#CoreWeave tuyên bố không cần vay nợ nhưng sau đó lại phát hành trái phiếu 4 tỷ USD
About CoreWeave tuyên bố không cần vay nợ nhưng sau đó lại phát hành trái phiếu 4 tỷ USD
Theo bài đăng đại diện, ban lãnh đạo CoreWeave ($CRWV) cho biết trong cuộc họp qua điện thoại rằng công ty có đủ tiền mặt và không cần vay thêm nợ, điều này đã gây ra hiện tượng ép giá bán khống, khiến giá cổ phiếu tăng khoảng 20% vào ngày hôm sau; tuy nhiên, sau khi đóng cửa, công ty lại thông báo kế hoạch huy động ít nhất 4 tỷ USD thông qua khoản vay cấp đầu tư, do Morgan Stanley và Mitsubishi UFJ dẫn dắt, với lợi suất dự kiến từ 7% đến 7,5%, giá cổ phiếu sau giờ giao dịch đã từng giảm 6%. Điểm chú ý của thị trường là sự tương phản giữa việc phục hồi hình ảnh tín dụng và việc tăng đòn bẩy trở lại, cũng như ảnh hưởng tiếp theo đến khẩu vị rủi ro từ mảng sức mạnh tính toán AI (báo cáo tài chính Nvidia, v.v.).
Nổi bật
Mới nhất
CoreWeave tuyên bố không cần vay nợ nhưng sau đó lại phát hành trái phiếu 4 tỷ USD Bài đăng phổ biến
CoreWeave’s proposed GPU- and contract-backed loans expose a useful distinction: a $132.5B backlog signals demand visibility, not automatically cash available for expansion. Funding at 7%-7.5%, versus roughly 14%-15% unsecured-bond yields, could improve carrying costs. The remaining test is whether contracted demand converts into cash flow quickly enough to support new data centers.
#CoreWeave4BFinancing
$CRWV vs $NBIS is a fascinating case study in "narrative follows price"
The fintwit consensus has been "Nebius is the superior neocloud with better financing"
In reality, the price performance differential has likely been largely structural flows:
- Leopold et al. bid billions into NBIS
- Magnetar, Intrator et al. sold billions out of CRWV
- Since Leopold liquidation, CRWV/NBIS ratio has been largely flat, the relative outperformance ended.
NBIS outperformance earlier in the year led to the invention of various narratives in support of Nebius, which have not held up under further scrutiny.
1. "Nebius has better financing, Coreweave is a mess of debt."
The inverse seems true now if anything. It now seems like Coreweave was quite sharp to secure debt financing and establish lender partnerships at reasonable rates. On the other hand, Nebius has a multibillion dollar financing gap with no access to debt markets.
2. "Nebius executes better, Coreweave has poor execution."
While this was true in the past, they were also at very different points in the buildout ramp. Nebius was executing off a much smaller, lower complexity base, and is now running into delays at the bigger part of the ramp. On the other hand, Coreweave seems to have pushed past difficulties in their ramp and hit their stride.
3. "Nebius is stronger technically, Coreweave has the weaker technical team."
This has never really matched the facts on the ground. They are held in similar regard throughout the industry, with similar uptimes, negotiation leverage in contracts, managed inference offerings, etc. Coreweave was the first to overcome various technical challenges, like bringing Vera Rubin to production workloads. Doesn't seem like there is a clear winner here.
I have traded both of these stocks in the past, but haven't owned either for awhile now. I don't identify with either side in the CRWV vs NBIS wars. Just think that the narratives painted around these stocks are very interesting.
Saso Capital
$NBIS notes from earnings
Pay attention to this. About $325M of that revenue was non-recurring, booked when partners deployed Nebius architecture on capacity they funded themselves (the asset-light model). Strip it out and Q3 was roughly $656M, up about 13% from Q2's $582M. That is below consensus. It likely flatters the margin too, since enabling a partner's capex shouldn't carry much cost (my inference, they don't break it out).
Prepayments on Q3 deals covered more than 100% of the associated capex, up from 50 to 60% in Q2. Over 90% of contract value signed carried a prepayment. Payback on comparable 1 to 3 year deals dropped to 19 months from 22, before counting any prepayment. ACV (annual contract value) per MW topped $25M, and pricing per GPU-hour rose more than 20% QoQ.
Volozh addressed the elephant directly, noting the industry has been "asked questions about how the AI build-out will be financed." Their answer is $10B of committed prepayments in 2026, rising to $17 to 20B in 2027.
Enterprise logos showed up. Databricks, Datadog and Snowflake signed multi-year deals, and Palantir named Nebius a preferred AI cloud partner. Until now the base was frontier labs and AI natives.
Token Factory, their managed inference platform, grew contract value 17x QoQ and is approaching a multi-hundred-million run rate less than a year after launch. Selling inference on top of owned infrastructure is how they earn more per MW than a pure GPU landlord.
They are shifting to owned sites. Missouri, Lappeenranta, a second Mantsala site and Alabama start deploying in 2027, over 3 GW at full scale. Owned capacity goes from over 10% of today's base to over half of the next gigawatt and over 75% of the one after.
So what does the market have to chew on? Two things.
First, the 2026 guide slid. Revenue now lands at the low end of $3.0 to 3.4B because third-party colocation capacity, including the remaining New Jersey tranches for Microsoft, moved into 1H27. The $7B+ ARR target moved from December to February. Simple math: $3.0B minus $1.96B through nine months implies about $1.04B in Q4, barely above a Q3 headline that included the one-off.
Second, 2026 capex is now about $27B, above the $20 to 25B range. 2027 capex is $40 to 45B, versus roughly $29B analysts were modeling three weeks ago. Prepayments and asset-backed financing on investment-grade contracts cover 75 to 85%, which still leaves $6 to 11B to find elsewhere. They already did $5.75B of converts in August and about $1.0B through the ATM in Q3.
A neocloud can have eager customers, busy GPUs, and an uncomfortable call with its lender, all in the same week. A financing squeeze can arrive before a demand slump.
In the latest Sequence Opinion I look at why.
There are four pressures. Money goes out before it comes in: sites, power, and GPUs get paid for before customers pay. Delivery is harder than signing, because a cluster only bills once the grid connection, permits, and cooling are done. GPUs have an uncertain economic shelf life. And "diversified" demand often leads back to the same few wallets: CoreWeave's top three customers were 72% of its Q2 revenue.
Take a cluster that collects $100 and pays $40 to operate and $50 in debt service. If collections drop to $85, a $10 cushion becomes a $5 shortfall, even though the cluster is still busy.
My view is that this changes ownership and business models before it proves AI infrastructure was a mistake. Buyers will pick off power connections, sites, and contracts rather than whole companies. The telecom buildout is the closest parallel: society got the networks, but the original investors didn't always get their money back.
The question I keep coming back to: is delivered capacity turning into cash fast enough to cover the debt?
AI infrastructure requires serious capital.
$CRWV’s reported $4B financing is another reminder of how expensive the AI cloud race has become.
More funding can support expansion, but it also raises questions about debt, costs, and long-term profitability.
Growth is exciting. Sustainable growth is what matters.
CoreWeave $CRWV plans to raise at least $4B through an investment-grade loan, Bloomberg reports.
Morgan Stanley & MUFG are leading the proposed deal, expected to be backed by GPUs & customer contracts.
Pricing is seen at 7%-7.5%, about half the yield on its junk bonds. It could launch next week.
CoreWeave May Have Just Saved AI Stocks $CRWV #markets #stockmarket #finance
