Neocloud short sellers rack up $6 billion as AI debt fears deepen

Short selling is a trading strategy built around a falling stock price. An investor borrows shares and sells them at the current price. Later, the investor tries to buy those shares back for less and return them to the lender. The difference, after costs, is the potential profit. If the price rises instead, the short seller can lose money.

CoreWeave provides a clear example. Its short interest reached 23.7% of free float, while its shares fell almost 30% this month. Reuters reported that short sellers made nearly $1.7 billion in paper profits from that decline. The profit is called “paper” because it reflects an unrealised gain while the position remains open.

Short selling can express a negative view about a company’s valuation or business outlook. Here, investors are questioning rapid, debt-backed data-centre expansion. Borrowing remained cheap for the five companies, which may help bearish positions stay open. These positions can also add pressure when many investors sell.

The bearish positioning is large by the measure used in the article. ORTEX said around $17 billion worth of stock across CoreWeave, IREN, Nebius Group, Applied Digital and Core Scientific was currently sold short. Short interest in each company stood between about 20% and 26% of its free float. Free float means shares available for public trading.

CoreWeave’s short interest rose to 23.7% from 19.6% at the start of October. Nebius reached 20% of free float, compared with 15.4% at the start of the year. Short sellers also made more than $1.2 billion from positions against IREN and Applied Digital combined. Core Scientific generated another $115 million in paper profits for bearish traders.

The pressure remained visible in trading. CoreWeave and Nebius fell more than 4% early Thursday. IREN dropped 6.1%, Applied Digital declined 3.7%, and Core Scientific slipped 2.4%. Cheap borrowing may allow these positions to remain open.

The concern is about the economics of expansion, not simply demand for AI services. Companies seen as major AI infrastructure beneficiaries are building data-centre capacity while using more debt. Investors are asking whether the resulting business can support those obligations. If expected revenue or cash generation falls short, debt can make a growth plan more vulnerable.

The article gives no company-by-company debt totals or cash-flow forecasts. It does show the market’s response. Short interest reached 23.7% of CoreWeave’s free float and 20% of Nebius’s. Across five companies, roughly $17 billion of stock was sold short. The article links these bearish bets to doubts about rapid, debt-backed expansion.

That scrutiny has changed trading quickly. CoreWeave and Nebius fell more than 4% in early Thursday trading, while IREN dropped 6.1%. Borrowing remained cheap for all five companies, potentially keeping bearish positions in place. The broader implication is that investors may demand stronger financial support for future expansion.

In general, a data-centre company needs revenue and cash flow from its facilities to cover operating costs and debt obligations. If new centres underperform, the company may have less cash available for interest or repayment. It might then need to delay projects, raise additional money, refinance, or sell assets. These outcomes are risks, not stated events in this article.

The article’s central warning is about “rapid, debt backed expansion.” Investors are questioning whether the economics can support the build-out. That means the key mechanism is a gap between the cost of new capacity and the money it ultimately produces. The source does not provide revenue targets, debt totals, or repayment schedules for the companies.

Market consequences can arrive before any missed payment. Investors may mark down shares if they expect weaker returns or greater financing pressure. This article shows that response: CoreWeave shares fell almost 30% this month, while several peers also declined on Thursday. Short sellers benefited as prices dropped.

Short sellers are market participants who borrow shares, sell them, and hope to repurchase them later at a lower price. Their potential profit comes from the price difference. Their potential loss comes when the stock rises instead. Short selling therefore reflects a bearish view, although it does not prove that a company’s outlook will worsen.

The companies in this article are associated with the AI infrastructure boom. They include CoreWeave, IREN, Nebius Group, Applied Digital and Core Scientific. Investors may target them because the article says their expansion is rapid and debt-backed. That combination creates a specific question: can new data-centre capacity generate enough economic value to support the financing behind it?

The positioning was substantial. Around $17 billion of their stock was sold short, and short interest ranged from about 20% to 26% of free floats. Cheap borrowing may let bearish positions remain open. As shares fell, short sellers recorded large paper profits, including nearly $1.7 billion on CoreWeave this month.

A data centre is a specialised facility that houses computer servers and the systems that keep them operating. It provides computing capacity for digital workloads. AI workloads are especially demanding because they use large amounts of processing power and data. They also require many machines to work together, so fast connections between equipment are important.

The source does not define data centres or list their equipment. In general, an AI facility needs computing hardware, buildings, electricity, cooling and networking. The hardware performs calculations, while networking moves data between machines. Electricity powers everything, and cooling removes heat. Without these supporting systems, the computing equipment cannot run reliably at scale.

This helps explain why AI infrastructure companies face financial scrutiny. Building capacity requires rapid expansion of physical facilities, not just software changes. The article says investors are questioning whether this expansion can be supported by rising debt. It also reports pressure across five data-centre-related stocks, showing how concerns about infrastructure economics can affect market sentiment.

AI data centres require large upfront investment because several systems must be built or purchased together. Companies need suitable buildings, computing equipment, electricity connections, cooling systems and networking. These costs arise before a facility can generate revenue. Rapid expansion therefore requires substantial capital, which can include debt. The source specifically refers to debt-backed data-centre expansion.

Each part supports the others. Computing equipment performs AI calculations. Electricity powers it, while cooling manages the heat it produces. Networking connects machines so workloads and data can move efficiently. Buildings contain and protect the equipment. If one supporting part is missing or undersized, the facility may not deliver its planned capacity. These mechanisms are established background, not details provided by the article.

The financial question is whether the completed centres earn enough to justify their cost and support their financing. Investors are increasingly asking that question about the companies discussed. Around $17 billion of their stock was sold short, and borrowing remained cheap, potentially sustaining bearish positions. Their share declines show how quickly sentiment can change.