After CoreWeave’s (CRWV) stock price was cut in half, it still rose 24%: Can the AI compute cloud giant challenge AWS, Azure, and Google Cloud?

On July 13, 2026 (Beijing time), CoreWeave (CRWV) closed at $88.88 on the previous trading day, down 0.91% from the day before. After-hours trading inched down slightly to $88.81. This AI infrastructure company, centered on GPU cloud computing as its core business, has seen its stock price skyrocket since its 2025 listing—from the offering price to a peak of $187—before pulling back to its current range. Although the share price is about half of its peak, it is still up 24.12% year-to-date, and its market capitalization remains around $48.49 billion.

Divergent valuation views in the capital markets reflect investors’ deeper questioning of the AI cloud computing business model: does the explosive growth in AI model training demand provide enough support for the long-term expansion of GPU cloud services? Can a “new cloud” focused on AI compute power like CoreWeave carve out a breakthrough in the ecosystem moat built by AWS, Azure, and Google Cloud? And when will the erosion of profitability caused by sustained, high-intensity capital expenditures finally turn?

The answers to these questions concern not only the valuation of CoreWeave, but also the underlying judgment driving the entire logic of AI infrastructure investment.

A structural shift in AI compute demand: from “buying GPUs” to “renting cloud”

The exponential rise in the cost of training large models is fundamentally changing how enterprises obtain compute power. According to SemiAnalysis data, the one-year GPU rental contract price for NVIDIA H100 has risen from a low of $1.70 per GPU per hour in October 2025 to $2.35 in March 2026, an increase of nearly 40%. The research also shows that half of the surveyed GPU suppliers said H-series chips have no inventory on hand, and that all of the new-generation Blackwell series production capacity expected to come online before August to September 2026 has already been reserved.

The continued upward movement in GPU rental prices essentially reflects a market mapping of supply-demand imbalance. On one hand, AI labs, hyperscalers, and enterprises continue to drive rapid growth in demand for compute power. On the other hand, GPU chip supply is constrained by production ramp-up cycles, making it difficult in the short term to match the pace of demand growth. Against this backdrop, the trend of enterprises accelerating their shift from “buying GPUs” to “renting AI compute power” is gathering momentum—buying means high upfront capital expenditures, long delivery cycles, and depreciation risks caused by chip iteration; renting, by comparison, offers greater financial flexibility and more efficient resource allocation.

According to data from market research organizations, the GPU as a Service (GPU as a Service) market is estimated to be about $7.36 billion in 2026, up from $5.7 billion in 2025—an increase of 29%. It is expected to grow to $26.43 billion by 2031, representing a compound annual growth rate of 29.12%. Another firm, Mordor Intelligence, is more optimistic, forecasting that the GPU cloud market will grow from $7.73 billion in 2025 to $15.62 billion in 2026, reaching $37.69 billion by 2031.

Regardless of which set of data is used, the conclusion is consistent: the AI compute leasing market is in the early stage of high-speed expansion. CoreWeave is one of the most closely watched independent players in this segment.

CoreWeave’s business model: a closed loop of GPU procurement, data center construction, and compute leasing

CoreWeave’s business chain can be broken down into four steps: GPU procurement → data center construction → AI compute leasing → model companies and enterprise customers. The model may look simple, but each step forms competitive barriers and sources of risk.

On the GPU procurement side, CoreWeave has established deep cooperation with NVIDIA, allowing it to secure priority supply rights and large-scale procurement discounts. This is a clear advantage over smaller and mid-sized GPU cloud service providers—in today’s environment of GPU shortages, supply-chain certainty itself is a core competitive strength. On the data center construction side, the company continues to expand its infrastructure footprint; based on publicly available information, it operates about 40 AI data centers.

On the compute leasing side, CoreWeave provides a dedicated cloud service for AI workloads (model training and inference), not general-purpose cloud computing. Its customer base includes some of the world’s largest AI model makers, such as OpenAI, Anthropic, Meta, Google, and Microsoft, as well as AI application platforms like Perplexity AI and Cursor, along with enterprise customers such as Siemens and Salesforce.

The company’s first-quarter 2026 earnings report reveals the stage-by-stage outcomes of this business model. The company achieved revenue of $2.08 billion, up 112% year over year, exceeding analysts’ expectations of $1.97 billion. More importantly, as of March 31, 2026, CoreWeave’s revenue backlog reached $99.4 billion. This figure indicates that the company has already locked in a high degree of contracted revenue for the coming years—management said on the earnings call, “All full-year 2026 capacity has been sold out.”

However, behind the high growth is equally high loss. In the first quarter of 2026, the company’s adjusted net loss widened to $5.89 billion. Ongoing infrastructure deployment investment is squeezing profit margins, which is also where the market is most concerned about its profitability.

CoreWeave’s differentiated competition versus AWS, Azure, and Google Cloud

To directly compare CoreWeave with AWS, Azure, and other traditional cloud giants, the first step is to clarify a core difference: they have different positioning.

AWS, Azure, and Google Cloud are comprehensive cloud computing platforms that provide end-to-end solutions ranging from compute, storage, and databases to AI services. According to Synergy Research Group data, in the global cloud infrastructure services market in the first quarter of 2026, AWS held about 28% of the market and ranked first, Microsoft Azure accounted for 21%, and Google Cloud accounted for 14%. Together, these three companies occupy more than 60% of market share, with an enormous ecosystem scale and a very large customer base.

CoreWeave’s positioning is more focused—AI-dedicated cloud infrastructure. Its advantage lies in end-to-end optimization for AI workloads: from the interconnect architecture of GPU clusters, to the IO performance of storage systems, to scheduling software and model deployment toolchains—everything is designed around the needs of large-model training and inference. This “AI-only cloud” strategy enables it, for certain AI workloads, to deliver better cost-effectiveness and superior latency performance than general-purpose clouds.

But focusing also means concentrated risk exposure. CoreWeave’s customers are highly concentrated in the AI sector. If AI model training demand slows cyclically, or if enterprises shift from renting to building their own compute, the company’s revenue growth will face direct headwinds. By contrast, AWS and Azure cover a broader range of scenarios—from traditional enterprise IT to cutting-edge AI—making them more resilient across cycles.

Another difference worth noting is the pricing model. CoreWeave’s compute rental prices are dynamically adjusted according to market supply and demand, while traditional cloud providers’ AI compute pricing is more often bundled as part of their integrated cloud services. Strategically, they tend to emphasize ecosystem lock-in rather than maximizing short-term profit. This means that in a price war, CoreWeave may face greater pressure on profitability.

The three core questions investors care about most

Can AI capital expenditures be converted into sustainable revenue?

This is the most critical variable in CoreWeave’s valuation logic. With Q1 2026 revenue of $2.08 billion and a $99.4 billion revenue backlog, these figures prove the reality of demand. But the issue is that continuously expanding data center infrastructure requires a steady stream of capital investment, and depreciation and operating costs will erode profits over the long term. A profitability inflection point will only occur when revenue growth stays consistently higher than the growth rate of capital expenditures, and the utilization rate of existing infrastructure remains high. As of now, this inflection point has not arrived— the company is still in the “trading losses for growth” stage.

Will the GPU cloud market exist long-term?

This involves a more fundamental judgment: is AI compute leasing a structural demand, or a temporary phenomenon caused by specific supply-demand mismatches? If GPU supply increases significantly in the coming years, or if improvements in large-model training efficiency reduce the demand for compute, the growth slope of the GPU cloud market may slow. But based on current trends, explosive growth in AI inference demand is becoming a new growth driver for compute leasing—management explicitly positioned inference as “a way to monetize AI” during the Q1 call. Unlike the one-time characteristic of training demand, inference demand is sustained and benefits from scale effects, which could provide longer-term support for the GPU cloud market.

Does CoreWeave have network effects?

Network effects are the deepest moat for cloud service providers. AWS’s ecosystem advantage lies not only in the scale of its infrastructure, but also in the developer tools, third-party services, and enterprise data accumulated on top of it. CoreWeave currently does not have an ecosystem network at a comparable scale—its competitiveness comes more from optimizing the hardware supply chain and dedicated infrastructure. However, it is worth noting that as more AI companies deploy their workloads on the CoreWeave platform, the toolchains and optimization practices around the platform are also accumulating. If this accumulation can form a positive feedback loop, CoreWeave could establish its own ecosystem moat within the vertical AI space.

Conclusion

At its core, CoreWeave’s story is a microcosm of the underlying logic of AI infrastructure investment. Amid the backdrop of explosive growth in AI compute demand, CoreWeave—focused on GPU cloud computing—leveraged its deep partnership with NVIDIA, its large-scale GPU procurement capability, and its rapidly expanding data center network to achieve 112% year-over-year revenue growth and a $99.4 billion revenue backlog in Q1 2026.

But the other side of high growth is persistent losses and heavy capital expenditures. The market’s disagreement over its valuation is essentially a disagreement over whether the “trading losses for growth” model is sustainable. In the cloud computing market dominated by AWS, Azure, and Google Cloud, whether CoreWeave can carve out its own territory with a differentiated positioning as an AI-dedicated cloud depends on three factors: the long-term growth rate of AI compute leasing demand, the pace of improvement in the company’s capital expenditure efficiency, and whether it can build a truly defensible ecosystem moat in the vertical AI space.

For investors, CoreWeave represents a form of “higher-purity exposure” to the AI infrastructure space. Unlike investing in traditional cloud giants, it does not need to take on risks from diverse businesses such as enterprise IT and consumer internet—but it therefore gives up the safety margin provided by diversification. In the AI infrastructure segment, where fundamentals are highly certain but the path is full of variables, whether CoreWeave becomes the frontrunner or an overexpansion risk-taker will be answered by time.

FAQ

Q1: What are CoreWeave’s main sources of revenue?

CoreWeave’s revenue mainly comes from AI compute leasing services, providing GPU cloud computing resources for AI companies and enterprises to handle model training and inference workloads. In the first quarter of 2026, the company’s revenue reached $20.8 billion, up 112% year over year. Its customers include major global AI model makers such as OpenAI, Anthropic, Meta, Google, and Microsoft.

Q2: What is the core difference between CoreWeave and AWS?

CoreWeave is a dedicated cloud platform focused on AI workloads, where the infrastructure—from GPU clusters to the software stack—is optimized for AI scenarios; AWS is a comprehensive cloud platform that provides full-stack services such as compute, storage, and databases. CoreWeave’s advantage lies in deep optimization and cost-effectiveness for AI scenarios, while AWS’s advantage lies in ecosystem scale and business diversification.

Q3: Is CoreWeave currently profitable?

Not yet. In the first quarter of 2026, although revenue reached $20.8 billion, its adjusted net loss widened to $5.89 billion. Continued data center expansion and high-intensity capital expenditures are the main reasons for the losses. The market widely watches when its profitability inflection point will arrive.

Q4: What does the “$99.4 billion revenue backlog” mean?

Revenue backlog (Revenue Backlog) refers to the total amount of future revenue the company has signed with customers, but has not yet recognized upon delivery and fulfillment. As of March 31, 2026, CoreWeave’s revenue backlog reached $99.4 billion. This figure means the company has a high degree of contractual revenue locked in for the coming years, but it is also important to note that the actual recognized revenue depends on delivery progress and whether customer needs are met.

Q5: What are the main risks of investing in CoreWeave?

Key risks include: the customer base is highly concentrated in the AI sector, making demand fluctuation risk higher; sustained high capital expenditures create long-term profit pressure; competitive pressure from traditional cloud giants such as AWS and Azure; and the risk that rental prices may fall after improvements in GPU supply. The company’s current trailing price-to-earnings ratio is negative, and its price-to-book ratio is about 13.39x.