Heavy neocloud debt loads funding artificial intelligence data centres have triggered sharp market declines across specialised infrastructure providers, as investors and credit markets question whether private operators can sustain tens of billions of dollars in borrowing if major technology customers trim their capital spending.
Key points
- Shares in specialised data centre operators including CoreWeave, Nebius and IREN have fallen by about 25% over the past month as debt markets demand higher yields.
- Lenders have backed massive hardware purchases by using long-term client contracts and graphics chips as collateral, leaving market leader CoreWeave holding $35.1bn in debt.
- Gartner analysts forecast that 80% of neocloud providers will cease independent operations by 2030 as borrowing costs climb and industry consolidation accelerates.
- Project cancellations and delayed share offerings in the UK and Australia point to growing operational friction around electrical grid access and equity valuations.
How borrowing built the neocloud sector
Neocloud operators build and run warehouse-scale data centres packed with specialised processors designed to train and run machine learning models. Instead of offering general cloud services like Amazon Web Services or Microsoft Azure, these firms buy high-end chips in bulk and lease compute capacity directly to artificial intelligence developers such as OpenAI, Anthropic and Meta.
Because advanced graphics processing units must be paid for upfront, operators depend heavily on debt to expand their footprints. Companies sign multi-year capacity agreements with technology clients and use those forward commitments, alongside the physical processors, as collateral to secure loans from private credit funds and commercial lenders.
Cory Johnson, an analyst at financial research firm The Quiet Period, said lenders frequently assess the credit standing of corporate clients rather than the balance sheet of the infrastructure firm itself. That underwriting pattern helped CoreWeave accumulate $35.1bn (£26.3bn) in debt despite operating at a fraction of the corporate scale of the hyperscale tech giants it serves.
This heavy reliance on debt across the supply chain comes as public markets reconsider the return on artificial intelligence spending. A recent tech stocks slide over AI capex concerns highlighted how fast equity values fall when corporate capital commitments outrun near-term business revenue.
Melissa Otto, head of Visible Alpha Research at S&P Global, described the aggressive expansion as an arms race between competing infrastructure groups. Otto said “There is just an enormous amount of spending getting injected into the system” as operators rush to meet demand from foundational model developers.
Crypto roots and the rush for compute
Many prominent neocloud data centres did not start as dedicated artificial intelligence providers. Several major operators trace their origins to the cryptocurrency mining sector, where companies constructed specialised warehouses designed to supply intense electric power and industrial cooling to process digital tokens such as Bitcoin and Ethereum.
CoreWeave began in 2017 as an Ethereum mining venture. Competitors including Core Scientific, Applied Digital and IREN also built their initial electrical connections and physical facilities for cryptocurrency networks. When generative artificial intelligence expanded rapidly in late 2022 following the launch of OpenAI’s ChatGPT, these companies owned physical real estate that could be converted to house enterprise artificial intelligence accelerators far faster than traditional commercial property developers could build new facilities.
Nebius entered the market along a different path. Formerly known as Yandex NV, the company operated as the Dutch parent entity of the Russian technology business before divesting its domestic assets and changing its name in 2024 to focus entirely on artificial intelligence cloud services.
The rapid pivot allowed neocloud data centres to secure urgent business from developers who could not get enough chip capacity from established hyperscalers. However, research analysts warn that relying on debt-funded, capital-intensive hardware leaves these businesses exposed whenever commercial lending conditions deteriorate.
What happens if Big Tech cuts back spending?
The central financial risk for neocloud data centres is their exposure to a small group of large technology customers. When major corporate clients trim their expansion budgets, the specialised operators that purchased expensive hardware with borrowed funds face immediate cash pressures.
Investor unease spread across the sector after Microsoft and Google reported capital expenditure projections that came in below Wall Street estimates. While Microsoft continues to tap public bond markets to support data centre construction, including a recent Microsoft bond sale for AI expansion, any pause in external leasing plans quickly impacts independent operators. Credit rating agencies are also monitoring balance sheets closely, with S&P recently revising its Meta credit rating outlook to negative over sustained artificial intelligence expenses.
Charlie Dai, vice president and principal analyst at Forrester, said the core vulnerability of the neocloud model is higher financial risk whenever client spending plans adjust. Bradley Tusk of venture capital group Tusk Ventures said that if current artificial intelligence growth projections stall, specialised cloud providers will suffer severe collateral damage.
Equity markets have reacted to these balance sheet risks with steep sell-offs. Shares in CoreWeave, Nebius, IREN and Applied Digital have each fallen by roughly a quarter or more over the past month. At the same time, private credit lenders have demanded higher interest rates before underwriting additional hardware purchases.
CoreWeave chief executive Mike Intrator defended his company’s position, saying the market is behaving irrationally and expressing confidence in its commercial agreements. Intrator said large enterprise clients continue to queue up for computing infrastructure and specialised software services.
Consolidation pressures and grid constraints
Rising interest rates and physical infrastructure constraints are also slowing down data centre rollouts in multiple international markets. Debt service costs increase as existing loans mature, while regional electricity grids face severe delays in connecting power-hungry server halls.
In the United Kingdom, Microsoft cancelled an infrastructure agreement with British neocloud firm Nscale due to electrical supply constraints that could extend through 2030. In Australia, data centre firm Firmus recently abandoned plans for an initial public offering after market sentiment cooled. These delays coincide with rising community scrutiny over data centre electricity consumption and water usage.
Elon Musk warned that rising borrowing costs could cause more than 200 firms to “see guppy-level death and consolidation”, leaving only a handful of well-capitalised operators standing. Analysts at Gartner predict that 80% of neocloud companies will cease independent operations by 2030, while Forrester expects at least three operators to be acquired in the short term as cash balances shrink.
Some analysts argue the sector will find support from basic market fundamentals. Analysts at Yorkville Ives said total demand for high-performance computing still outstrips available capacity, pointing out that specialised neocloud data centres fill an essential supply gap that hyperscalers cannot satisfy alone.
United States Treasury Secretary Scott Bessent also downplayed wider systemic risks during an appearance on CNBC, characterising neocloud debt as a limited component of the broader financial ecosystem. Industry attention is now turning to upcoming public share sales from OpenAI and Anthropic, whose commercial capacity contracts remain central to the debt structures funding these specialised sites.
