Accelerated Computing Land, Power, and Shell Guarantees

Accelerated Computing Land, Power, and Shell Guarantees are contingent credit-support arrangements NVIDIA uses to help selected AI customers and partners secure data-center sites and lease capacity. They are not the same thing as supply commitments, revenue backlog, debt already funded, or an expected-loss estimate.

The scale changed materially in August 2026. NVIDIA's fiscal Q2 2027 filing shows $108.5 billion of combined maximum gross exposure after including a new $105 billion capped guarantee tied to SB Energy's PORTS Technology Campus in Ohio, in addition to an existing $3.5 billion AI-cloud guarantee pool.

That headline is economically important, but it is also easy to misuse. Maximum gross exposure is the contractual ceiling before trigger conditions, recoveries, reimbursement rights, and exposure runoff. It is not a forecast that NVIDIA will pay $108.5 billion.

The legacy AI-cloud guarantee pool is a credit derivative, not a funded lease balance

As of July 26, 2026, NVIDIA reported $3.529 billion of notional value for land, power, and shell guarantees covering selected AI cloud partners' data-center lease obligations. The filing describes $3.5 billion of maximum gross exposure under those agreements.

The partners had placed $712 million in escrow to partially mitigate NVIDIA's potential exposure. That escrow equals about 20% of the $3.5 billion maximum gross amount, although it should not be treated as a simple subtraction that proves a fixed net loss ceiling.

NVIDIA says the guarantee exposure declines as partners make payments to lessors over terms ranging from five to seven years. The guarantees are classified as credit derivatives. Their fair values were not significant at July 26, 2026, with changes in fair value recognized in Other income, net.

That accounting treatment is one reason not to equate notional exposure with current expense. The notional amount describes the contractual reference exposure. Fair value reflects the market value of the contingent obligation under the applicable assumptions and contract terms.

The SB Energy arrangement changes the scale and the contract mechanics

On August 17, 2026, NVIDIA entered into multiple residual-value guarantees with SB Energy related to approximately 4.25 gigawatts of IT load at the PORTS Technology Campus. OpenAI is the tenant, while NVIDIA provides credit support to help secure the land, power, and shell capacity.

NVIDIA's aggregate payment obligation for the initial commitment is capped at $105 billion. Combined with the pre-existing $3.5 billion guarantee pool, the filing reports $108.5 billion of maximum gross exposure. The SB Energy arrangement therefore represents about 96.8% of that combined maximum amount.

The $105 billion cap does not become an unconditional payable on signing. An agreement generally becomes effective when the applicable lease begins, and NVIDIA's obligations are subject to conditions that include the lessor satisfying ready-for-service requirements for the relevant premises. The company said those conditions were expected to begin being met in 2028.

NVIDIA also has the option, in its sole discretion, to provide credit support for roughly another 3.8 gigawatts at the site. That optional capacity is not the same thing as the initial 4.25-gigawatt commitment and should not be added automatically to the $105 billion capped obligation.

Payment depends on a trigger event and the recovery value of the site

The SB Energy agreements define specific trigger events. NVIDIA's payment obligation can arise if OpenAI becomes insolvent and defaults under a lease, or if OpenAI fails to make required lease payments.

Even after a trigger, the relevant payment is generally tied to a shortfall between the guaranteed minimum lease value and amounts recovered through a replacement lease or sale. That means the contract contains a recovery mechanism rather than treating the entire capped amount as an automatic cash payment.

NVIDIA can also choose among several remedies after a trigger. The company may assume the lease, require the lessor to seek a replacement tenant, initiate a sale process, allow the lease to terminate, or defer those remedies for up to one year while paying specified project-agreement costs.

The agreements also give NVIDIA reimbursement protection. OpenAI agreed to reimburse and indemnify NVIDIA for amounts NVIDIA actually pays to the lessor under the guarantees.

Those provisions do not eliminate risk. They change the shape of it. Investor analysis should focus on tenant credit quality, the recoverability and re-leasing value of the facilities, technology obsolescence, timing of site delivery, and the enforceability and value of reimbursement rights rather than treating the contractual cap as a current cash outflow.

Maximum gross exposure is different from upstream supply commitments

NVIDIA separately reported $279 billion of supply and capacity commitments used to secure memory, manufacturing capacity, components, and related infrastructure. The $108.5 billion guarantee total is about 39% of that amount numerically, but the two balances represent different economic risks.

The legacy $3.5 billion guarantee pool was small beside NVIDIA's $279 billion of supply and capacity commitments, at about 1.3% of that balance. Even then, it was contingent exposure, not an immediate expense or a forecast of losses. The new SB Energy arrangement changes the scale of the downstream exposure without changing that accounting and risk distinction.

Supply and Capacity Commitments are principally upstream obligations supporting NVIDIA's production plan. Land, power, and shell guarantees are downstream credit support intended to help customers or partners secure deployment capacity.

Adding the two balances together does not produce a useful expected-loss estimate. One can create procurement and inventory risk if demand is overestimated; the other can create contingent credit and residual-value exposure if a supported tenant defaults and recoveries are insufficient.

For demand context, pair the guarantee analysis with Data Center Revenue Growth. Rapid AI infrastructure demand can make additional deployment capacity strategically valuable, but the guarantee structure means NVIDIA is accepting a separate balance-sheet and contract risk to help that capacity get built.

Primary source: NVIDIA fiscal Q2 2027 Form 10-Q.

Additional source: NVIDIA August 17, 2026 Form 8-K.

Explore more topics in the Financial Research Encyclopedia.