AI Demand Is Holding Up. AI Financing Is the Market’s New Stress Test

The current market tell is not that enterprise AI demand has disappeared. It is that investors are beginning to separate usage growth from the financing structure used to build the capacity behind it.

CoreWeave has become the clearest pressure point. Its shares were indicated at $52.91 in premarket trading at 08:27 ET on October 26, down 2.44% from the October 23 regular-session close.^[1] By contrast, Datadog and Snowflake were still above their October 23 closes in the latest available premarket snapshots: DDOG at $271.31, up 0.62%, and SNOW at $325.69, up 0.08%, as of their respective quoted times on October 26.^[1] That is not proof of a permanent winner-and-loser split, but it is a useful market signal: software monetization and infrastructure financing are being priced as different risks.

The fault line is credit, not simply AI enthusiasm

Recent reporting describes a sharp deterioration in the financing backdrop for highly leveraged AI infrastructure. CNBC reported that CoreWeave’s unsecured debt was yielding about 15%, while its five-year credit-default-swap spread stood at 11.45%; the company also had approximately $20.7 billion of debt due through 2030.^[2] The same report said data-center financing was facing additional uncertainty around permitting and political pushback ahead of the U.S. midterm elections.^[2]

A separate CNBC review reported that CoreWeave’s shares had fallen 36% during October through October 21 and that the company’s CDS spread had risen 179% over the prior year.^[2] Those figures describe a financing stress event, not a verified collapse in demand for compute. They do, however, change the economics of adding capacity: contracted revenue is less valuable to equity holders if the capital required to deliver it becomes prohibitively expensive.

The pressure is spreading beyond one issuer. CNBC reported that October losses among several neocloud companies had exceeded 30%, while SK Hynix discontinued the planned IPO of its Solidigm unit, citing market volatility and share-price trends.^[3] The article also reported record-high CDS spreads for several large technology companies as investors looked for ways to hedge the broader AI buildout.^[3]

The demand evidence is still constructive

Datadog’s latest available earnings-call commentary presents a different part of the chain. In its August 2026 call, management said AI activity was increasing cloud consumption, that more than 750 AI customers were using Datadog, and that agent-related tool calls had grown more than 22 times compared with Q4 2025.^[4] The company also described GPU monitoring and observability as ways for customers to improve utilization, efficiency and operational reliability.^[4]

Snowflake’s September 2026 call likewise pointed to continuing enterprise adoption rather than an abrupt pullback. Management reported 37% year-over-year product-revenue growth in the quarter, continued customer additions and a broadening set of AI products.^[5] Snowflake also emphasized cost and governance controls for AI workloads, an important detail as customers become more sensitive to consumption economics.^[5]

This is the distinction the market is now trying to make. Demand can remain strong while returns on new capacity become less certain. Customers may want more AI workloads, but they can still negotiate harder on price, optimize usage, delay deployments that lack a clear return, or favor vendors with stronger balance sheets and more flexible infrastructure.

Why this matters for the wider scope

The same test reaches beyond CRWV, DDOG and SNOW. For consumer-facing names such as AN, RH, WSM and LZB, the connection is less direct, but higher rates and tighter credit can still affect household confidence, big-ticket purchases and valuation multiples. ETH sits in a different risk bucket, where liquidity and technology sentiment can transmit quickly without the same operating linkage to enterprise cloud demand.

The market’s response suggests that investors are not treating every AI-adjacent exposure equally. CoreWeave’s premarket move was negative on October 26, while DDOG and SNOW were comparatively stable to positive in the same snapshot.^[1] That relative behavior is consistent with a rotation toward businesses that can monetize demand through software and services rather than having to finance enormous physical capacity first. It remains an inference, not a settled verdict.

The counterargument is important. Infrastructure bottlenecks can create scarcity value, and a company with contracted customers, specialized capacity and multiple funding channels may ultimately benefit if demand continues to outrun supply. CoreWeave’s CEO has said its model combines infrastructure-backed borrowing, corporate debt, equity and customer prepayments.^[2] The question is whether those channels remain available on terms that preserve shareholder value.

What to watch next

1. CoreWeave’s earnings and funding plan. The key evidence will be cash usage, committed capacity, customer concentration, debt maturities and the cost of new financing—not just revenue growth.
2. Credit-market confirmation. Watch unsecured yields, CDS spreads and the ability of AI infrastructure companies to issue or refinance debt. A stabilization in credit would matter more than a one-day equity rebound.
3. Permitting and power decisions. Data-center approvals, grid access and local political resistance can determine whether contracted demand becomes delivered revenue.
4. Software consumption quality. For DDOG and SNOW, the important questions are whether AI workloads move into production, whether usage expands across existing customers and whether optimization limits revenue growth.
5. Customer return on investment. Strong workloads with tighter governance and measurable productivity gains would support the durable-demand case. Capacity announcements without evidence of utilization would reinforce the financing-risk case.

Bottom line

The thesis that resilient demand can support DDOG, SNOW, CRWV, AN, RH, WSM, ETH and LZB over the next year is not disproved by the current move—but it is being stress-tested in a more specific way. The evidence still supports real AI and cloud usage growth at DDOG and SNOW. The warning signal is that leverage, refinancing and permitting can overwhelm that demand at the infrastructure layer.

The next phase of the trade will likely reward evidence of cash-generating demand and punish growth that requires continuously receptive credit markets. That is a narrower, more actionable market question than whether AI is simply “real” or “a bubble.”