CoreWeave's impressive backlog and partnerships face significant risks, including high capital expenditure, competition from hyperscalers, and securing power permits for expansion.
Risk: Securing power permits for expansion and converting backlog into revenue while managing high capital expenditure and funding costs.
Opportunity: Capturing a genuine moat by solving power logistics before hyperscalers do.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Key Points
- CoreWeave is expanding beyond cloud rental into inference, agentic AI, and enterprise services through its Weights & Biases integration and Physical AI Field Engineering program.
- Its customer base now spans AI labs, hyperscalers, and enterprises like Meta, Caterpillar, and Bentley Systems, backed by a $104 billion revenue backlog.
- CoreWeave is rapidly scaling power capacity, …
Read more
Key Points
- CoreWeave is expanding beyond cloud rental into inference, agentic AI, and enterprise services through its Weights & Biases integration and Physical AI Field Engineering program.
- Its customer base now spans AI labs, hyperscalers, and enterprises like Meta, Caterpillar, and Bentley Systems, backed by a $104 billion revenue backlog.
- CoreWeave is rapidly scaling power capacity, now near 4.2 gigawatts contracted, with Nvidia's backing supporting a build-out of more than 5 gigawatts by 2030.
- 10 stocks we like better than CoreWeave ›
The last few years have seen a flurry of flashy IPOs, but I'm more interested in the companies building the infrastructure behind the boom. CoreWeave (NASDAQ: CRWV) -- which went public in 2025 -- is one of them, and I think its AI cloud business could make it an outperformer by 2028. To see why, it helps to start with what the company actually does behind the scenes.
Missed AI’s "Act 1"? Act 2 Could Be 15x Bigger. Most investors think they missed the AI boat because they didn't buy Nvidia in 2005. But according to our analysts, we’re only at the end of "Act 1"—the R&D phase. "Act 2" is the global rollout. Continue »
CoreWeave is becoming more than just a cloud provider
The way to think about CoreWeave is to look at what happens behind the scenes when a company wants to train or run a massive model. It needs enormous amounts of compute power, networking, and storage. And with demand for cloud compute outpacing supply, companies find it increasingly challenging to secure it.
CoreWeave packages all those resources in a single AI platform. That matters because building AI infrastructure from scratch is expensive and time-consuming, and most companies lack the expertise required. So, instead of waiting years and spending heavily up front to build their own capacity, companies can simply rent access to infrastructure from neoclouds such as CoreWeave.
But the company has been expanding its AI platform well beyond that. In fact, its recent launches cover inference, agentic workloads, and AI development. It's also expanding its software capabilities through Weights & Biases, which it acquired in 2025. This means the company is trying to expand across AI workflows rather than just renting out compute.
CoreWeave's launch of Physical AI Field Engineering is another good example. The company is now adding engineers to customer teams to help them translate proprietary data into AI systems, especially in industries such as automotive, aerospace, and manufacturing.
With its recent launches, CoreWeave is making it harder for potential customers to ignore its AI platform, which is already resonating with an expanding roster of clients.
The customer list keeps growing across major industries
Meta Platforms recently expanded its relationship with CoreWeave through a multiyear AI infrastructure deal. Anthropic also signed an agreement for compute to support Claude's development. Wall Street market maker Jane Street announced a multibillion-dollar AI cloud agreement, while companies like Caterpillar, Bentley Systems, and Grammarly have also become customers.
This broad mix of clients shows that CoreWeave is serving more customers spanning AI labs, trading firms, and even hyperscalers with demanding compute needs. That diversification is valuable because it reduces the company's dependence on any single revenue source or industry.
Strong demand is already showing up in the company's backlog. CoreWeave reported a revenue backlog of $104 billion as of the end of the second quarter. That doesn't include the $25 billion in additional net customer commitments it added early in the third quarter.
Granted, it won't be able to turn that backlog into revenue instantly. CoreWeave still has to build the capacity to deliver on those contracts. Still, it does provide strong visibility into its future revenue growth.
CoreWeave is racing to expand capacity to meet demand
The other side of the equation is capacity. CoreWeave expanded active power to roughly 1.5 gigawatts, and it had 3.7 gigawatts of total contracted power as of the end of the second quarter. From there, contracted power climbed further to 4.2 gigawatts by August, which will make it easier for CoreWeave to fulfill its contracts and meet customer demand.
Not only that, the company announced that it deployed a multi-rack Nvidia Vera Rubin NVL72 cluster, consolidating hundreds of GPUs into a single system optimized for intensive AI workloads.
Speaking of Nvidia, the chipmaker invested $2 billion in CoreWeave earlier in 2026, while the two companies expanded their relationship around deploying more AI factories. CoreWeave said the partnership is meant to accelerate the build-out of more than 5 gigawatts of AI infrastructure by 2030.
Why CoreWeave looks positioned to outperform by 2028
The question isn't whether AI demand will keep growing, but which companies will capture the spending from that growth. CoreWeave stands out for providing the infrastructure that companies need to train and run their models.
Heading into 2028, the setup is already strong. CoreWeave is growing its customer base, adding capacity, and pushing into inference and enterprise AI. So, while investors chase flashier new IPOs, 2025 IPO CoreWeave is better positioned to stand out through 2028.
Should you buy stock in CoreWeave right now?
Before you buy stock in CoreWeave, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and CoreWeave wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $384,839! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,385,657!
Now, it’s worth noting Stock Advisor’s total average return is 936% — a market-crushing outperformance compared to 213% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of September 24, 2026. *
Rick Orford has positions in Meta Platforms. The Motley Fool has positions in and recommends Caterpillar, Meta Platforms, and Nvidia. The Motley Fool recommends Bentley Systems. The Motley Fool has a disclosure policy.
AI Talk Show
Four leading AI models discuss this article
Opening Takes
“The bull case for CoreWeave rests on rapid conversion of a $104B backlog into sustained revenue and healthy margins, which may not materialize given the company’s capital-intensive model and potential demand volatility.”
CoreWeave is pitched as a behind-the-scenes AI infrastructure winner, with a growing backlog and aggressive capacity builds backed by Nvidia. The thesis pins growth on AI demand, software moat via Weights & Biases, and services like Physical AI Field Engineering. That sounds compelling, but the picture is incomplete: a $104B backlog is not revenue; monetization depends on capex-heavy build-out, utilization, and long-cycle deployments. The moat may erode if hyperscalers and incumbents scale their own infra or if demand moderates. Valuation risk remains high for a company that only recently IPO’d; execution risk to reach 5GW by 2030 and maintain margins is material.
The strongest counter is that backlog ≠ revenue and near-term margins; massive capex and working-capital cycles could compress returns if AI demand slows, GPUs become cheaper or more commoditized, or customers delay deployments.
“CoreWeave's long-term viability hinges on its ability to maintain premium margins during the transition from training-heavy workloads to more competitive, lower-margin inference services.”
CoreWeave’s $104 billion backlog is staggering, but investors must distinguish between 'contracted capacity' and 'profitable utilization.' While the pivot to inference and agentic AI through the Weights & Biases acquisition suggests a move up the value chain, CoreWeave remains a capital-intensive utility play tethered to Nvidia’s supply chain. The real risk isn't demand—it's execution and margin compression. If hyperscalers like AWS or Azure successfully internalize their custom silicon stacks, CoreWeave’s neocloud model faces a brutal race to the bottom on pricing. I expect volatility as the market realizes that building 5 gigawatts of infrastructure is a liquidity-draining endeavor that requires constant, dilutive financing.
The immense backlog and deep integration with Nvidia’s latest hardware provide a structural moat that protects CoreWeave from generic cloud price wars, effectively making them the 'Intel Inside' of the AI infrastructure layer.
“CoreWeave has demand visibility but faces a structural margin compression risk if hyperscalers shift from renting to owning infrastructure, which the article never addresses.”
CoreWeave's $104B backlog and 4.2GW contracted capacity look impressive on paper, but the article conflates *committed capacity* with *revenue certainty*. The real test is conversion velocity and gross margins. A $104B backlog over 5+ years at typical cloud margins (20-30%) is materially different from the growth trajectory implied here. Nvidia's $2B investment is bullish for validation but also creates vendor lock-in risk. The Weights & Biases acquisition adds software stickiness, but integration execution risk is underplayed. By 2028, the question isn't demand—it's whether CoreWeave captures margin expansion or races to the bottom competing with hyperscalers' own infrastructure.
Meta, Microsoft, and Google are all building captive AI infrastructure at scale; CoreWeave's customer diversification masks the fact that its largest revenue driver (hyperscalers) has every incentive to internalize these workloads over time, collapsing the addressable market.
“CoreWeave's backlog and capacity targets are unlikely to translate into sustained outperformance by 2028 due to hyperscaler competition and buildout risks.”
The article pitches CoreWeave (CRWV) as an AI infrastructure winner with a $104B backlog, 4.2GW contracted power, and Nvidia support for 5GW+ by 2030, plus expansion into inference and enterprise via Weights & Biases. Yet it underplays conversion risk: turning contracted power and backlog into actual revenue requires flawless execution on data-center builds amid power constraints and rising interest costs. Hyperscalers (AMZN, MSFT, GOOG) are simultaneously scaling their own AI clouds, creating direct competition that could compress margins. The 2025 IPO timing also leaves valuation opaque relative to growth.
Even with execution hurdles, CRWV's specialized GPU clusters and early enterprise wins could still capture outsized share if Nvidia supply remains tight and hyperscalers prioritize internal workloads.
The Debate
Responding to Gemini
“Backlog alone isn’t a moat; ongoing capex and financing costs could erode returns, making margin upside uncertain even if demand holds.”
I’d push back on treating backlog as a moat. The real hidden variable is cash burn and capital structure: 4.2GW today, 5GW by 2030, implies ongoing heavy capex and working capital. If funding costs stay high, IRR on new capacity may stay well below cloud margins, pressuring returns even with software moat via Weights & Biases. Hyperscalers slow or Nvidia pricing/supply loosens could widen the funding risk, not just demand risk.
Responding to Claude
“CoreWeave's primary risk is power infrastructure deployment speed rather than hyperscaler competition.”
Claude, your focus on hyperscalers internalizing workloads misses the immediate bottleneck: power and cooling. CoreWeave isn't just selling compute; they are selling 'time-to-market' for companies that cannot wait for Azure or AWS to clear their internal backlogs. The real risk isn't hyperscaler competition, but the grid. If CoreWeave fails to secure power permits for that 5GW target, their backlog becomes a liability, not an asset. They are essentially a power-arbitrage play disguised as software.
Responding to Gemini
“Power availability, not GPU supply or hyperscaler competition, is the binding constraint on CoreWeave's 5GW thesis—but the article never specifies whether backlog includes secured power or just compute orders.”
Gemini's power-grid constraint is the hardest variable to model and the one most likely to break the bull case. But it cuts both ways: if CoreWeave solves power logistics before hyperscalers do, they've captured a genuine moat, not just a temporary arbitrage. The real question is whether their backlog reflects *signed power agreements* or just compute commitments. The article doesn't clarify this distinction—that's a massive disclosure gap.
Responding to Claude
“Power permit uncertainty directly heightens funding risks for CoreWeave's capacity buildout.”
Claude flags the power agreement disclosure gap, which amplifies ChatGPT's capital structure risks. Delayed permits would stall the 5GW expansion, sustaining elevated capex burn without revenue conversion and pressuring IRRs below the 20-30% cloud margins referenced earlier. This linkage makes Nvidia's validation less protective than assumed if grid access slips.
Panel Verdict
NEUTRAL No ConsensusCoreWeave's impressive backlog and partnerships face significant risks, including high capital expenditure, competition from hyperscalers, and securing power permits for expansion.
Capturing a genuine moat by solving power logistics before hyperscalers do.
Securing power permits for expansion and converting backlog into revenue while managing high capital expenditure and funding costs.
Related Signals
This is not financial advice. Always do your own research.