Everything About AI
NextFin News - CoreWeave shares rallied into the Tuesday close as investors weighed a raised revenue forecast and a $104 billion order backlog against a growing debate over whether the artificial-intelligence infrastructure boom is a structural shift or a credit-fueled cycle. The stock touched an intraday high of $51.02 before settling at $48.73, down 0.51% on the session, and extended gains of more than 2% in after-hours trading to $50.04.
The move captures the central tension now dividing the market. CoreWeave's revenue machine is firing on every cylinder: second-quarter sales more than doubled, guidance moved higher, and the backlog nearly doubled in a year. Yet the company is funding that growth with tens of billions of dollars of capital spending and a balance sheet that has drawn fresh scrutiny from analysts who see financing costs, not demand, as the constraint that will decide the winners in AI infrastructure. The rally is a bet that demand wins.
The Numbers Behind the Rally
CoreWeave reported second-quarter revenue of $2.58 billion, up 112% from a year earlier and ahead of the roughly $2.54 billion to $2.55 billion analysts had expected, marking a fifth consecutive quarter of record revenue. The company posted an adjusted loss per share of about $1.03, narrower than the prior-year loss, while adjusted EBITDA doubled to $1.5 billion at a 59% margin and adjusted operating income reached $128 million, a 5% margin.
The guidance is what moved the stock. Management raised its full-year 2026 revenue forecast to between $12.4 billion and $13.2 billion, up from a previous range of $12 billion to $13 billion, and lifted adjusted operating income guidance to between $960 million and $1.15 billion. For the third quarter alone, the company expects revenue of $3.45 billion to $3.6 billion, a 158% year-over-year surge at the midpoint, with adjusted operating income of $200 million to $260 million. To fund that growth, CoreWeave raised its 2026 capital expenditure target to between $35 billion and $39 billion, up from $31 billion to $35 billion, and guided third-quarter capex to between $11.5 billion and $13.5 billion.
The backlog behind the forecast carries meaningful detail. Revenue backlog ended the quarter at $104 billion, excluding more than $25 billion of net new customer commitments added early in the third quarter. No single customer accounts for more than 35% of the backlog, a decrease from earlier in the year, and more than 60% of the backlog is tied to investment-grade customers. Major commitments include an additional $21 billion from Meta Platforms, a multi-year agreement with Anthropic, and a $6 billion commitment from quantitative trading firm Jane Street.
The company is also scaling the physical inputs that make the revenue possible. CoreWeave reached 1.5 gigawatts of active power by the end of the second quarter, adding nearly 500 megawatts in the period, and management lifted its target for annualized run-rate revenue at the end of 2026 to between $18.5 billion and $19.5 billion. The year-end active power target now stands above 1.85 gigawatts.
On the call, Chief Executive Mike Intrator framed the quarter as a turning point in the company's scale.
"Demand continues to exceed supply across sectors, geographies, and generations of infrastructure," Intrator said. "The opportunity ahead is generational. CoreWeave is the essential cloud for AI. Our conviction in our strategy has never been stronger, and our execution continues to reinforce it."
The stock's intraday range tells the story of the two-sided trade. Shares opened at $48.10, climbed to $51.02, dipped to a low of $47.06, and closed at $48.73 on volume of 71.5 million shares, above the 10-day average volume of roughly 55.6 million. The shares remain well below their 52-week high of $138.25 and sit just above their 52-week low of $40.58, a range that reflects how completely the AI trade has been repriced this year.
Why the Market Is Still Buying the AI Infrastructure Story
The rally is not just a reaction to a single quarter. It is a bet that the shift in computing demand has further to run, and that CoreWeave's backlog is a claim on future cash flow rather than a monument to overbuilding. The mechanism runs through the hyperscalers and large technology companies that cannot build AI data centers fast enough on their own. CoreWeave buys advanced graphics processing units from Nvidia, installs them in dedicated data centers, and leases the capacity back to customers including Microsoft, OpenAI, Meta Platforms, and Alphabet under long-term contracts. Those contracts become the backlog, and the backlog gives the company the visibility it needs to raise debt and equity for more buildout.
This is where the structural argument gains its force. Unlike a cyclical commodity boom driven by a temporary supply squeeze, the AI infrastructure buildout is anchored in a multi-year transition. Enterprises and cloud providers are rewiring their computing stack around generative AI, and that rewiring cannot be undone without abandoning the applications being built on top of it. Analysts at Jefferies made the point directly in a research note: "AI compute represents a new era of computing which ushers in vast complexities in scaling and reliably hosting this compute, something CoreWeave has been able to do, while incumbents have struggled, and we believe potentially underestimating the long-term opportunity."
There is also a pricing argument embedded in the guidance. Bank of America, which raised its revenue growth forecasts for CoreWeave, expects revenue per megawatt to rise from about $10 currently to $14 by the end of 2028, supported by deployments of Nvidia's next-generation computing systems. If that plays out, the same physical capacity generates more revenue over time, and the backlog converts at higher margins than the market assumes.
The customer mix is improving in a way that supports the structural read. With no single customer above 35% of the backlog and more than 60% tied to investment-grade counterparties, CoreWeave is less exposed to a single hyperscaler pulling back than it was earlier in the year. That diversification matters because it means the demand base is broadening beyond a handful of the largest technology companies.
The Counter-Thesis: Financing Costs and the Sentiment Problem
The strongest bear case does not dispute the revenue growth. It attacks the economics of financing it. CoreWeave's capital expenditure guidance of $35 billion to $39 billion for 2026 implies a cash burn that must be funded through debt and equity markets. Free cash flow was negative $5.7 billion in the most recent quarter, and total liabilities have swelled to roughly $72 billion. The company has missed earnings estimates in five of the past six quarters.
Rosenblatt Securities cut its price target on the stock by 36%, from $250 to $160, while maintaining a Buy rating, citing rising financing costs. The analyst raised assumptions for borrowing costs on CoreWeave's delayed-draw term loan and increased the discount rate used to value future cash flows. The concern is not abstract: third-quarter interest expense is guided at between $860 million and $940 million, a figure that sits alongside adjusted operating income guidance of just $200 million to $260 million for the same period.
The sentiment problem runs deeper than one firm's target cut. After a bruising stretch for artificial-intelligence stocks, CoreWeave faces a high bar to convince Wall Street that the AI trade is alive and well. Investors are focused less on the quarterly results themselves and more on what management says comes next, particularly in light of political pressure over data-center construction, with many voters opposing new facilities near their communities. That constraint is real: a company whose growth depends on adding gigawatts of power capacity needs permits, grid connections, and local approval, any of which can slow the buildout.
There is also a voice from inside the AI ecosystem adding to the skepticism. OpenAI Chief Executive Sam Altman recently said he was seeing "the first signs of what feels to me like unsustainable silliness of random new neoclouds popping up," warning that some players claim they will build gigantic amounts of compute without the revenue to support it or a buyer for it. Altman did not name CoreWeave, but the comment has become a reference point for traders reassessing the publicly traded AI compute buildout.
The bear argument, stated at its strongest, is this: CoreWeave's revenue growth is real, but it is being purchased with leverage at a time when borrowing costs are rising. If demand for AI compute proves cyclical rather than structural, the company is left with stranded capacity, heavy debt service, and a valuation that assumes perpetual hypergrowth. The stock's decline from its 52-week high of $138.25 to around $48.73 reflects that repricing.
Is AI Demand Cyclical or Structural? The Call That Matters
This is the judgment that decides the stock, and it requires separating two forces that are operating at the same time. The short-term leg is cyclical: sentiment toward AI infrastructure has swung hard, financing costs have risen, and the stock has been repriced on fears that the buildout has run ahead of demand. That leg can and probably will mean-revert as each quarter's revenue print either confirms or disappoints.
The long-term leg is structural. Three pieces of evidence point in that direction. First, the customer concentration has improved, with no single customer exceeding 35% of the backlog. Second, more than 60% of the backlog is tied to investment-grade customers, which reduces counterparty risk in a downturn. Third, the backlog-to-market-cap ratio remains elevated: a $104 billion backlog against a market capitalization of roughly $26.9 billion implies that the market is pricing in a meaningful probability that a substantial portion of that backlog never converts to revenue.
But the structural call has conditions attached. It holds only if CoreWeave can finance the buildout without diluting shareholders to the point of destroying equity value, and only if revenue per megawatt rises as next-generation systems are deployed. Chief Financial Officer Nitin Agrawal pointed to one early signal that the company is moving up the value chain: booked annual recurring revenue for managed inference rose from $1 million to more than $100 million within a quarter, and management expects about $250 million of ARR from the business by year-end. That matters because inference workloads diversify revenue beyond the large training contracts that dominate the backlog.
If borrowing costs continue to climb and capacity additions slow, the structural thesis weakens into a cyclical one: a boom that built too much, too fast, funded with debt that must be refinanced at higher rates. The company's own guidance contains the tension: revenue of $12.4 billion to $13.2 billion looks spectacular, but it comes with $35 billion to $39 billion of capital spending and third-quarter interest expense guided at $860 million to $940 million. Growth is not the question. The quality of that growth, measured in free cash flow per dollar of revenue, is.
The Second-Order Trade: Who Wins If CoreWeave Is Right
If the structural thesis holds, the beneficiaries extend well beyond CoreWeave. Nvidia remains the primary supplier of the advanced chips that make the capacity valuable, and every gigawatt of new AI data-center capacity implies continued demand for its graphics processing units. Power providers and utilities that can deliver the electricity these facilities require become the bottleneck asset class; a company that can add 500 megawatts in a single quarter is competing for a scarce input that no amount of capital can instantly create.
The hyperscalers that lock in capacity today also secure a scarce input for the AI applications that will drive their own revenue for the rest of the decade. That is the second-order logic behind the backlog: it is not just CoreWeave's revenue visibility, it is a claim by the largest technology companies on the compute they believe they will need to remain competitive.
The second-order risk runs the other way. If CoreWeave's financing costs rise faster than revenue per megawatt, the company faces a margin squeeze even as top-line growth accelerates. That is the trap embedded in the guidance. Interest expense guided at $860 million to $940 million for the third quarter against adjusted operating income of $200 million to $260 million means the company is, for now, paying more to finance its buildout than it earns from operating it. That gap must close, and it must close through revenue per megawatt rising and borrowing costs falling, not through growth alone.
What To Watch Next
Three signals will determine whether this rally is the start of a recovery or a bear-market bounce. First, watch the pace of net new commitments: the market needs to see the more-than-$25 billion quarterly pace sustained, not a slowdown. Second, watch borrowing costs on CoreWeave's delayed-draw term loan; a further rise would validate the Rosenblatt concern and pressure the valuation. Third, watch revenue per megawatt as next-generation systems come online; if it fails to rise toward the $14 per megawatt that some analysts expect by 2028, the margin story breaks.
The falsifying signal is specific and quantifiable: if CoreWeave reports a quarter in which revenue backlog growth stalls below 20% year over year while interest expense rises above 40% of adjusted operating income, the structural-demand thesis is wrong and the stock is pricing a boom that has already peaked. A backlog growing below 20% would signal that the hyperscalers are slowing their commitments; interest expense above 40% of adjusted operating income would signal that financing costs are consuming the operating leverage the company is counting on.
There are also external signals. If OpenAI's Altman is right that neocloud capacity is being built without buyers, utilization rates across the industry will fall before revenue does, and CoreWeave's pricing power will erode. If local opposition to data-center construction slows permitting, the company's ability to add gigawatts of power capacity will lag its capital spending plans, and the revenue per megawatt thesis will take longer to play out.
The Bottom Line Across Time Horizons
In the short term, the stock is a sentiment and liquidity trade. After a brutal repricing from $138.25 to the low-$40s, any positive earnings surprise or guidance raise can produce a sharp bounce, and Tuesday's move into the close and after-hours is consistent with that pattern. The risk is that the bounce runs into the same financing concerns that drove the selloff.
In the medium term, the fundamental question is whether revenue converts into free cash flow. The base case is that CoreWeave delivers on its $12.4 billion to $13.2 billion revenue target, that managed inference adds $250 million of ARR by year-end, and that the backlog continues to grow as hyperscalers lock in capacity. The upside case is that revenue per megawatt rises toward $14 by 2028 and financing costs stabilize, allowing operating leverage to close the gap between interest expense and operating income. The downside case is that borrowing costs keep rising, capacity additions slow, and the backlog growth rate decelerates below 20% year over year, turning the stock's backlog-to-market-cap ratio from an asset into a liability.
In the long term, the question is whether AI infrastructure is a regime shift or a cycle. The evidence — a diversified, investment-grade backlog, rising revenue per megawatt, and demand that management says continues to exceed supply across sectors and geographies — points toward a structural shift. But the financing structure built on top of that shift is fragile, and fragility has a way of turning structural stories into cyclical ones when credit conditions tighten.
The rally says the market is willing to pay for growth today and ask about free cash flow tomorrow. That trade works until the bill for the financing comes due.
Explore more exclusive insights at nextfin.ai.
