Nvidia CEO Jensen Huang said Thursday during the GTC conference in Berlin that market demand for computing power from “neocloud” operators is “extremely strong,” but what is really holding back these emerging cloud service providers’ expansion is not orders — it is the working capital needed to support growth.
Neoclouds are emerging cloud operators that buy large quantities of servers equipped with AI chips and then lease the computing power to AI developers. Huang described these companies’ technical capabilities as “extraordinary,” saying their fundamentals are completely sound and the problem lies in the middle of the funding chain: “The missing piece in the middle right now is the working capital needed to fuel the growth of these companies… because demand is so strong across every area of artificial intelligence, there is a shortage of capital.”
He did not specify a concrete solution or timeline, but said with conviction that the scale of capital required by the AI industry far exceeds the financing companies can currently obtain, and “this problem will ultimately be solved.”
Notably, even as Huang remains bullish on the industry’s prospects, capital markets have turned noticeably colder on the neocloud group. Shares of operators such as CoreWeave (CRWV) and Nebius (NBIS) have continued to weaken recently. In Thursday’s U.S. premarket session, Nebius fell about 3% to 3.6%, while CoreWeave and IREN (IREN) dropped 4%, and Applied Digital (APLD) also fell nearly 4%.
The selling pressure stems from two concerns. First, investors are beginning to question whether the massive capital expenditures on AI infrastructure will take longer to pay back than previously expected. Second, there are doubts about financial structure — CoreWeave, one of the largest neocloud operators, is financing data center construction with substantial borrowing, and its leverage level has raised market caution.
Neocloud OperatorThursday Premarket Stock MoveNebius (NBIS)Down about 3%–3.6%CoreWeave (CRWV)Down about 4%IREN (IREN)Down about 4%Applied Digital (APLD)Down nearly 4%
Note: Stock moves reflect Thursday U.S. premarket trading data; different reports cited slightly different declines.
For Nvidia, any operational troubles among neocloud operators would be a double blow: AI chip sales could shrink as a result, and the financial support Nvidia previously provided to these operators could also face losses. The risk could also spread along the AI supply chain and affect the broader AI ecosystem.
Beyond money, power is also a bottleneck. Huang noted that electricity supply is one of the main constraints on the AI boom, especially in Europe, where power prices are relatively high. With Europe’s energy supply already tight, how to allocate limited electricity to neocloud companies such as Nebius and Mistral is crucial to their growth.
He further outlined the path for Europe to develop an AI economy: foreign investment can be brought in, but the precondition is that Europe must first have its own infrastructure. If local neocloud operators can build a computing foundation in Europe, an AI ecosystem and AI economy can grow on top of it, setting the growth flywheel in motion in Europe and ultimately forming a complete closed loop that keeps the results in Europe. He said bluntly: “All of Europe needs more neocloud service providers.”
Nvidia is stepping up its push into Europe’s AI industry this month, and Huang has been giving密集 interviews to local media. The company has committed nearly $10 billion (about NT$330 billion) to European companies, universities and research institutions. During a visit to London last week, it announced stakes of more than $1 billion (about NT$33 billion) in five British startups, and it is also investing in several French startups.
Huang said Nvidia is working to kick-start Europe’s AI economy because Europe has a wealth of advanced technology, deep industrial knowledge and experience, and outstanding scientific and technical strength. “Europe is ready to participate in building and developing the AI economy here,” he said.
From an industry-chain perspective, Huang’s remarks essentially frame the neocloud group’s problem as a “capital mismatch” rather than “fake demand.” Judging by recent stock moves, the market is clearly more focused on the payback efficiency of capital expenditure and debt risk than on whether end-user demand for computing power is solid. That also means whether neocloud operators can obtain cheaper, longer-term funding will be a key variable determining whether this wave of AI infrastructure investment can be smoothly converted into revenue. And for Nvidia, its European buildout and equity investments in startups are, to some extent, precisely filling in this missing segment of the funding chain for its downstream customers.