Everything About AI
NextFin News - The AI trade is offering investors two true facts at once: CoreWeave has turned demand for computing into a $132.5 billion contracted backlog, while the same business remains dependent on enormous capital spending and external financing. Anthropic’s proposed public offering adds a second vote of confidence in AI demand, but it also raises the valuation bar for the entire sector. The immediate relief is therefore cyclical, not proof that the financing problem has disappeared.
The contradiction matters because the market is no longer asking only whether companies want more computing. It is asking whether the revenue created by that demand can outrun depreciation, interest expense and the cost of building power-hungry data centers. CoreWeave’s third-quarter figures supplied the bullish answer on growth. Its net loss and capital intensity supplied the bearish answer on cash economics. All figures in this article are anchored to information available through Nov. 11, 2026.
The Relief Rally Has a Narrower Foundation Than the Headline Suggests
CoreWeave’s third-quarter revenue reached $3.661 billion, up from $1.365 billion a year earlier, a 168% increase. Adjusted operating income was $379 million, compared with $217 million in the year-ago quarter, and the company said adjusted operating margins reached 10%. Management raised full-year 2026 revenue guidance to $13.0 billion-$13.3 billion and adjusted operating income guidance to $1.275 billion-$1.375 billion.
Those figures answer the first question confronting the market: is AI infrastructure demand still arriving in the income statement? For CoreWeave, the answer is yes. The company’s earnings presentation showed a $132.5 billion revenue backlog, up 138% year over year. It also said it added more than 900 megawatts of contracted power during the quarter and still expected more than 1.85 gigawatts of active power by year-end.
Michael Intrator, CoreWeave’s co-founder, chairman and chief executive officer, framed the result as an operating inflection.
“This was a breakout quarter for CoreWeave, with revenue increasing by more than $1 billion sequentially and adjusted operating margins reaching double digits a quarter ahead of plan,” Intrator said in the company’s earnings release.
The company’s release also reported a $709 million net loss for the quarter. The contrast is the story: operating momentum is improving, but equity value still depends on whether that momentum survives the financing burden.
Anthropic’s proposed listing adds a different kind of evidence. Its SEC registration statement confirms plans to list Class A common stock on Nasdaq under the symbol ANT. The offering would give public investors direct exposure to a frontier-model developer whose demand for computing is part of the demand chain supporting data-center operators, chip designers and power suppliers. A public listing would also create a daily market price for the economics of advanced AI rather than leaving those economics inside private funding rounds.
That distinction matters. A private funding mark can absorb uncertainty through strategic investors and negotiated terms. A listed company must disclose revenue growth, losses, capital commitments and risks that could impair future cash generation. Anthropic’s listing is thus both a demand signal and a valuation test, even before a final share count or price range is set.
The semiconductor proxy highlighted in the supplied market reference had already shown the market’s instability: a 12% drop was nearly erased by two days of gains before the index turned lower again. That sequence is not a clean trend. It is a market switching between two discount rates—one for the future cash flows of AI and one for the cost of financing the infrastructure needed to produce them.
Why the First-Order Bull Case Is Not Enough
The first-order case is simple: Anthropic needs models and CoreWeave supplies computing, so stronger Anthropic demand should support CoreWeave and other AI infrastructure stocks. The transmission mechanism is more complicated. Demand becomes revenue only when computing capacity is powered, installed, delivered and paid for; revenue becomes equity value only after depreciation, interest and dilution are absorbed.
CoreWeave’s backlog is large relative to current revenue, but backlog is not the same as near-term free cash flow. The company defines revenue backlog as remaining performance obligations plus estimated amounts expected to be recognized under committed contracts, subject to delivery and availability requirements. That qualification is material. The company must convert contracts into operating capacity, and it must finance that capacity before recognizing the full economic benefit.
The capital cycle is visible in the same quarter that produced the bullish revenue number. CoreWeave’s earnings presentation highlighted $14.5 billion of capital expenditures. It also pointed to approximately $9 billion of customer prepayments and deposits expected over the following three months. Those deposits can reduce external funding needs, but they do not eliminate the need to build and maintain the assets that support the contracts.
The second-order effect runs through credit markets. When lenders believe customer contracts are durable, they can finance servers and facilities against contracted revenue. When lenders question renewal, power availability or customer credit quality, the same infrastructure becomes a long-duration asset with uncertain residual value. Borrowing costs then rise precisely when a company needs more capital. The equity market may celebrate backlog growth while the credit market asks whether the backlog can be converted without another large financing round.
This is why the current relief should be classified as cyclical. The evidence is the sequence of the market’s own behavior: a 12% fall in the semiconductor proxy, a rapid partial recovery, and another reversal. The short-term driver is a change in expectations about demand and financing conditions, not a new rule that permanently changes industry economics. The mean-reverting element is the risk premium.
But the demand buildout itself is structural. AI models require more computation, and the infrastructure chain is expanding beyond chips into data centers, electricity, networking and specialized cloud services. CoreWeave’s addition of more than 900 megawatts of contracted power in one quarter is evidence of an industry reorganizing around scarce physical capacity. The structural fact does not guarantee that every provider earns attractive returns. It means the demand for capacity can persist even while the ownership of the profits changes.
The distinction is crucial. A cyclical rally can lift the group because investors fear less about near-term demand. A structural shift can create winners and losers because customers may eventually demand lower prices, more flexible contracts or direct control of power and facilities. The first supports sentiment. The second determines margins.
Anthropic’s Listing Raises the Sector’s Expectation Gap
Anthropic’s proposed IPO can validate the AI spending cycle, but it can also expose how much of the cycle is already embedded in valuations. The company’s official announcement said that the number of shares and price range had not yet been determined. A valuation discussed before final pricing should therefore not be treated as an established market value.
The public-market process changes the questions asked by investors. A private AI lab can emphasize capability, customer adoption and strategic partnerships. A listed company must disclose revenue concentration, losses, capital commitments and risks that could impair future cash generation. The SEC filing’s risk discussion makes the point in business terms: advanced models operate in sensitive contexts, and failures can create liability, reputational damage and customer-retention risks.
The expectation gap therefore has two sides. Bulls see Anthropic as evidence that enterprise demand is broad enough to support a new public-market category. Bears see an enormous capital requirement moving from private balance sheets into public portfolios. Both can be right in sequence. A successful offering may initially lower the perceived cost of capital for the AI ecosystem. If the stock later trades poorly because revenue growth does not translate into cash flow, it could raise the cost of capital for every infrastructure provider tied to the same spending chain.
The second-order market effect is a reallocation of bargaining power. If frontier labs can access public equity, they may fund more of their own infrastructure, negotiate better prices from cloud providers or secure long-term power directly. That would be positive for demand volume but potentially negative for the margins of intermediaries. Conversely, if public investors reward asset-light model companies more than data-center operators, capital may move toward software and away from the physical layer even as total AI spending rises.
That is the counter-intuitive possibility: a successful Anthropic IPO could validate AI demand while weakening the relative investment case for some suppliers. The sector does not move as one balance sheet. It is a chain in which each participant competes for the same dollar of future cash flow.
The strongest counter-thesis is that financing risk is overstated because customer prepayments, contracted revenue and rising utilization can create a self-funding flywheel. CoreWeave’s management explicitly cited multiple funding sources, including financing backed by customer contracts, operating cash flow and approximately $9 billion in expected customer prepayments and deposits. If that model scales, the company can grow without relying on continuously receptive capital markets.
That argument deserves more than a dismissal. A backlog up 138% year over year and revenue up 168% show that the contracts are not merely theoretical. Adjusted operating income rose to $379 million, and the company raised guidance. Those are operating facts, not promotional projections.
But the counter-thesis still leaves a measurable vulnerability: conversion and funding must improve together. The signal that would falsify the cyclical-relief judgment would be two consecutive quarters in which backlog grows while revenue conversion slows, adjusted operating margin falls below 10%, and customer prepayments fail to cover a rising share of capital expenditure. That combination would show that demand is not solving the capital structure; it is enlarging it.
What the Market Is Actually Pricing
Without a quantified, authoritative consensus range for the proposed Anthropic offering in the available primary materials, the market’s expectation cannot be reduced to a single defensible valuation number. The more reliable baseline is operational: CoreWeave’s own raised guidance of $13.0 billion-$13.3 billion in 2026 revenue and $1.275 billion-$1.375 billion of adjusted operating income. The equity market must decide how much of that guidance deserves a durable multiple after interest and capital spending.
That is a narrower claim than saying the market has priced a particular IPO valuation. It is also more useful. The key expectation is that contracted demand will convert into higher utilization and operating leverage before financing costs overwhelm the benefit. The share-price reaction becomes fragile if investors treat backlog as a substitute for cash flow.
There is a technical reason the tug-of-war can persist. AI equities are exposed to a common discount-rate factor, but their operating sensitivities differ. A chip designer may monetize demand through volume and gross margin. A neocloud operator must purchase or lease the physical capacity first. A model company must spend on training and inference while proving that customers will renew. Higher rates compress all three valuations, but they do not affect the same line item.
That creates cross-industry dispersion even when the headline AI narrative is bullish. Strong model demand can benefit chipmakers and cloud vendors while pressuring smaller infrastructure providers if customers use their bargaining power to force lower prices. Conversely, a supply shortage can protect infrastructure margins for a period, but only until new capacity arrives or customers build internally.
The market’s rapid reversals are therefore rational rather than inconsistent. Investors are not changing their view about whether AI exists. They are changing their view about the distribution of returns and the time required to earn them. A $132.5 billion backlog can coexist with a $709 million quarterly loss because the former is a contractual opportunity and the latter is a current accounting outcome. The valuation question sits between them.
The Time-Horizon Test
Over the short term, the signal is liquidity and positioning. A better-than-feared CoreWeave quarter can encourage investors who reduced exposure to rebuild positions, particularly after the semiconductor proxy’s 12% decline and partial recovery. Anthropic’s listing process can add a headline catalyst. This is the horizon on which sentiment can dominate fundamentals.
Over the medium term, conversion is decisive. Investors will need to see whether CoreWeave turns the $132.5 billion backlog into revenue without proportionally larger depreciation, interest expense and new equity issuance. The company’s raised guidance and 10% adjusted operating margin are constructive, but the quarterly net loss shows that operating improvement has not yet settled the cash-economics debate.
Over the long term, the structural question is control of scarce inputs. More than 900 megawatts of contracted power added in one quarter points to electricity and data-center capacity becoming strategic assets. Companies that control power, cooling, networking and reliable capacity may capture a larger share of AI economics. Companies that merely resell rented capacity may face margin pressure once supply expands.
The upside scenario is a self-funding infrastructure cycle: customer prepayments arrive as expected, utilization rises, margins remain at or above 10%, and Anthropic’s public offering broadens the equity capital available to the ecosystem. In that case, the current pullback would look like a cyclical reset inside a structural expansion.
The base case is more uneven. CoreWeave and a small group of infrastructure providers continue to grow rapidly, but investors separate backlog quality from backlog quantity. Model companies, cloud platforms, chip suppliers and power owners receive different multiples because their capital intensity and bargaining power differ. The AI trade remains a powerful theme but unstable as a basket.
The downside scenario is a financing-led break. If borrowing costs rise, customer deposits disappoint, or the market questions whether contracts can be delivered on schedule, the backlog becomes a reminder of future obligations rather than a cushion. Anthropic’s IPO could then transmit pressure in both directions: a weak offering would damage confidence, while a strong offering could leave public investors holding a valuation that later proves too demanding.
The next decisive evidence will not be another headline about AI demand. It will be the relationship between revenue, adjusted operating margin, capital expenditure and customer funding in the next two reporting cycles, alongside the final terms and trading behavior of Anthropic’s offering. If revenue growth stays above 100% year over year while margins hold at 10% or higher and prepayments fund a growing portion of investment, the bullish structural case strengthens. If backlog rises but conversion and margins weaken, the market’s relief will prove cyclical.
AI demand has survived the selloff. The financing model has not yet earned the same verdict.
Explore more exclusive insights at nextfin.ai.
