CoreWeave closed at $53.74 on the Nasdaq on November 3, 2026, up 7.16% after an 8.60% slide the session before. The bounce barely dents months of losses. Shares in the US company, which rents out cloud infrastructure for artificial intelligence, sit 61.1% below their 52-week high of $138.25, set on May 6, even as the business keeps growing at explosive rates.

Contracted work in the pipeline tops $100 billion, and demand for computing power should climb further. The investment needed to serve it dwarfs revenue. CoreWeave is now worth about $29.6 billion, and some of its bonds yield around 15%, a measure of what the credit market charges for the risk in this expansion.

A new market low and Cathie Wood’s buying

The decline steepened in October as investors worried about interest rates and about how the most heavily indebted players in AI will fund their data centers. The stock fell 15.46% on October 14 and 11.91% two sessions later, on October 16. On November 2 it slid to $49.46, a new 52-week low, before closing at $50.15.

The $141.79 peak of October 28, 2025 has now dropped out of the one-year comparison. Further back still is the all-time intraday record of $187, reached on June 20, 2025, from which the stock has shed 71.3%.

At the depths of the sell-off, in the sessions of October 14 and 19, Cathie Wood’s ARK Invest funds bought 321,163 CoreWeave shares for roughly $20.4 million.

Revenue up 112%, but the loss widens to $626 million

The latest figures cover the second quarter of 2026. Revenue reached $2.575 billion, up 112% year over year, with adjusted EBITDA of $1.510 billion and a 59% margin. The net loss widened to $626 million over the same period, more than double the $290 million of a year earlier.

Infrastructure depreciation and net financial charges drive the red ink. Together they reached $640 million in the quarter. First-half losses totaled $1.366 billion, and while operations generated $3.663 billion in cash, investments absorbed $14.874 billion.

For full-year 2026 the company expects revenue of $12.4 billion to $13.2 billion and capital spending of $35 billion to $39 billion. That mismatch ties its fate to its ability to keep raising money in the markets.

CoreWeave’s debt and a 2031 bond yielding about 15.4%

As of June 30, CoreWeave carried roughly $35.1 billion in financial debt, plus $16.3 billion in operating lease liabilities. The repayment schedule published on that date showed $4.413 billion due by the end of 2026 and a further $6.184 billion in 2027, about $10.6 billion in all. New issuance or repayments may have changed those figures since.

A bond with a 9.75% coupon maturing in 2031 was changing hands at around 81 cents on the dollar on November 3, with a yield to worst close to 15.4%. That is not the average cost of the debt CoreWeave pays today, but it gauges what the company would pay to borrow again on an unsecured basis.

In September the company placed $4.2 billion of convertible bonds with a 2.875% coupon maturing in 2033, and set up a program to sell up to 35 million new shares into the market. Each new share dilutes existing holders. The convertibles come with hedges that limit the dilution.

The $104 billion order book and Nvidia’s backing

The backlog, meaning revenue already under contract, stood at $104.2 billion as of June 30, more than triple the $30.1 billion of a year earlier. More than $25 billion in net new commitments were added in early July, though these turn into revenue only as data centers are built and brought online.

The flip side is dependence on a handful of large customers. The top three alone accounted for 72% of revenue in the second quarter. Nvidia acts as a safety net: both a supplier to and a shareholder in CoreWeave, it has committed to buying any capacity left unused under an agreement initially worth about $6.3 billion.

Contracted power capacity rose to roughly 4.2 GW in September from 3.7 GW at the end of June. Late October brought the announcement of new capacity in Denmark, and on November 2 the company extended its partnership with CrowdStrike to protect customers’ AI workloads.

HSBC targets $39, Cantor Fitzgerald $176

Across the latest available recommendations tracked by TipRanks, analysts split 22 Buy, 5 Hold and 3 Sell, with an average twelve-month target of $130.92. Against the November 3 close that works out to theoretical upside of 143.6%, a gap that says as much about Wall Street’s confidence as it does about the market’s skepticism.

The extremes are almost diametrically opposed. HSBC rates the stock a sell with a target of $39, 27.4% below current prices. Cantor Fitzgerald goes as high as $176, which would mean a 227.5% gain.

October brought a flood of cuts in between. Raymond James nearly halved its target on October 26, from $125 to $75, while keeping a Buy rating. Three days earlier UBS had come down from $120 to $105, Piper Sandler from $153 to $124 and BNP Paribas Exane from $169 to $117, all with positive ratings. Goldman Sachs also cut its estimate, from $139 to $106, while raising its recommendation to Buy.

Barclays confirmed Hold with a $105 target on October 27, and Morgan Stanley stopped at Equal Weight and $70. Bank of America is more upbeat, reiterating Buy and $140 on October 22. Measured against the November 3 close, these targets imply potential gains ranging from Morgan Stanley’s 30.3% to Bank of America’s 160.5%, taking in Raymond James at 39.6%, UBS and Barclays at 95.4%, Goldman Sachs at 97.2%, BNP Paribas Exane at 117.7% and Piper Sandler at 130.7%.

Technical picture: the stock stays below its moving averages

At the November 3 close the stock traded below all the main simple moving averages: 17.3% under the 20-day ($65.01), 31.1% under the 50-day ($78.05) and 40.9% under the 200-day ($90.89). The 14-day RSI, around 35, confirms the weakness but is not yet low enough to call a bottom.

The $49.46 low of November 2 is the first support to watch. On the upside the stock would need to reclaim the $55-$56 area first, then the $59-$60 and $64-$65 bands. A genuine reversal would require a series of higher highs and higher lows.

November 10 earnings put liquidity under the microscope

CoreWeave has confirmed it will report third-quarter results on November 10, 2026, after the close, with a conference call at 5 p.m. New York time. That moves up the indicative date of November 16 that had previously circulated.

Beyond the data center build-out and the new contracts, investors will watch how much cash the investment program is burning relative to the financial resources available, the real test for a stock the market has already punished hard.

Editor’s note

This article was originally published in Italian on money.it by Gerardo Marciano on November 04, 2026 as «CoreWeave crolla del 61% dai massimi. Cathie Wood compra e gli analisti vedono un rialzo del 143,6%». It has been translated and adapted for an international audience by the Money.it International desk.