Companies that once focused on cryptocurrency mining are rapidly becoming suppliers of AI computing capacity, posting revenue growth that rivals the early expansion of major cloud platforms while taking on the heavy capital costs of data centers, chips, and power contracts.
CoreWeave illustrates both sides of that transition. The company reached $2.6 billion in revenue in roughly 25 quarters, compared with the 40 quarters Amazon Web Services took after launch to reach the same revenue level, based on company-reported results cited in the source material. Yet CoreWeave’s shares were down about 16% over the previous year, reflecting concern over the cost of building GPU infrastructure as well as the growth opportunity created by AI demand.
The emerging group, often called “neoclouds,” includes operators that sell specialized computing capacity to AI developers and enterprises. Several began in or around the crypto-mining sector, where access to power infrastructure, cooling systems, data-center sites, and large-scale hardware operations offered a foundation for a pivot into AI workloads.
Revenue growth is arriving with unusually high costs
CoreWeave, Nebius, and Applied Digital are growing from a far smaller base than the major cloud providers. Hyperscalers continue to generate quarterly cloud and infrastructure revenue measured in tens of billions of dollars, leaving neoclouds well below their scale and financial resilience.
That gap shapes how markets assess them. Smaller companies such as Nebius and Applied Digital have traded at higher price-to-sales multiples than CoreWeave, partly because their reported growth rates were cited at roughly 400% to 450%. CoreWeave’s revenue was still doubling, but its larger revenue base and more visible infrastructure costs have produced a lower valuation multiple.
Price-to-sales ratios provide only a partial view of these businesses. Unlike software companies, GPU-cloud operators must spend heavily before they can sell capacity. They buy or lease advanced chips, install networking equipment, secure power, build data-center space, and often borrow to finance the process. A fast-growing revenue line can therefore coexist with rising cash needs and weaker margins.
CoreWeave was cited as an example of that pressure: capital expenditures increased faster than revenue as the company expanded capacity. Depreciation on chips and related equipment accounted for more than half of revenue in the cited analysis, while interest expense rose alongside borrowing used to fund the buildout.
Those expenses make the business model sensitive to utilization. A GPU cluster can produce substantial revenue when customers keep it busy, but idle capacity, delayed construction, lower rental rates, or a faster-than-expected shift in hardware generations can quickly affect returns. The sector’s growth story is therefore closely tied to whether AI customers continue signing large, long-duration computing contracts.
Mining infrastructure gives operators a starting point, not a guarantee
The appeal of former mining operators lies in their physical footprint. Bitcoin mining companies already understand energy procurement, site construction, industrial cooling, hardware deployment, and operations in regions with comparatively low power costs. Those capabilities can shorten the path to operating AI data centers.
They do not eliminate the challenges of serving AI customers. Training and inference workloads require high-performance networking, reliable uptime, sophisticated cluster management, and access to the most sought-after GPUs. Cloud customers also expect service levels and software tools that go beyond the operational requirements of a mining site.
The shift places former miners in a more demanding market, where electricity remains central but is only one component of a broader infrastructure stack. A data-center operator needs power availability, but it also needs chip supply, financing, construction capacity, network connectivity, and customers willing to commit to compute contracts.
That has created a sharp divide between companies with active AI deployments and those whose transitions remain largely aspirational. Rapid revenue growth can validate the model, but high depreciation and debt costs show why revenue alone does not settle the question of long-term profitability.
Software spending is spreading beyond infrastructure
AI-related spending is also showing up in enterprise software. Atlassian reported cloud revenue growth of 31% year over year in the figures cited in the source material. Customers using its Rovo AI assistant were said to be increasing spending at nearly twice the rate of customers not using the tool.
That pattern offers a different form of AI monetization from GPU leasing. Infrastructure providers earn revenue by supplying computing power, while software companies aim to turn AI features into higher subscription spending, better retention, or broader product adoption. The two models are connected: rising enterprise use of AI software can increase demand for the computing capacity beneath it.
Horizontal software companies were among the stronger performers over the prior 30 trading days within the constituents of the IGV software ETF, according to the market-performance data cited in the material. Some of those gains later retreated. Expected price-to-sales multiples for many horizontal SaaS companies nevertheless remained below the levels implied by a typical growth-to-valuation trendline, except for several higher-priced names.
The contrast with neoclouds is clear. Software companies can potentially expand AI-driven revenue without buying warehouses full of hardware. Neocloud operators capture demand closer to the physical layer, but each additional unit of capacity carries substantial upfront cost.
Lower model costs may expand use rather than reduce spending
AI companies are also working to reduce the cost of serving model requests. Databricks has described its Smart Router system as a tool that assigns tasks to different AI models depending on their difficulty and cost. The company said the system can reduce average task costs by more than 30% while maintaining output quality, largely by directing simpler tasks to less expensive models.
Silicon Data’s Token Price Intensity Index similarly showed declining cost intensity for AI tokens, linked to a larger role for lower-cost open models. The index measures the cost per unit of model usage rather than total spending. As prices fall, businesses may run more queries, add AI features to more products, or serve more users, allowing total token consumption and total AI spending to keep growing.
Ramp’s spending data, which is skewed toward technology companies, showed a wide gap in AI adoption. The top 10% of companies by AI spending recorded per-employee AI outlays about 50 times larger than the median company in its dataset. Boston Consulting Group, analyzing 107 public companies, found that firms in the top two quintiles of token usage had faster revenue growth than the remainder of the group.
Those findings do not establish that higher token use causes faster growth. They do show that companies extracting more value from AI are often deploying it at a substantially larger scale, reinforcing demand for the computing infrastructure that neocloud operators are racing to build.
Explore how AI and blockchain converge to understand the tech shift behind neoclouds evolving from crypto mining to AI infrastructure.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
