CoreWeave Earnings: It's Losing Money — So Why Did the Stock Jump 16%?

 

• $104B in Booked Revenue: CoreWeave's Answer to the "AI Bubble" Crowd
• The Company OpenAI and Microsoft Line Up For: CoreWeave Q2 Earnings, Fully Explained
• Inside the $104B Backlog: Breaking Down the Neocloud Leader
• First to Run NVIDIA's Newest GPUs: The Secret Behind CoreWeave's Blowout Quarter

📌 The Scorecard in Three Lines

On August 11, 2026 (U.S. Eastern Time), AI-focused cloud provider CoreWeave (Nasdaq: CRWV) released its fiscal Q2 2026 results (April–June). Nearly every headline metric came in above Wall Street's expectations, and the stock jumped roughly 16% in after-hours trading right after the announcement.

Here's the scorecard, based on the official SEC filing.

Metric	Actual	Consensus	Verdict
💰 Revenue	~$2.58B ($2,575M)	~$2.56B	+112% YoY, beat
📉 Loss per share (EPS)	-$1.14 (GAAP)	~-$1.49	Smaller loss than expected
📊 Adjusted EBITDA margin	59%	~55.4%	Beat
🎯 Full-year guidance	Raised (~$12.4B–$13.2B)	—	Guidance up
📦 Revenue backlog	~$104B	—	+46% YoY

One clarification: the "-$1.03" EPS shown on the summary card is a non-GAAP estimate, while the actual GAAP loss per share was -$1.14. The "$2.56M" expected revenue figure is a units typo; the correct number is $2.56B. Everything below follows the official filing.

🏢 CoreWeave — You've Heard the Name, But What Does It Do?

Before digging into the numbers, let's cover the company itself. Understanding what CoreWeave is makes it far clearer why this news made such a splash.

Founded in New Jersey in 2017, CoreWeave is an AI-specialized cloud company that went public on the Nasdaq in March 2025. Its tagline is "The Essential Cloud for AI."

Let's unpack a few unfamiliar terms first.

☁️ The cloud refers to renting computing power from massive remote data centers over the internet, rather than installing software on your own machine. Amazon (AWS), Microsoft (Azure), and Google (GCP) are the established giants here.

🧠 So what sets CoreWeave apart? If Amazon and Google are "all-purpose" clouds that handle every kind of task, CoreWeave is a "specialist" cloud built solely for AI computation. This new breed of AI-only providers is known in the industry as neoclouds — a term for companies that rebuilt the cloud playbook around GPUs for the AI era.

🎮 Why do GPUs matter so much? A GPU (graphics processing unit) was originally the chip that rendered game visuals, but it excels at crunching enormous volumes of calculations in parallel. That makes it perfect for training and running AI like ChatGPT — so much so that GPUs are now called "the oil of the AI age."

In short, CoreWeave's business model is straightforward: secure top-tier NVIDIA GPUs in bulk, install them across data centers, and rent them out to AI companies by the hour. OpenAI, Microsoft, and NVIDIA are among the heavyweight names on its client roster.

🔍 So Why Did This News Blow Up?

There's solid reasoning behind the double-digit stock pop and the breaking-news treatment. Follow the points management emphasized on the earnings call and the picture sharpens.

📈 First, revenue more than doubled in a single year. This quarter's $2.58B is over twice the $1.21B posted a year earlier — a 112% increase. The company said demand for AI cloud infrastructure is "exceptionally strong" and is spreading across industries, regions, workload types, and GPU generations rather than concentrating anywhere. In fact, immediately available GPU capacity is essentially sold out, which management views as a springboard to grow market share over the coming years. A quick note: a workload is a "chunk of work" a computer must handle — training an AI or having a chatbot generate an answer each counts as one.

🤖 Second, managed inference exploded. A standout moment on the call was the announcement that the managed inference business grew 100x in just one quarter. It helps to know what inference is. Building an AI splits into two phases: training, where you feed it massive data to make it smart, and inference, where the trained AI actually produces answers to questions. The moment you ask ChatGPT something and get a reply is inference. As AI services spread, inference demand surges — and CoreWeave is posting the steepest curve here. "Managed" means CoreWeave handles the operations so customers don't have to fuss with complex setups.

📦 Third, "future revenue" is piling up like a mountain. The term backlog may be unfamiliar; it refers to revenue that's already locked in through signed customer contracts. Think of it as the total value of reservations a restaurant has already booked — the food hasn't been served (service not yet delivered), but the money is essentially guaranteed. CoreWeave reported that this backlog reached $104B, up 46% year over year, as AI cloud demand exploded. Even more striking: that figure doesn't yet include over $25B in new contracts added early in Q3. For a company with annual revenue still in the ~$13B range, holding contracts worth eight times that is remarkable. This "revenue visibility" is exactly what has investors excited.

🎯 Fourth, the company raised its own targets. Guidance is a company's self-issued forecast of its results, so raising it is effectively management publicly declaring, "We expect to do even better than we thought." CoreWeave lifted its annual guidance and its year-end target metrics as well. A 59% adjusted EBITDA margin added fuel to the rally.

💵 EBITDA measures operating cash-generating power before interest, taxes, and depreciation. Capital-heavy businesses like data centers often look unprofitable on a net basis due to depreciation, but EBITDA reveals whether the core business is actually making money once equipment factors are stripped out. A 59% margin signals a very healthy underlying profit engine. Management also acknowledged that component and equipment procurement costs are rising, yet noted that the value each GPU generates is climbing faster than costs — so margins are actually widening. As a result, contribution margin rose roughly 5–10 percentage points from the prior quarter. 📐 Contribution margin is revenue minus the variable costs directly tied to generating it — a gauge of "how much is actually left when you run one set of equipment."

🧩 Three Things Easy to Miss from the Summary Card

There are three points that a quick skim of the summary might gloss over, yet they represent the real substance of these results.

💡 First, don't be fooled by the "loss" on the surface. This quarter's net loss was $626M, actually larger than a year ago. On the surface it reads as "they lost even more money," but the internals tell a different story. The main culprits behind the loss are interest expense ($640M) and depreciation (~$1.39B). CoreWeave took on heavy debt to build GPUs and data centers (hence the interest), and it writes off purchased equipment in annual slices for accounting purposes (depreciation). In other words, this loss stems not from a failing business but from aggressive investment in the future. In fact, operating cash flow flipped to a positive $679M this quarter.

🤝 Second, the customer base is broadening from "mostly AI labs" to "mainstream enterprises." CoreWeave had long been criticized for over-reliance on a handful of AI giants like OpenAI. This quarter, though, a wave of enterprises across industries joined — Caterpillar (heavy machinery), Bentley Systems (design software), Grammarly (writing AI), and Isomorphic Labs (Google's drug-discovery subsidiary), among others. As management put it, CoreWeave sits at a "crossroads of information flow," serving both hyperscalers (mega-scale cloud firms) and ordinary enterprises, letting it flexibly handle any task — from large-scale training to real-time responses. A thicker customer base means steadier revenue.

🏆 Third, it earned a "world-first" badge on the technology front. CoreWeave became the first cloud provider in the world to build and validate NVIDIA's next-generation platform, the Vera Rubin NVL72. Simply put, it got NVIDIA's just-released, cutting-edge AI supercomputing gear up and running ahead of rivals. The effort also incorporated data center operating techniques dubbed "rack management innovation." That translates into a premium as "the company that gets NVIDIA's newest weapons first." ⚡ On top of that, it added roughly 500MW (megawatts) of active power capacity this quarter alone, bringing the total to 1.5GW (gigawatts), while maintaining its plan to secure at least 8GW by 2030. It also rolled out seven new AI platform features, and cumulative tracked model training runs surpassed one billion. But why is power capacity a performance metric?

Running tens of thousands of GPUs demands enormous electricity. So in the AI cloud world, "secured power" equals "the total compute capacity you can sell going forward." At 8GW — comparable to several large nuclear power plants — power effectively sets the ceiling on revenue.

🎤 Management's Message on the Call

CEO Michael Intrator described the quarter as "a critical inflection point where our scale has finally begun translating into expanding operating leverage." 🔧 Operating leverage refers to a structure where profits grow faster than revenue because costs don't rise in lockstep. Once you've built a data center, taking on more customers barely adds cost — so margins improve as scale grows. The CEO's remark expresses confidence that "now that we're large enough, our earnings will grow far faster from here."

Management also voiced strong conviction in the long-term ROI (return on investment) of the products underpinning its industry-leading AI cloud services. The logic: "Once we deliver a cluster (a bundle of GPUs) to a customer, the contracted revenue lands predictably, and on that foundation we can pay down debt while generating enormous free cash flow." The "Still call it an AI bubble?" line on the breaking-news screen reads as a confident rebuttal — answering the "AI bubble" skeptics with hard results.

🧭 Wrapping Up — How to Read These Results

CoreWeave's latest scorecard proved with numbers that demand for AI infrastructure hasn't cooled. Doubled revenue, a $104B backlog, and raised guidance all lined up, and the market responded immediately.

⚠️ That said, a cool-headed view still matters. The company remains in the red, rooted in heavy debt and interest burdens. Because sustaining growth requires continually raising more debt, it also carries the risk of a shifting rate environment or a downturn in GPU demand. Ultimately, how smoothly that dazzling backlog converts into actual revenue and cash will be the next thing to watch.

Note: This article is for informational purposes and is not investment advice recommending the purchase or sale of any security. The final decision and responsibility rest with the individual investor.