TL;DR · 30-second read
The Short Version
NVIDIA, whose chips run many artificial intelligence systems, took in $96 billion in three months, more than double a year earlier.
The surprise is what it has signed up for. It promised $279 billion to suppliers, up from $119 billion three months ago, largely to lock in memory chips. It also agreed to cover up to $105 billion if OpenAI, the company behind ChatGPT, fails to pay for a giant Ohio computer campus planned to use about as much electricity as four large nuclear plants produce.
NVIDIA is no longer just selling chips. It is taking on the risk that the boom runs short of parts, power or money.
NVIDIA reported record revenue of $96.2 billion on August 26, 2026, for its second quarter of fiscal 2027, which ended July 26. Revenue rose 18% from the prior quarter and 106% from a year earlier. Data center revenue reached $89.0 billion, up 18% and 117% on the same comparisons. NVIDIA guided to third-quarter revenue of $108.0 billion, plus or minus 2%, with no data center compute revenue from China assumed.
CFO Colette Kress’s commentary, filed with an 8-K the same day, put supply commitments at $279 billion, up from $119 billion a quarter earlier and primarily tied to memory procurement. An 8-K filed August 17 disclosed guarantees capped at $105 billion covering about 4.25 gigawatts of computing load at SB Energy’s PORTS-Pike campus in Ohio, with an OpenAI affiliate as tenant.
Executive Summary
The headline figures point to strong demand. NVIDIA’s data center business added roughly $13.5 billion of revenue in a single quarter, gross margin (the share of sales left after direct production costs) held at 75.0%, and the outlook calls for about $11.8 billion more in the third quarter. Nothing in those results indicates that shortages of memory or power stopped NVIDIA from shipping this quarter.
The commitments tell a different part of the story. In 13 weeks NVIDIA added $160 billion in supply obligations and agreed to backstop a data center campus whose first capacity is expected in 2028. Its CFO commentary says securing land, power and shell “has become the next critical phase in the AI infrastructure buildout” and that many AI clouds and model makers “are growing faster than their balance sheets and long-term credit profiles can support.”
That matters to anyone who builds, powers or finances AI capacity. The company at the center of the build-out is using its financial strength to secure memory supply and data center sites before they become constraints, and in doing so it is taking on exposure that would traditionally sit with suppliers, tenants and lenders.
Demand Is Strong, but the Breadth Claim Needs Context
Jensen Huang said in the release that where one lab once drove the build-out, multiple frontier labs now scale in parallel, alongside startups and open-model developers. The segment data offers some support. Revenue from ACIE (AI Clouds, Industrial and Enterprise) rose 138% from a year earlier and 25% sequentially, while Hyperscale, the category for the largest cloud operators, more than doubled year over year and grew 13% sequentially.
Two details limit how far that evidence goes. CFO Colette Kress attributes ACIE growth partly to “hyperscalers utilizing AI clouds,” so some of that revenue traces back to the largest buyers. NVIDIA also moved one unnamed company from ACIE to Hyperscale this quarter and recast prior periods. And the campus NVIDIA chose to backstop has a single named customer, OpenAI, for up to 8 gigawatts.
The outlook implies slower sequential growth: $108.0 billion would be about 12% above the second quarter, compared with 18% just reported. In dollars, that is about $11.8 billion of added revenue versus $14.6 billion. Guidance is a forecast rather than a result, and it excludes China data center compute, where Hopper product shipments were already under 1% of data center revenue. It should not be read as evidence of a demand or supply ceiling on its own.
The $279 Billion Line Is Where Risk Changes Hands
The most consequential number in the quarter may be the rise in supply commitments from $119 billion to $279 billion. The $160 billion increase equals about 1.7 times the quarter’s revenue, and the total equals more than 11 quarters of NVIDIA’s current $24.1 billion cost of revenue. NVIDIA says the commitments secure critical components “for the next several years” and that they relate primarily to memory.
This is the mechanism behind the headline. A supplier that receives a long-term commitment has more certainty about selling its output, while the buyer takes on more of the risk that future demand, pricing or product needs turn out differently. NVIDIA is making that trade at very large scale, alongside a multiyear technology partnership with SK hynix on next-generation memory. Inventories also rose to $31.6 billion from $21.4 billion in January. Memory makers gain visibility from this kind of arrangement. Other buyers, if supply stays finite, may find less uncommitted output to compete for.
The results show memory pressure directly in only one place, and it is outside the data center. Kress says slower consumer PC sales were “tempered by elevated memory and systems prices,” even as Edge Computing revenue grew on workstation demand. So the answer to whether memory is limiting the AI build-out is nuanced: the data center numbers do not show a cap, but NVIDIA’s purchasing suggests it takes the possibility seriously enough to commit a sum far larger than a quarter’s revenue.
Power and Land Have Become a Credit Problem
The Ohio transaction shows the same pattern in physical infrastructure. Under the August 17 8-K, NVIDIA entered residual value guaranties with SB Energy for leases covering about 4.25 gigawatts of IT load, meaning the power used by the computing equipment itself. Its obligation is capped at $105 billion for the initial commitment. It also holds an option to support roughly 3.75 gigawatts more, a figure the 8-K rounds to about 3.8. In return, the campus will exclusively host NVIDIA infrastructure, subject to limited exceptions, and NVIDIA is separately investing $1.5 billion in SB Energy.
The structure turns a site-and-power problem into a credit solution. If OpenAI becomes insolvent and defaults, or fails to make lease payments, NVIDIA pays any shortfall between a guaranteed minimum lease value and what can be recovered through re-leasing or sale. The cap is therefore a ceiling, not an expected loss. OpenAI has agreed to reimburse NVIDIA, although the events that trigger the guarantee are also those in which a tenant’s capacity to reimburse could be most in question. The obligations end after 20 years or earlier on certain events, including OpenAI achieving a satisfactory credit rating.
Scale helps put this in perspective. NVIDIA earned $63.7 billion in GAAP operating income in the quarter and held about $99.4 billion in cash and marketable securities at quarter-end, so the $105 billion cap, which phases in and could only be called under specific conditions, does not present an immediate solvency question. The point is the change in role. With ready-for-service conditions expected beginning in 2028, the guarantee does not address near-term capacity. It is a long-dated effort to ensure future sites exist for future NVIDIA hardware, and a further $3.5 billion of maximum gross exposure covers other AI cloud partners’ lease obligations.
Reading Revenue When the Seller Also Helps Finance the Buyers
NVIDIA’s support for its ecosystem now extends across several channels. Its equity investments focus on AI model makers, infrastructure financiers and other private companies. Non-marketable securities on its balance sheet rose to $51.2 billion from $22.3 billion in January. It has announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR intended to mobilize over $500 billion of third-party capital over time, subject to definitive agreements. Some AI cloud agreements also let NVIDIA share in revenue if certain criteria are met.
None of this means the demand is not real, and CFO commentary describes extraordinary demand from customers constrained by their credit profiles. But it does complicate using NVIDIA’s revenue as a clean measure of independent end demand, because some buyers’ access to capacity increasingly depends on NVIDIA’s own support. Investors can see a related effect in earnings: GAAP net income rose only 2% sequentially, against 18% on a non-GAAP basis, because equity gains swing the GAAP figure. Net equity gains were $7.8 billion in the quarter.
For data center operators and enterprise buyers, the practical takeaway is that access to future capacity may increasingly come packaged with a specific vendor stack and financing structure. Long-term debt, meanwhile, rose to $32.4 billion from $7.5 billion in January, which adds another figure to monitor as these commitments mature.
Background
NVIDIA reports through two markets, Data Center and Edge Computing, and sells complete computing platforms that combine processors, networking, systems and software. Its fiscal year runs well ahead of the calendar, so fiscal 2027’s second quarter ended in July 2026. Blackwell is the architecture generating most current revenue, and Vera Rubin is its successor platform, now ramping. Within Data Center, Hyperscale covers the largest cloud operators, while ACIE covers AI clouds, industrial and enterprise customers.
AI data centers require far more than chips: advanced memory, high-speed networking, buildings and large amounts of electricity must all arrive on compatible schedules. SB Energy is the developer, owner and operator of the PORTS-Pike campus in Pike County, Ohio, where NVIDIA’s credit support and OpenAI’s tenancy link the chip supplier, the customer and the developer in one long-term structure. Source: NVIDIA Announces Financial Results for Second Quarter Fiscal 2027, NVIDIA Newsroom’s report of quarterly results, outlook and business highlights. Primary sources: NVIDIA 10-Q filed 2026-08-26; NVIDIA 8-K filed 2026-08-26; NVIDIA Announces Financial Results for Second Quarter Fiscal 2027 (Exhibit 99.1 to 8-K filed 2026-08-26); CFO Commentary on Second Quarter Fiscal 2027 Results (Exhibit 99.2 to 8-K filed 2026-08-26); NVIDIA 8-K filed 2026-08-17; NVIDIA Guarantees SB Energy’s PORTS-Pike Technology Campus in Ohio to Exclusively Host NVIDIA AI Compute (Exhibit 99.1 to 8-K filed 2026-08-17).Sources