to a lack of financing, but during the current boom cycle in AI, abundant funds have extended their lifespans. Data highlights the harsh reality of this market. Despite the influx of capital, data analysis firm Demand Sage reports that about 90% of startups ultimately fail. A report from management consulting firm Kearney shows that since 2010, the number of "zombie" companies globally has grown by about 9% annually, with an expected total of 2,370 by 2024. A recent report from Fortune magazine also noted that there are already 574 such companies in the venture capital space. Joseph Favorito stated that nearly half of venture capital flows into the AI sector, allowing these companies that should have gone bankrupt to survive far longer than expected. He believes that the challenge for capital allocators lies in determining whether the funds are being wasted or if they can yield long-term value returns. The Hidden Real Economic Benefits Industry insiders believe that cheap venture capital, government subsidies, and credit provided by cloud service providers are masking the fragile fundamentals of AI startups. Abdur Rehman Arshad, CEO of Capidel Consulting, pointed out that these subsidies artificially keep company costs low for three to six years, This has obscured the "true unit economic benefits." Arshad predicts that AI-related costs will grow three to ten times. He warns that by 2030, many companies will face an $800 billion revenue gap, which will directly lead them to become "zombie" companies. Brian Londono, founder of Resume Tailor AI and former venture capital analyst, shares a similar view, believing that many claims about AI "efficiency improvements" are mixed with speculative growth, with funds trapped in slowly developing companies. Once funds are released or subsidies stop, these companies will face the risk of sudden devaluation due to prolonged mispricing of risks. Furthermore, it is expected that by 2026, the federal government's annual funding for AI will reach $32 billion. Arshad believes that while this can extend the operational cycle of companies, it may also encourage the survival of unsuccessful enterprises. Profit Dilemma Under High Costs In addition to capital misallocation, the high operational costs of AI companies and unclear profit paths are also key factors leading to "zombification." The Harvard Business Review points out that generative AI has high variable costs and low variable returns, and the profit path for many companies is "unclear." Andy Wu, an associate professor of business administration, analyzes that the startup costs for AI startups are skyrocketing, with pre-seed round costs expected to rise from $50,000 to $2 million by 2026. Meanwhile, productivity improvements have not translated into revenue as expected. Research from MIT shows that 95% of AI application companies report no significant revenue growth. Arshad adds that early productivity declines are common, and this gap between input and output has bred a large number of companies relying on subsidies rather than actual output. Talent Competition and Market Reshuffling Talent loss is another major survival threat faced by AI startups, making it difficult for them to convert technology into commercial value even if they possess it. Former economics associate professor Nikki Finley points out that if top talent is poached by large companies, startups will lose acquisition opportunities. The case of Windsurf (formerly Codeium) in July 2025 exemplifies this. During its acquisition by AI startup Cognition, its leadership was poached by Google. Finley believes that while this may be a win-win for individual engineers and the overall economy, it also means that the value of startups as independent entities is hollowed out. In the face of the current frenzy, Joseph Favorito compares this to the "dot-com bubble" era, when large amounts of capital were blindly directed toward any project with ".com." He predicts that history will repeat itself, and with the inevitable consolidation of the industry, those companies unable to adapt to the real market environment will ultimately disappear through bankruptcy or fire sales