For years, artificial intelligence (AI) startups have been pitched as the vehicles of the next productivity boom.
But as product delivery lags behind the hype, some AI companies are slipping into a quieter, more troubling category: startups that are functional but no longer viable.
Economists and financial institutions describe these firms as “zombie” companies—businesses that can’t cover their debt, operating costs, or generate sufficient returns, yet continue to survive through repeated injections of fresh capital, debt restructuring, or investor reluctance to accept losses.
Venture capital, financial, and AI insiders say that signs of these zombie companies are increasingly visible in the AI startup community.
AI and machine learning venture capital deals accounted for more than 65 percent of all U.S. venture capital project funding in 2025, totaling $222 billion, according to the National Venture Capital Association.
That’s a lot of money flooding into an investment market with a high rate of failure. Approximately 90 percent of startups fail, according to analysis by data analytics company Demand Sage.
While a universally accepted count of U.S.-based AI firms operating as zombies is unavailable, a recent Fortune report put the number of venture capital zombies at 574. An analysis by management consulting firm Kearney reported that the number of zombie companies worldwide has grown by about 9 percent annually since 2010, with a total of 2,370 as of 2024.
Concern is no longer limited to investor losses but also broader economic effects, such as misdirected capital and talent being tied up in underperforming companies. Some believe this could slow AI productivity and future innovation.
“Typically, unproductive entities would fail quickly due to the inability to secure further funding and service debt. But with the AI boom, nearly half of the venture capital funding is going into all things AI, and it’s extended their lifespan beyond what normally would be expected,” Joseph Favorito, founder of Landmark Wealth Management, told The Epoch Times.
“At any given point in time, there is a limit to how much capital will be allocated towards innovation. If capital is allocated to an insolvent entity, those dollars could have been put to better use elsewhere in a place that would promote innovation and productivity,” he said.
“But this always happens on some level. That is the nature of capitalism. The challenge is that every entity starts off in debt. It is up to those that are allocating their capital to determine if they are throwing good money after bad or if there is a valuable longer-term reward.”
Pay the Piper
“When the subsidy era ends, the companies that survive will be the ones that saw it coming,” Abdur Rehman Arshad, CEO of Capidel Consulting, told The Epoch Times.
He also said he anticipates that costs related to AI will climb threefold to tenfold.
“Many will face an $800 billion revenue shortfall by 2030, turning them into ‘zombie’ outfits,” Arshad said.
However, he stressed that factors such as venture capital, grants, and cloud credits—a type of virtual currency offered by cloud service providers—can also make the difference between a “bankrupt founder and the next unicorn.”
Companies that qualify as AI “unicorns” are those valued at more than $1 billion. As of December 2025, there are 308 AI unicorns, including OpenAI, which owns ChatGPT.
Brayan Londono, founder of Resume Tailor AI and a former venture capital analyst, suspects that many of the AI “efficiency gain” claims are mixed in with speculative growth. The result is a need to scale and push for more “fertile stories” to continue funding nonviable businesses.
“It has been my experience that credits from the cloud, government, or enterprise contracts, while delaying infrastructure costs, have hidden the weaknesses in fundamentals,” Londono told The Epoch Times.
“I have observed how a lot of money is trapped in low-trajectory AI start-ups, and the moment it unwinds would result in a sudden devaluation, as those distortions will also have been exposed, partly through having risk mispriced for so long.”
“Not unlike the dot com era, there were large amounts of capital being allocated to anything that had dot com in its name,” Favorito said.
“Ultimately, you’ll see the same thing happening. These companies will disappear through fire sales or bankruptcies, and there will be winners that reap the rewards as the space consolidates.”
—Autumn Spredemann; Stacy Robinson
BOOKMARKS
California Attorney General Rob Bonta is going after “ghost guns,” suing two companies and more than 100 individuals for allegedly distributing computer code to 3D-print the untraceable firearms. The defendants’ conduct “enables unlicensed people who are too young or too dangerous to pass firearm background checks to illegally print deadly weapons without a background check and without a trace,” Bonta said.
Ghislaine Maxwell pleaded the Fifth at a House Oversight Committee hearing on Monday. “After months of defying our subpoena, Ghislaine Maxwell finally appeared before the Oversight Committee and said nothing,“ Rep. Robert Garcia (D-Calif.) said in a statement.
Russian Foreign Minister Sergey Lavrov has warned that his country does not intend to attack Europe, but will respond to an attack with full force. Any attack on the Russian Federation would be met with “a full-scale military response using all available military means, in accordance with the relevant doctrinal documents,” he said.
Israeli authorities have repealed laws that made it difficult for Jews to purchase land in certain parts of the West Bank. “From now on, it will be possible to purchase land in Judea and Samaria through a simpler and more transparent procedure,” Israeli Finance Minister Bezalel Smotrich said.
The Department of Justice is seeking to dismiss charges against Steve Bannon, stemming from his refusal to appear before Congress in 2022. “The government has determined in its prosecutorial discretion that dismissal of this criminal case is in the interests of justice,” prosecutors said in their court filing.
—Stacy Robinson
