CoreWeave Data Center Bonds Raise $1.1 Billion, but Investors Demand a 9.25% Yield

CoreWeave data center bonds raised $1.1 billion, but investors demanded a 9.25% yield to finance the associated Virginia project. The borrowing cost turns a routine construction deal into a revealing test of confidence in the AI infrastructure boom.

The five-year bonds finance the first building at the Digital Drive campus near Richmond. CoreWeave has leased the planned facility’s entire 76 megawatts of critical IT capacity for 15 years.

That long contract gives bondholders visible rental income. Yet the securities were issued at 98.5 cents on the dollar, according to the original bond pricing report. Their yield was about 2.7 percentage points above the average for similarly rated debt.

The gap captures the central conflict. Investors still want exposure to AI data centers, but they are charging heavily for tenant concentration, construction risk, and CoreWeave’s speculative-grade credit.

The issuer is not CoreWeave itself. DDC 01 Propco LLC, a project company sponsored by affiliates of Blue Owl Capital, Cedarwood Investment Group, and PowerHouse Data Centers, sold the debt.

That separation matters because the bonds depend on one building, one lease, and one primary tenant. The arrangement shifts construction financing away from CoreWeave’s direct corporate balance sheet, but it does not eliminate CoreWeave risk.

CoreWeave Data Center Bonds Fund One Building and One Tenant

The bond sale converts CoreWeave’s lease commitment into financing for a data center that is still under construction.

DDC 01 is developing Building 1 at the Digital Drive campus in Chesterfield County, Virginia. The project will provide 76 megawatts of critical IT load, the electricity available directly to computing equipment.

The issuer raised approximately $1.1 billion through senior secured notes. Senior secured debt gives investors first-ranking claims over specified assets, contracts, and cash flows if the borrower defaults.

CoreWeave has committed to lease all available capacity for 15 years. The agreement includes two additional five-year extension options that CoreWeave can exercise.

The base lease is reportedly worth $2.94 billion. It also includes a 3.5% annual rent escalator, which increases payments throughout the initial term.

That structure offers more revenue visibility than a data center built without an anchor tenant. It also creates a straightforward financing mechanism: expected lease payments support debt service while the building becomes operating infrastructure.

The lease uses a triple-net structure. CoreWeave must cover operating expenses, property taxes, and insurance in addition to rent.

CoreWeave also lacks an ordinary right to terminate the lease for convenience. Before rent commencement, qualifying termination would require repayment of secured debt and 120% of the landlord’s non-debt investment.

After commencement, the termination payment would equal the present value of remaining rent, discounted at 6.5%. These provisions give the project meaningful contractual protection.

They do not guarantee performance. A contract is valuable only while the tenant can meet its obligations and still needs the capacity.

The project’s sponsors are placing that question directly before bond investors. Buyers accepted the exposure, but only after receiving a yield far above comparably rated debt.

The 9.25% yield is therefore not evidence that funding disappeared. It is evidence that funding remains available when borrowers compensate investors for identifiable risks.

That distinction matters for the wider AI buildout. Developers need enormous amounts of capital before a single GPU starts generating cloud revenue.

Power connections, cooling systems, buildings, transformers, and networking equipment require funding years before utilization becomes certain. A long CoreWeave lease helps bridge that gap.

However, the lease also concentrates the project’s economics around one company. The entire facility is designed around a tenant whose own growth depends on sustained demand for rented AI computing capacity.

Investors are not simply underwriting a Virginia building. They are underwriting the durability of CoreWeave’s business model over a period much longer than many end-customer contracts.

The 9.25% Yield Is the Story Behind the Headline

Investors funded the project, but the yield shows they did not treat a long AI lease as low-risk infrastructure income.

The bonds carry a five-year term and were issued below their face value. That discount lifted the yield to 9.25%, even though bondholders received security over project assets and contractual cash flows.

S&P Global Ratings assigned DDC 01 a B+ issuer rating. It rated the senior secured notes BB-, one level higher because their collateral improves expected recovery.

Both ratings remain below investment grade. That classification explains why the offering belongs in the high-yield, or junk-bond, market.

S&P estimated a 75% rounded recovery in a default scenario. Its full credit assessment identified both the lease protections and the project’s heavy dependence on CoreWeave.

The rating agency expects stable rental income after construction, supported by the escalator and long lease. However, CoreWeave’s B+ credit rating constrains the project during construction.

That constraint reflects the funding mechanism. The project relies partly on payments from CoreWeave to cover remaining construction costs.

S&P said CoreWeave was already paying a monthly amount equal to a 10% yield on cost. This support strengthens construction funding, but it also deepens the project’s reliance on the tenant.

The bond pricing provides a market judgment beyond the ratings. A yield approximately 270 basis points above similarly rated debt indicates that buyers wanted an additional premium.

One basis point equals one-hundredth of a percentage point. The reported gap therefore equals about 2.7 percentage points.

Several risks help explain that premium. The project remains unfinished, its income depends on one tenant, and the site lacks the depth of alternative demand found in larger data center markets.

CoreWeave also operates an unusually capital-intensive cloud business. It must acquire computing equipment while securing buildings, power, and customer commitments on overlapping schedules.

Its expansion has produced a large contracted backlog, but it has also required extensive borrowing. CoreWeave reported $35.55 billion in scheduled principal payments at June 30, 2026.

That total included debt owed across future periods rather than an amount immediately due. Still, it shows the scale of the obligations supporting the company’s expansion.

CoreWeave also disclosed $103.7 billion of remaining performance obligations, or contracted revenue not yet recognized, at the same date. Forty-one percent was expected within the following 24 months.

Those figures show both sides of the credit argument. CoreWeave has substantial contracted demand, but it needs substantial financing to build the infrastructure required to fulfill it.

During the first half of 2026, CoreWeave spent $14.1 billion on property and equipment. Its net investing cash outflow reached $14.9 billion.

Financing supplied $14 billion of net cash during that period. This near correspondence illustrates how closely infrastructure delivery remains tied to access to capital.

The company’s quarterly filing said future investments would continue to require a mix of debt, equity, equipment financing, and available cash.

Digital Drive extends that model beyond CoreWeave’s direct borrowing. A separate project company owns the facility, raises the construction debt, and receives rent from CoreWeave.

The structure can match specific financing with a specific asset. It can also make the project’s main risk easier to see.

If CoreWeave performs, the bondholders receive debt service from a protected stream of escalating rent. If it struggles, the project must find another tenant or enforce contractual remedies.

The yield represents the price investors demanded for living with that uncertainty.

Long Leases Cannot Remove CoreWeave’s Credit Risk

The lease protects the project from ordinary cancellation, but it cannot make the tenant stronger than CoreWeave itself.

CoreWeave’s 15-year commitment is much longer than many contracts between cloud providers and their customers. That maturity mismatch sits at the center of the financing risk.

CoreWeave disclosed that committed contracts accounted for 98% of its revenue during the second quarter of 2026. These take-or-pay agreements require customers to purchase contracted capacity regardless of actual usage.

However, the company also warned that customers might shift toward pay-as-you-go arrangements. Such a change would make future cash flows less predictable.

CoreWeave’s newer financing arrangements are already accommodating shorter customer commitments. In August, it closed a $2.6 billion delayed-draw term loan with an approximate five-year maturity.

The customer contracts supporting that facility averaged approximately three years. CoreWeave said the structure allowed it to finance shorter engagements while assuming renewal or remarketing risk.

That loan structure demonstrates why a long building lease cannot be evaluated independently from shorter cloud contracts.

CoreWeave must continually renew customers, replace expiring workloads, or find new demand. Its rent obligation to DDC 01 continues even if a customer contract ends earlier.

S&P highlighted this mismatch directly. It said weaker AI demand or excess infrastructure supply could make CoreWeave’s long lease obligations harder to honor.

The Digital Drive parties have designed a possible enhancement. DDC 01 signed an August memorandum concerning the proposed transfer of the lease to a CoreWeave financing subsidiary.

That subsidiary would hold a contract with an investment-grade hyperscaler. Project rent would receive payment priority before distributions to CoreWeave’s parent company.

A springing corporate guarantee from CoreWeave would also support the structure under specified conditions. S&P said the reassignment would improve the lease, but would not equal a direct investment-grade tenancy.

That qualification is important. The hyperscaler contract might strengthen the payment chain without making the hyperscaler directly responsible for rent.

Investors therefore remain exposed to legal structure, contract performance, and CoreWeave’s continuing role between the end customer and the building owner.

This project finance model is not unique to Digital Drive. Developers increasingly pair long leases with secured borrowing to fund specialized AI campuses.

Blue Owl, Chirisa Technology Parks, and PowerHouse previously sponsored another Virginia project leased solely to CoreWeave. A bank group supplied $600 million in debt financing.

That earlier Virginia financing supported a site expected to reach 120 megawatts of total capacity. It demonstrated lender appetite for facilities with CoreWeave as the anchor tenant.

Other developers have also tapped bond markets. Applied Digital raised high-yield debt for capacity connected to CoreWeave, while Galaxy Digital financed a much larger Texas campus.

Galaxy’s Helios expansion offers a useful comparison. Its second phase is planned for 260 megawatts of critical IT load under a 15-year CoreWeave lease.

The company’s project presentation cited $10.4 billion of minimum contracted lease payments for that phase. It also included annual rent escalation.

These transactions show a repeatable pattern. A developer controls land and power, CoreWeave signs a long lease, and capital providers fund construction against the resulting cash flows.

The pattern can accelerate capacity delivery without forcing CoreWeave to own every building. It also spreads exposure across project companies, lenders, bondholders, and infrastructure investors.

However, diversification among financiers does not automatically diversify the operating risk. Multiple projects can still depend on the same tenant and the same market for AI computing.

The Digital Drive yield suggests investors recognize that connection. They are willing to support the model, but not at conventional infrastructure borrowing costs.

Construction Delays Add Risk Before the GPUs Arrive

Digital Drive must still turn contracts, purchased equipment, and power commitments into a functioning facility.

The project was already behind schedule when S&P published its assessment. As of July 24, 2026, its three construction phases were between 111 and 126 days late.

S&P attributed the delays mainly to a longer county permit process and adverse winter weather. The project responded by accelerating its construction schedule.

Work expanded to six days each week, including four ten-hour days. The general contractor is DPR Construction, which has worked with the project’s sponsors on other nearby data centers.

Equipment procurement appears more advanced. By September 21, the project had acquired all owner-furnished equipment intended for contractor installation.

That total included all long-lead equipment, a category covering components that require early orders because manufacturing and delivery can take many months.

Power remains another critical dependency. The project has a 100-megawatt electricity service agreement with Dominion Energy.

A 300-megawatt substation serving the wider campus has been completed and handed to Dominion for energization. Transmission work for the project was scheduled for April 2027.

Neither a completed substation nor an electricity agreement guarantees timely service. Transmission construction, testing, permits, and final connections must still align with the building schedule.

CoreWeave’s lease offers DDC 01 unusual protection against ordinary delays. The tenant does not receive rent credits and cannot cancel merely because delivery slips.

Certain extended force majeure events can create termination rights. Power delays also qualify as force majeure under the agreement.

S&P said CoreWeave would generally need to face 24 consecutive months of qualifying unavailability before terminating on that basis. This gives the project time to correct setbacks.

The construction contractor may also owe liquidated damages for delays. These predetermined payments can offset some financial harm, but they cannot create computing capacity on time.

A delayed data center affects more than the developer. CoreWeave coordinates building availability with GPU deliveries, networking, customer deployment dates, and financing draws.

If one schedule slips, equipment can sit unused or customers can wait for contracted capacity. Both outcomes reduce the economic value of rapid expansion.

The project also faces conventional construction pressures. S&P identified additional weather, supply-chain disruption, skilled-labor shortages, and scheduling complexity as continuing risks.

Those risks become more consequential in AI facilities because of their power density and specialized cooling requirements. The building must support equipment configurations that place exceptional demands on electrical and thermal systems.

Digital Drive’s Virginia location offers advantages, including established fiber networks, tax benefits, and expanding power infrastructure. Yet S&P does not consider Richmond a tier-one data center market.

That judgment matters if CoreWeave cannot occupy the facility. A property in Northern Virginia’s largest data center clusters would likely have a deeper pool of replacement tenants.

Digital Drive sits near an emerging market rather than inside the country’s most liquid leasing hub. Finding another tenant could therefore take longer or require costly modifications.

The property’s 76-megawatt scale is also large enough to narrow the replacement pool. Few companies can absorb that volume of high-density capacity on short notice.

The lease mitigates this problem while CoreWeave performs. It does not solve the remarketing challenge following a serious tenant failure.

This is why secured collateral and estimated recovery cannot replace credit analysis. Bondholders might recover substantial value after default while still experiencing delayed payments or principal losses.

The 9.25% yield compensates investors for taking those possibilities. It does not make them disappear.

AI Infrastructure Funding Has Become a Credit Test

The Digital Drive financing shows that AI data center demand remains bankable, but the market now distinguishes growth from credit quality.

CoreWeave’s contracted demand provides a strong foundation for new construction. Its June backlog and take-or-pay revenue mix suggest customers still want substantial AI capacity.

The company also reported $15.55 billion of liquidity at June 30, including cash and borrowing availability. That amount helps support near-term deployment plans.

CoreWeave’s scale, however, magnifies execution risk. Its future debt payments, equipment commitments, leases, and joint venture investments all compete for cash.

The company held interests in unconsolidated data center ventures and had committed up to $1.7 billion to two such projects by June. These arrangements expand capacity while distributing ownership.

They also make the complete financing picture more complicated. Investors must evaluate corporate debt, project debt, leases, equipment financing, and affiliated entities together.

Digital Drive gives the market a relatively clean case study. One project company owns one building and depends on rent from one CoreWeave lease.

The project has meaningful protections. It holds contracted capacity, escalating rent, secured collateral, a triple-net structure, and restrictions on tenant termination.

It also has clear weaknesses. Construction is late, the tenant is below investment grade, and the local market offers limited replacement depth.

That combination explains why the sale succeeded while still offering a rich yield. Investors did not reject AI infrastructure, but they refused to treat it like low-risk real estate.

The outcome also pressures future borrowers. Comparable projects will now face Digital Drive’s pricing when negotiating their own bond terms.

A developer with a stronger tenant, better location, completed construction, or more diversified income should seek a lower yield. A weaker project might have to pay more.

CoreWeave is also under pressure because its name affects financing costs beyond its direct debt. A downgrade or weaker operating results could raise capital costs for its landlords and development partners.

That feedback loop can reach cloud customers. Higher financing costs can influence capacity prices, deployment timelines, and the economics of long-term AI workloads.

It can also change which projects get built. Capital providers may favor campuses with secured power, advanced construction, and contracts linked to investment-grade customers.

Smaller or speculative developments could struggle, even while overall AI infrastructure spending remains high. Aggregate demand does not guarantee financing for every site.

The Digital Drive transaction therefore marks a transition from broad enthusiasm to project-level discrimination. Investors want to know who pays, when power arrives, and what collateral survives a default.

They also want compensation when those answers depend heavily on one fast-growing company.

Three Signals Will Show Whether the Premium Was Justified

Construction progress, the proposed lease transfer, and CoreWeave’s financial performance will determine whether these bonds look cautious or expensive.

The first signal is Digital Drive’s delivery schedule. Investors should watch whether Dominion completes the transmission connection and whether the project prevents further construction delays.

Progress near the planned 2027 power milestone would support S&P’s stable outlook. Additional delays or cost overruns would weaken expected coverage before rental operations fully begin.

The second signal is the proposed lease assignment. Moving the lease to a financing subsidiary with an investment-grade hyperscaler contract would improve the project’s payment structure.

Investors should examine the final terms rather than assume that assignment creates an investment-grade obligation. Payment priority and direct legal liability are not the same protection.

A direct guarantee from the hyperscaler would materially change the credit story. The described memorandum does not provide that outcome.

The third signal is CoreWeave’s balance between contracted revenue and capital consumption. Backlog must translate into operating cash before debt and lease obligations overwhelm financing flexibility.

Future filings should reveal changes in remaining performance obligations, investing cash flow, total debt, liquidity, and customer concentration. Those measures matter more than headline capacity announcements.

The strongest outcome would combine timely completion, a strengthened lease, and improving cash generation. That would suggest the 9.25% yield adequately rewarded investors for temporary construction and credit concerns.

A weaker outcome would involve more delays, continued dependence on CoreWeave’s parent credit, and faster growth in obligations than in cash generation. That would validate the market’s premium.

For developers and enterprise AI buyers, the lesson is practical. Physical capacity does not arrive when a lease is signed or a bond is sold. It arrives when financing, equipment, construction, power, and customer demand work on the same timetable.

The CoreWeave data center bonds show that capital is still available for that coordination challenge. They also show that investors now attach a visible price to every weak link. Watch those three signals before treating the sale as either confidence in AI demand or a warning about excess debt.