CoreWeave grew 112% and still lost $626m — here is what that teaches you

When a company reports earnings, most people instinctively look at one number: profit. If it is up, the company is doing well. If it is down, something has gone wrong. But that rule breaks down quickly when you encounter a company like CoreWeave, whose second-quarter 2026 results contained two figures that seem to contradict each other — and yet both tell an important part of the same story.

CoreWeave is an AI cloud computing company. It rents out computing power — specifically, racks of expensive specialist processors called GPUs — to the companies building and running artificial intelligence systems. In the second quarter of 2026, it generated $2.58 billion in revenue, up 112% from the same quarter a year earlier. Yet it posted a net loss of $626 million, and that loss was wider than the loss it recorded a year before. Far from panicking, many investors responded positively. To understand why, you need to understand the difference between revenue, profit, and the particular way capital-intensive businesses grow.

What CoreWeave actually does

CoreWeave began as a cryptocurrency mining company and pivoted around 2019 to focus on renting out GPU computing power to artificial intelligence researchers. It turned out to be one of the most well-timed strategic shifts in recent technology history. As AI became the dominant investment theme of the mid-2020s, demand for exactly the kind of computing CoreWeave provides exploded.

Its customers are companies that train and run large AI models — the kind of work that requires thousands of the most powerful computer chips available, running for months at a time. Building and operating that infrastructure themselves would cost each AI company billions. Renting it from CoreWeave is cheaper, faster, and more flexible. CoreWeave listed on the Nasdaq stock exchange in early 2025 in one of that year’s highest-profile initial public offerings, or IPOs.

By the time it reported Q2 2026 results on 11 August, CoreWeave had confirmed deals with several of the largest technology companies in the world, including a $21 billion agreement with Meta to supply AI cloud capacity through 2032, and a separate multi-year contract with Anthropic to provide the computing power for its Claude AI models.

Revenue and profit are not the same thing

This distinction sounds basic, but it is frequently misunderstood — and CoreWeave’s results make it vivid.

Revenue is the total money customers pay you. Profit is what remains after you have paid all your costs. The gap between them is your expenses. If your expenses are less than your revenue, you make a profit. If they are greater, you make a loss.

CoreWeave’s revenue was $2.58 billion. But running and expanding its business cost significantly more than that. The two main cost drivers were:

• Capital expenditure on infrastructure — CoreWeave is building data centres filled with Nvidia H100 and H200 GPUs, the specialist chips that power AI training. These are extraordinarily expensive: a single Nvidia H100 chip costs roughly $30,000, and a large AI training cluster needs tens of thousands of them. CoreWeave guided investors to expect between $35 billion and $39 billion in capital expenditure for the full year 2026.
• Debt financing costs — because CoreWeave has borrowed heavily to fund its rapid build-out, it pays significant interest on that debt. Those interest payments flow directly through the income statement as a cost, widening the reported loss even when the core business is growing strongly.

In short, CoreWeave is spending enormous amounts of money now — building data centres, buying chips, taking on debt — in order to generate revenue for years into the future. During that build-out phase, losses are not a sign that the business is broken. They are the cost of growing very fast.

What a backlog means for investors

One of the most important numbers in CoreWeave’s results was not the revenue or the loss figure. It was the revenue backlog: approximately $104 billion as of the end of June 2026, excluding a further $25 billion or more in new customer commitments secured in the early weeks of the third quarter.

A backlog is the total value of contracts that customers have already signed but that CoreWeave has not yet delivered on. Think of it like a restaurant that has taken bookings months in advance. It has not yet served those meals and received payment, but it knows those customers are coming. For investors trying to work out whether a company’s rapid spending will eventually generate proportionate returns, the backlog is critical evidence. It demonstrates that customers have already committed to pay for the capacity CoreWeave is building right now.

That $104 billion backlog, against roughly $10 billion in annual revenue the company was tracking towards, meant CoreWeave had multiple years of future revenue already locked in through signed contracts. That transforms the picture of the business: it is not simply borrowing and spending on the hope that customers will appear. The customers have already signed up, and the infrastructure is being built to service them.

Adjusted figures and what to pay attention to

As with many company results, CoreWeave reported both a headline loss and an adjusted figure. The adjusted loss per share for Q2 was $1.03 — better than the $1.20 loss per share that analysts had been forecasting. This beat on the adjusted measure was one reason the market reacted relatively well.

Adjusted figures strip out certain costs — often share-based compensation paid to employees in the form of stock, or one-off charges — to give a picture of the recurring operating performance. Critics argue that these adjustments can be used to present an artificially flattering view of a company’s finances. The honest answer is that they are useful, but should always be read alongside the full reported figures. The full net loss of $626 million is real cash that left the business. The adjusted figure helps strip out noise to show the trend in underlying operations. Both matter. A good investor reads them together.

For more on how to interpret these two measures when reading any company’s results, the post on why shares move on earnings news covers this in detail.

Why investors were broadly comfortable

Growth investors — those who focus on companies with the potential for rapid future earnings rather than strong current profits — evaluate businesses differently from income or value investors. Rather than asking “how much did this company earn last quarter?” they ask “how large could this company become, and is the price I am paying today reasonable given that potential?”

In CoreWeave’s case, the investment case rests on three pillars:

1. The market is enormous and growing rapidly. AI infrastructure spending is expected to scale dramatically through the late 2020s. CoreWeave is one of only a handful of companies capable of providing large-scale specialised AI compute at short notice.
2. The contracts are long-term and binding. A $21 billion deal with Meta through 2032 is not the same as a month-to-month customer. It is contracted, recurring revenue that is much harder to cancel than a typical cloud subscription.
3. The losses are the cost of speed. Building data centres takes money. Borrowing to build them faster than competitors costs even more money. But if the AI infrastructure market is winner-takes-most over the next decade, the cost of falling behind competitors now is far higher than the cost of the current losses.

That reasoning might turn out to be wrong — many fast-growing companies that seemed to have unassailable positions later found their business models disrupted. But it is the reasoning the market was applying. To understand the difference between this kind of growth investment and a more traditional value approach, the post on growth vs value stocks explains the two styles and when each tends to perform well.

A note on risk

It would be incomplete to explain CoreWeave’s results without noting the risks that the same figures highlight. The company carries very large amounts of debt, on which it pays significant interest. Its business is almost entirely dependent on the continued rapid growth of AI infrastructure spending — any slowdown in that trend, or any loss of major customers, would immediately affect results. It also relies heavily on a single supplier for its most critical hardware: Nvidia. If Nvidia faces production constraints, if chip prices fall sharply, or if a competitor develops a chip that makes Nvidia’s products less attractive, CoreWeave’s costs and pricing power could shift overnight.

None of these risks means CoreWeave is a poor business. But they are the kinds of questions a thoughtful investor asks whenever a company’s growth story depends on many future assumptions all proving correct simultaneously. In the US earnings season for Q2 2026, CoreWeave was one of the most talked-about results — and understanding it helps you read the broader context of how that earnings season unfolded.

What this means in the Student Investor Challenge

The Student Investor Challenge uses a virtual portfolio of £100,000 invested in real shares. Understanding how to evaluate a company like CoreWeave — which may not be directly available in the challenge’s stock list — still matters, because the same principles apply to any growth company you might hold.

When a company you hold reports results, practise asking these questions:

• What is the gap between revenue and profit, and why? Is the company investing heavily to grow, or is it simply struggling to control costs?
• What does the forward guidance suggest? A company losing money now might be expected to reach profitability within one or two years. That is very different from one with no clear path to profit at all.
• What does the backlog or order book look like? For businesses that sign long contracts (construction, defence, cloud computing), the backlog can be more informative than the current quarter’s revenue.
• Are the losses getting smaller or larger as revenue grows? If the loss as a percentage of revenue is shrinking, the business is likely on a path towards profitability. If the opposite is true, the model deserves more scrutiny.

You can explore how to apply these questions to your virtual portfolio in the challenge. The most important skill in investing is not knowing which companies will win — it is asking the right questions about the ones you hold.

FAQ

Why do investors buy shares in companies that are losing money?

Because share prices reflect expectations about the future, not just today’s results. If a company is growing revenue very quickly and has large, long-term contracts already signed, investors may be willing to accept current losses in exchange for the potential profits years from now. The key question is whether the losses are the temporary cost of rapid growth, or evidence of a business model that will never generate sustainable profit.

What is a revenue backlog?

A backlog is the total value of contracts a company has already signed but not yet delivered on. It is a measure of future revenue that is already committed. CoreWeave’s backlog of approximately $104 billion as of June 2026 meant customers had already agreed to pay that much for services yet to be provided — giving investors confidence that the current spending was matched by future income.

What is capital expenditure and why does it cause losses?

Capital expenditure (capex) is money spent building or buying long-term assets — in CoreWeave’s case, server farms full of expensive GPU chips. This spending shows up immediately as a cost, but the revenue from it arrives gradually over years. During a fast build-out, spending can far exceed current income, producing losses even as revenue grows rapidly. The critical test is whether the assets being built will eventually generate returns that justify the cost.

What is the difference between revenue growth and profitability?

Revenue is the total income received from customers. Profit is what remains after all costs — staff, equipment, debt interest, overheads — have been paid. A company can grow revenue very quickly while still posting a loss if its costs grow even faster, especially when spending heavily to build infrastructure. Revenue growth tells you about customer demand. Profitability tells you about whether the business model works at scale.