DATA CENTER BUBBLE? MUSK WARNED THAT 200 AI CLOUD COMPUTING COMPANIES ARE AT RISK OF COLLAPSE.
Elon Musk warned that 200 AI cloud computing companies are at risk of collapse! The data center bubble may face major challenges by 2027. The Neocloud battle between Elon Musk and NVIDIA’s Jensen Huang unfolds amid rising AI demand but shrinking capital availability. Analysis of risks to data center investments in Thailand by Dr. Methangkorn Sermsuk (CIIM – KMITL).
The artificial intelligence (AI) industry is facing a major warning from Elon Musk, CEO of SpaceX and Tesla, who stated that over 200 next-generation AI cloud service providers—known as Neoclouds—could face widespread failures and consolidations in the near future.
This warning appeared in a CNBC report on October 26, 2026, in which Musk used the phrase “death and merger on a rainbow fish scale level,” comparing it to the collapse of numerous small companies in an intensely competitive market.
The core issue is not whether AI will stop advancing, but whether the many companies borrowing massive sums to purchase GPUs, build data centers, and lease computing power can generate sufficient revenue and cash flow to repay their debts.
In recent years, investment in AI infrastructure has been driven by expectations of ever-increasing computational power—for both training AI models and inference (delivering AI services to end users).
However, as industries must invest heavily in expensive equipment upfront, while revenues and profits depend on uncertain future demand, financial risks are becoming increasingly apparent.
This situation holds significant implications for Thailand, which is rapidly attracting massive investments into hyperscale and AI data centers.
Therefore, the critical question is not merely how many more data centers Thailand will need, but how Thailand will be affected if the global AI and cloud computing industry enters a correction phase—and whether we should implement policies now to mitigate these risks.
Neocloud refers to cloud infrastructure providers focused exclusively on leasing specialized AI computing power, such as GPU computing, AI model training, and AI inference.
This business model is growing due to demand for high-performance GPUs, particularly NVIDIA’s hardware, which plays a central role in the AI computing market.
Neocloud companies attracting investor attention include CoreWeave, Nebius, and IREN—firms building infrastructure for AI and computing services.
The key distinction between Neoclouds and traditional hyperscale cloud providers like AWS, Microsoft Azure, and Google Cloud lies in their business models and revenue diversification.
Hyperscale cloud providers offer a broad range of services—from data storage and databases to enterprise software, cybersecurity, and AI services.
In contrast, many Neocloud companies rely heavily on customers using GPU and AI computing to generate a significant portion of their revenue.
Therefore, if demand for GPU leasing declines, computing rental prices fall, or newer GPU technologies render older equipment obsolete, some Neocloud providers may face greater financial pressure than diversified cloud providers with multiple revenue streams.
However, each Neocloud company has unique capital structures, customer contracts, and risk profiles; thus, it would be incorrect to assume all face equal risk of failure.
According to a CNBC report on October 26, 2026, Elon Musk commented on X that in recent weeks, the bond market has largely shut down financing opportunities for companies lacking substantial revenue from businesses beyond AI and data centers.
Thus, Musk estimated that over 200 Neocloud companies could face failure or consolidation, with only a few surviving.
This warning followed remarks by Chamath Palihapitiya, founder of Social Capital, who suggested that by 2027, a large volume of AI computing power may have already been purchased and deployed—but may fail to generate sufficient revenue to offset its investment costs.
He believes 2027 could mark a major market correction, with companies that made poor investments forced to exit or merge.
However, the figure of over 200 companies is an estimate based on Musk’s commentary—not a confirmed count of bankrupt entities.
Moreover, SpaceX, the company Musk leads, owns its own infrastructure and computing businesses, giving it a competitive position in the AI computing market.Therefore, such warnings need to be analyzed alongside financial data and actual AI usage demands, rather than being interpreted as evidence that the entire Neocloud market is on the brink of collapse.
According to CNBC, following Musk’s warning, shares of Neocloud companies CoreWeave, Nebius, and IREN declined in early trading on October 26 before recovering and rising again by the end of the day.
Previously, CoreWeave and Nebius shares had dropped more than 30% over the course of about a month, while IREN shares fell approximately 40%.
Investor concerns stem not only from demand for artificial intelligence but also from companies’ capital structures, which require substantial borrowing to expand computing capacity.
In particular, CoreWeave is under close scrutiny after its credit default swap (CDS) costs and bond yields rose, reflecting market concerns about credit risk.
However, a decline in stock prices or a rise in CDS indicators does not necessarily mean a company has defaulted.
The market is assessing whether these companies can generate sufficient cash flow to support future investments and debt obligations.
AI data centers differ from traditional data centers because a significant portion of investment is not only allocated to buildings, power systems, and cooling, but also to high-performance GPUs and networking infrastructure, which are extremely expensive.
In the colocation hosting industry, operators can invest in buildings and supporting systems, then allow customers to install their own servers.
However, many Neocloud sites require direct investment in GPU clusters to enable customers to rent computing power directly.
Thus, risks have increased across multiple dimensions.
First, GPUs have rapidly evolving technology cycles. Newer models may offer higher performance per watt than older ones, reducing the competitiveness of previously invested GPUs.
Second, companies must invest before achieving maximum revenue. If customer utilization falls below expectations, GPU utilization rates will decline.
Third, borrowed investments require companies to pay interest and repay principal, even if rental revenue from facilities fails to meet targets.
Fourth, electricity, cooling, maintenance, and grid connectivity are ongoing operational costs.
Finally, price competition may intensify as multiple service providers simultaneously expand capacity upon entering the market.
These factors mean AI data centers are not merely digital real estate businesses—they are enterprises simultaneously exposed to technological, energy, and financial risks.
Assume Neocloud invests 1 billion baht in a GPU cluster and supporting systems, with an average cost of capital of 10% per year and an economic asset life of four years.
If the investment cost is amortized evenly over the project’s life, approximately 315 million baht in net cash flow per year would be required to recover the investment and generate the assumed return.
These figures do not include detailed assessments of taxes, residual value, or other risks.
If the company can lease GPUs at target maximum capacity, it may generate sufficient revenue to cover these costs.
However, if utilization drops from 90% to 50%, the number of billable processing hours could decrease by approximately 44% compared to the original scenario.
Meanwhile, investment costs and debt burdens do not decline at the same rate.
If GPU rental prices also fall simultaneously, the company could face severe cash flow pressure.
This example illustrates that Neocloud’s risk does not depend on the number of GPUs owned but on the ability to generate revenue from those GPUs over their investment lifecycle.
Installed GPUs that remain unused by customers will not generate the profits investors expect—even if the data center’s power and cooling infrastructure is fully operational.
While Elon Musk warns of potential failures among many Neocloud companies, NVIDIA CEO Jensen Huang argues that demand for Neocloud services remains extremely high.
According to a CNBC report from the GTC event in Berlin, Huang explained that a key challenge for the industry is a lack of working capital to meet the growing demand for artificial intelligence.
In other words, NVIDIA sees a market with strong demand for computing power, but service providers must invest enormous capital to rapidly build that capacity.
Meanwhile, executives from Nebius and IREN confirm that customer demand remains robust, with ongoing discussions about utilizing computing capacity in 2027 and 2028.
Thus, the differing perspectives between Musk and Huang reflect two potential risks that may emerge simultaneously.
On one hand, demand for artificial intelligence (AI) may indeed grow rapidly.
On the other hand, companies investing in AI infrastructure may face financial difficulties—even as the broader market continues to expand.
Rapidly growing industries do not necessarily mean every company within them will be profitable.The history of the technology industry has shown that market growth can coincide with the failure of some entrepreneurs.
If Musk’s and Palihapitiya’s predictions are at all accurate, 2027 could be the year when the AI infrastructure market begins to distinguish between truly competitive companies and those primarily expanding through capital investment.
Companies with long-term customer contracts, stable revenue, low cost of capital, and efficient power delivery capabilities may have a higher chance of survival.
Companies with high debt levels, insufficient contracted computing capacity, or high GPU and electricity costs may face liquidity risks.
Mergers and acquisitions could occur as larger companies acquire assets from financially distressed businesses.
In some cases, data centers and GPUs previously invested in may be sold below their original book value.
On the other hand, mergers and acquisitions could enable financially stronger enterprises to expand processing capacity at lower costs.
Thus, the restructuring of the Neocloud market does not necessarily mean the end of the AI industry—it could also be a process of competitive realignment.
By Methangkorn Sermsuk