Long-Term Projections: The $5–$7 Trillion Range Forecasts for the total investment needed through 2030 vary by institution but consistently land in the multi-trillion dollar range: 

• Project Cancellations: Gartner predicts that over 40% of agentic AI projects will be canceled by 2027 due to costs and lack of value.  Historically, widespread project cancellations precede vendor insolvencies by 12–18 months.

The AI Data Center Boom Doesn’t Survive Basic Physics

https://www.youtube.com/@jovanhallInfrastructure and Permitting Failures Despite hundreds of billions in committed capital, nearly half of planned U.S. data center builds in 2026 have been delayed or canceled.  The primary constraints are power grid shortages, with the U.S. peak electricity supply projected to fall short by 2028, and supply chain bottlenecks for critical electrical components like transformers and switchgear, many of which originate from China.  Additionally, local community opposition has become a critical barrier, with 75 projects worth $130 billion stalled or rejected in Q1 2026 due to concerns over water use, energy consumption, and local impactClimate and Operational Risks Severe weather and heatwaves are increasingly disrupting operations, with extreme weather now the leading cause of loss in U.S. data center builders’ risk portfolios.  79% of global data center capacity faces elevated risks from climate hazards such as flooding and wildfires, which increase downtime and insurance costs,

• McKinsey & Company estimates that data centers specifically equipped for AI processing will require $5.2 trillion in capital expenditures by 2030, contributing to a total global data center requirement of nearly $7 trillion. 

further threatening the long-term viability of these infrastructure investments Critics warn this dynamic mirrors the dot-com bubble, where massive overinvestment in fiber optics led to stranded assets; today, the risk involves trillions in potentially stranded data center capacity that cannot be powered or monetized before the hardware depreciates

As of mid-2026, no major “frontier” AI lab (such as OpenAI, Anthropic, or xAI) has filed for insolvency; instead, these entities have shattered fundraising records, securing nearly $300 billion in the first quarter of 2026 alone.  However, a wave of insolvencies has devastated the secondary layer of the AI ecosystem, specifically “wrapper” startups, hardware ventures, and companies accused of “AI theater.”

Notable AI Insolvencies and Collapses (2025–2026) The most high-profile failure is Builder.ai, a London-based “unicorn” valued at $1.5 billion, which filed for bankruptcy in May 2025 after investigations revealed its “AI” platform was actually powered by 700 offshore human developers.  Other significant casualties include:

• Monarch Tractor: Shut down operations in April 2026 after raising $240 million, citing safety issues with its autonomous driving modes. 

• Ghost Autonomy: Filed for bankruptcy in April 2024 despite raising $220 million and partnering with OpenAI, unable to solve the engineering complexities of self-driving. 

• Prophesee: The French computer vision firm went insolvent in October 2024 after raising €126 million. 

• Icon.com: An AI ad generator that folded in March 2026 after burning capital on a $12 million domain name. 

• Jasper AI and Copy.ai: While not strictly bankrupt, both suffered massive devaluations and were acquired for parts or merged, signaling the collapse of the standalone “copywriting AI” market. 

The “Neocloud” Debt Crisis While not insolvent, specialized AI data center operators like CoreWeave are facing severe financial stress that mirrors insolvency risks. CoreWeave’s long-term debt surged to $17.5 billion by March 2026, with total liabilities approaching $25–$29 billion.  The company is paying roughly 25% of its revenue just to service interest on debt with rates as high as 15%.  Analysts warn that a slowdown in AI spending or the loss of a major customer like OpenAI could trigger a default, potentially cascading through the pension funds and insurers holding its debt. 

Systemic Financial Risks The divergence between the “haves” and “have-nots” is sharpening. Frontier labs are burning billions (OpenAI lost $13.5 billion in the first half of 2025) but survive on massive equity injections. In contrast, smaller firms without access to trillion-dollar balance sheets are collapsing under the weight of inference costs and hardware depreciation. The failure rate for AI startups accelerated rapidly, with the number of recorded failures jumping from 14 in early 2025 to 47 by March 2026, driven by the realization that many business models were economically unviable once the initial hype faded. 

AI startup bankruptcies 2026

CoreWeave debt crisis 2026