NVIDIA Corporation (NVDA) FY2026 10-K and Q2 FY2027 10-Q Summary

Business

From the fiscal 2026 Form 10-K, accession 0001045810-26-000021.

NVIDIA pioneered accelerated computing and describes itself today as a data-center-scale AI infrastructure company. Its technology stack starts with the CUDA development platform, which runs on every NVIDIA GPU, and extends through hundreds of domain-specific software libraries, frameworks, SDKs and APIs used for AI model training and inference, data analytics, scientific computing, robotics and 3D graphics. Beginning with the Blackwell architecture, the company designs at data-center scale, chips, networking, systems, software and algorithms architected together so that hundreds of thousands of GPUs can be interconnected to operate as a single computer.

The company has invested more than $76.7 billion in research and development since inception. Landmarks it cites include the invention of the GPU in 1999, CUDA in 2006, the first Tensor Core GPU in 2017, the first autonomous-driving SoC in 2018, the Mellanox acquisition in 2020 (which added networking and the data processing unit, or DPU), the Grace data center CPU in 2023 and the Blackwell architecture in 2024. In fiscal 2026 NVIDIA launched and scaled Blackwell Ultra, optimized for agentic, reasoning and physical AI, and introduced NVLink Fusion, which lets hyperscalers and custom ASIC designers integrate their own CPUs and XPUs with the NVIDIA platform. It also unveiled the Rubin platform, built for agentic AI and reasoning, which the 10-K said was expected to enter production shipments in the second half of fiscal 2027 and to deliver up to a 10x reduction in cost per token versus Blackwell.

Reportable segments. NVIDIA reports two:

• Compute & Networking, Data Center accelerated computing and networking platforms, AI solutions and software, plus automotive platforms and autonomous/electric-vehicle solutions.
• Graphics, GeForce GPUs for gaming and PCs, and Quadro/NVIDIA RTX GPUs for enterprise workstation graphics.

End markets. The 10-K describes four end markets: Data Center, Gaming, Professional Visualization and Automotive. (As discussed below, the company changed its market-platform presentation in fiscal 2027 to Data Center, split into Hyperscale and AI Clouds, Industrial & Enterprise, and Edge Computing. The two reportable segments are unchanged.)

Data Center is the dominant business. The platform comprises compute and networking infrastructure typically delivered as rack-scale systems, subsystems or modules, together with software and services. Networking offerings include NVLink interconnects and switches and end-to-end InfiniBand and Ethernet platforms, network adapters, cables, DPUs, switch chips and systems. Customers include every major public and private cloud provider, AI model makers, enterprises, startups and public-sector entities, reached both directly and through OEMs, ODMs, system integrators and distributors. Software is monetized both as an integral part of the platform and through paid licenses to NVIDIA AI Enterprise and vGPU software.

Gaming products are GeForce RTX GPUs for desktop and laptop PCs, the GeForce NOW cloud gaming service and SoCs and development services for game consoles. Professional Visualization serves design, engineering and content-creation workflows through RTX PRO GPUs and ISV partnerships. Automotive is delivered under the DRIVE Hyperion platform, DRIVE AGX compute hardware, the DRIVE OS in-vehicle operating system, a reference sensor set and an open, modular software stack for autonomous driving, mapping and parking.

Strategy. Management frames five strategies: advancing the accelerated computing platform through full-stack innovation that outpaces Moore's Law; extending AI platform leadership across training and inference (GPUs, CPUs and DPUs, plus NVIDIA AI Enterprise with NIM, NeMo and AI Blueprints); extending computer-graphics leadership; advancing the autonomous-vehicle platform; and licensing its intellectual property through license and development arrangements. The ecosystem is a stated moat: more than 7.5 million developers worldwide use CUDA and NVIDIA's other software tools, and the platforms are available from virtually every major server maker and cloud provider.

Manufacturing. NVIDIA is fabless. It uses TSMC and Samsung for wafers; buys memory from SK hynix, Micron and Samsung; uses CoWoS packaging; and engages Hon Hai (Foxconn), Wistron and Fabrinet for assembly, test and packaging. Supply is concentrated in Asia, and the company is expanding into the U.S. and Latin America to add redundancy. In periods of growth it places non-cancellable component orders ahead of historical lead times and pays premiums or deposits to secure capacity.

Competition. Competitors named in the 10-K include AMD, Huawei and Intel in GPUs and accelerators; Alibaba, Alphabet, Amazon, Baidu, Huawei and Microsoft among large cloud providers designing internal silicon; Amazon, Huawei and Microsoft in Arm-based CPUs; Ambarella, AMD, Broadcom, Intel, Qualcomm, Renesas, Samsung and Tesla in SoCs; and AMD, Arista, Broadcom, Cisco, HPE, Huawei, Intel, Lumentum and Marvell in networking.

Other. Desktop gaming products typically see stronger revenue in the second half of the fiscal year. Issued patents have expiration dates from March 2026 to June 2045. As of the end of fiscal 2026 the company had approximately 42,000 employees in 38 countries, 31,000 of them in research and development; more than 80% of the workforce hold technical roles, more than half hold an advanced degree, over 40% of fiscal 2026 new hires came from employee referrals, and turnover was 3.7%. NVIDIA was incorporated in California in April 1993, reincorporated in Delaware in April 1998, and is headquartered in Santa Clara, California.

Risk factors

Condensed from Item 1A of the fiscal 2026 Form 10-K, accession 0001045810-26-000021, updated by Part II, Item 1A of the Form 10-Q, accession 0001045810-26-000075.

Export controls and China. This is the most heavily developed risk in both filings. Since August 2022 NVIDIA has been subject to expanding U.S. export restrictions, covering A100/H100 (2022–2023), a broad performance-threshold rule for China and Country Groups D:1/D:4/D:5 (October 2023), and an April 2025 license requirement for H20 and any circuit matching its memory or interconnect bandwidth, which produced a $4.5 billion charge in the first quarter of fiscal 2026 for excess H20 inventory and purchase obligations. Licenses granted from August 2025 would have permitted certain H20 shipments, but PRC-side restrictions prevented sales of that inventory. Beginning February 2026 the U.S. government licensed small volumes of H200 to specific China-based customers; PRC restrictions again limited sales, and NVIDIA took a $0.4 billion H200 excess-inventory and purchase-obligation charge in the first half of fiscal 2027. Shipments made under the H200 program were less than 1% of Data Center revenue in the most recent quarter, must pass a U.S. inspection process, and carry a 25% tariff on importation into the United States that the company has been unable to pass to customers. As of the end of the second quarter of fiscal 2027, NVIDIA states it was effectively foreclosed from China's data center compute market while still able to ship uncontrolled gaming and workstation GPUs, and that this foreclosure has helped competitors build larger developer and customer ecosystems worldwide. The January 2025 AI Diffusion interim final rule, which would have imposed worldwide licensing, was slated for rescission and replacement in May 2025; the replacement rule's scope and timing remain uncertain. Pending U.S. legislation (the GAIN AI Act, the Remote Access Security Act) could further constrain licensing flexibility and cloud services. On the Chinese side, antitrust regulators published a preliminary finding on September 15, 2025 that NVIDIA's compliance with U.S. export controls, which required offering degraded products in China, discriminated unfairly against Chinese customers and violated the terms of China's approval of the Mellanox acquisition; adverse conclusions could bring financial penalties or restrictions on the networking business in China.

Commitments, guarantees and counterparty exposure (new risk factor added this quarter). To secure supply and capacity, obtain cloud services and lease data center capacity, NVIDIA has committed capital requiring substantial payments over many years that it may be unable to reduce if demand or plans change. It also extends financial guarantees, credit support and data center leases to support customers' and partners' AI infrastructure buildouts. Counterparties may fail to fund commitments, secure capital or infrastructure, complete projects on time or within budget, or may enter financial distress; power constraints, government action, permitting delays or community opposition may impede data center development. Extended payment terms under large multi-quarter agreements with investment-grade customers will continue to affect the timing of operating cash flows. Under the AI cloud agreements, if the AI clouds do not sell committed capacity to third parties, NVIDIA has agreed to purchase that capacity and may lack sufficient demand or operational ability to use or resell it. The SB Energy guarantees tied to the PORTS Technology Campus may expose the company to substantial obligations over extended periods; if OpenAI does not perform or becomes insolvent and a guarantee is triggered, NVIDIA may assume the lease, seek a replacement tenant, pursue a sale or other remedies, may not find a replacement on acceptable terms, and may incur ongoing lease costs or substantial payments, OpenAI's agreement to reimburse and indemnify certain losses may not be recovered promptly or in full. The August 2026 memoranda of understanding with large capital providers are preliminary and may not lead to definitive agreements.

Indebtedness (new risk factor added this quarter). As of July 26, 2026 NVIDIA had $33.5 billion aggregate principal of senior notes outstanding plus a $25.0 billion commercial paper program with nothing drawn. Refinancing may come at less favorable terms or be unavailable; debt service, covenants and rating changes could constrain flexibility, liens, mergers and asset dispositions.

Demand, supply and manufacturing. Long lead times, sometimes beyond twelve months, combined with building inventory ahead of anticipated demand have repeatedly produced mismatches in both directions: shortages and excess inventory. Underestimating demand risks lost revenue and share; overestimating it has forced price reductions, inventory write-downs, cancellation penalties and impairments, and the risk is amplified by non-cancellable, non-returnable orders and by growing purchase obligations and prepaids. Because products are part of complex data center buildouts, a shortage of any single component can have a broad revenue impact. The company also depends on third-party foundries, assemblers and testers, reducing its control over quantity, quality, yields and delivery schedules, and product defects have caused significant remediation expense.

Customer and partner concentration. For fiscal 2026, one direct customer represented 22% of total revenue and another 14%, both primarily Compute & Networking. In the second quarter of fiscal 2027 one direct customer was 16% of revenue; for the first half, three direct customers were 16%, 15% and 13%. Five direct customers accounted for 22%, 14%, 13%, 11% and 10% of accounts receivable at July 26, 2026. Some indirect customers individually represent 10% or more of revenue, and one AI research and deployment company is estimated to have contributed a meaningful amount of revenue by purchasing cloud services from NVIDIA's customers. Most sales are on purchase orders that customers can cancel, change or delay with little notice and no penalty.

Investments. NVIDIA has invested heavily in its ecosystem, including illiquid, non-marketable early-stage private companies and publicly-held equity securities. These may generate large realized and unrealized gains or losses, the portfolio carries industry-sector concentration risk, and losses up to the full value of investments are possible. The fiscal 2026 10-K disclosed that the company was finalizing, with no assurance of completion, an investment and partnership agreement with OpenAI. The Groq intellectual property license required significant nonrefundable payments; incorporating the licensed technology into NVIDIA architectures requires substantial engineering effort, may not occur on expected timelines, and the associated costs may not be recovered.

Competition and technology change. Markets are intensely competitive with rapid technological change. Competitors may be cheaper or better-featured, some operate their own fabs and have longer histories and larger resources, and several large customers have in-house design capability and can substitute their own solutions. High-quality open-source foundation models broaden access to advanced AI; management believes this promotes use of its products, but if such models are deployed primarily on competitors' platforms, or if regulation limits NVIDIA's ability to support third-party applications and models (including Chinese-origin models such as DeepSeek, Qwen or Kimi), demand could fall.

Resource availability. The buildout of NVIDIA AI infrastructure by customers and partners depends on the availability of data centers, land, power, shell and capital. Expanding energy capacity is a complex multi-year process with regulatory, technical and construction challenges, and less-capitalized companies may struggle to finance large-scale projects, limitations that could delay deployments or reduce the scale of AI adoption.

Other named risks. Adverse economic conditions and macro factors including tariffs, inflation, interest rates and capital-market volatility; international operations, including approximately 6,200 employees in Israel supporting networking research and development, operations and sales; product, system-security and data-protection incidents and cyber-attacks; business disruptions; climate change; integration and realization risk on acquisitions; executive and key-employee retention; business-process and information-system changes, including a phased ERP upgrade under way; operating-result volatility versus analyst expectations; broad regulatory scrutiny, with information requests from competition regulators in the EU, U.S., UK, China and South Korea and a French Competition Authority inquiry into whether gaming and data center GPUs are separate product categories; the EU AI Act and new U.S. state AI laws effective January 1, 2026; corporate-sustainability scrutiny; responsible-AI issues; IP protection; data privacy; tax liabilities and rate changes; litigation, including the long-running securities class action described below; and anti-takeover provisions, including an agreement with Microsoft for the Xbox under which Microsoft may hold first and last rights of refusal if an offer is made to purchase at least 30% of NVIDIA's outstanding common stock.

Litigation. In the securities class action In re NVIDIA Corporation Securities Litigation (N.D. Cal., 4:18-cv-07669), which alleges materially false or misleading statements about channel inventory and the impact of cryptocurrency mining on GPU demand between May 2017 and November 2018, the district court granted the plaintiffs' motion for class certification on March 25, 2026, certifying a class of purchasers of NVIDIA common stock between August 10, 2017 and November 15, 2018. Related derivative actions in the Northern District of California, the District of Delaware and the Delaware Court of Chancery remain stayed or administratively closed pending resolution of that action. As of July 26, 2026 no contingent liabilities were accrued for these matters, as management believes losses are reasonably possible but not probable and cannot be reasonably estimated.

Management's discussion, fiscal year 2026

From Item 7 of the fiscal 2026 Form 10-K, accession 0001045810-26-000021, for the year ended January 25, 2026.

Revenue was $215.9 billion, up 65% year over year, driven by data center compute and networking platforms; Blackwell architectures represented the majority of Data Center revenue. By reportable segment, Compute & Networking revenue was $193.5 billion, up 67%, and Graphics revenue was $22.5 billion, up 57%. Within Compute & Networking, Data Center computing revenue grew 59% on Blackwell demand and Data Center networking grew 142% on the introduction and ramp of the NVLink compute fabric for GB200 and GB300 systems plus Ethernet and InfiniBand growth. On the end-market basis then in use, Data Center revenue was up 68%, Gaming up 41%, Professional Visualization up 70% (helped by the DGX Spark launch) and Automotive up 39%. Segment operating income was $130.1 billion for Compute & Networking (up 57%) and $9.2 billion for Graphics (up 80%).

Gross margin fell to 71.1% from 75.0% as the business model shifted from selling Hopper HGX systems to full-scale Blackwell datacenter solutions, and because of the $4.5 billion H20 charge in the first quarter. Inventory and excess purchase-obligation provisions totalled $7.2 billion (versus $3.7 billion in fiscal 2025), including the H20 charge, against provision releases of $1.5 billion (versus $689 million); the net effect on gross margin was an unfavorable 2.6% in fiscal 2026 and 2.3% in fiscal 2025.

Operating expenses were $23.1 billion, up 41%, research and development $18.5 billion (up 43%) and sales, general and administrative $4.6 billion (up 31%). The R&D increase reflected a 29% rise in compensation and benefits on employee growth and a 79% rise in compute and infrastructure spending. Interest income rose on a larger cash and debt-securities balance. Other income, net benefited from unrealized gains on non-marketable and publicly-held equity securities, including gains on the previously announced investment in Intel common stock.

Income tax expense was $21.4 billion, an effective rate of 15.1% versus 13.3% in fiscal 2025; the increase came from a smaller relative contribution of stock-based compensation, FDDEI and federal research-credit benefits against much higher pre-tax income. The One Big Beautiful Bill Act, enacted July 2025, was reflected in fiscal 2026 results. A $711 million valuation allowance on certain state deferred tax assets was released at year end. Unrecognized tax benefits were $4.0 billion including $374 million of interest and penalties; fiscal years 2023 and 2024 are under IRS examination.

Operating cash flow was $102.7 billion (from $64.1 billion), investing used $52.2 billion (from $20.4 billion, driven by equity-investment purchases and the Groq non-exclusive license), and financing used $48.5 billion (from $42.4 billion, mainly higher buybacks). Capital expenditures were $6.1 billion versus $3.4 billion, with management expecting a further increase in fiscal 2027. Cash, cash equivalents and marketable securities were $62.6 billion at year end; senior notes outstanding were $8.5 billion; the commercial paper program was raised to $25.0 billion in January 2026 with nothing outstanding.

Groq. In December 2025 NVIDIA entered into a non-exclusive license agreement with Groq, Inc. for its language processing unit technology and hired certain Groq employees; no customer contracts, existing products or equity interests were purchased. On the arrangement NVIDIA recorded $14.4 billion of goodwill, attributed primarily to the workforce and the future development of the licensed technology, and a $2.5 billion developed-technology intangible asset valued using a cost-to-recreate methodology with a five-year useful life, both in the Compute & Networking reporting unit. Total consideration was $13.0 billion paid at closing plus $4 billion, inclusive of imputed interest, payable within one year and carried in accrued and other current liabilities; the goodwill is tax deductible. Goodwill on the balance sheet rose to $20.8 billion from $5.2 billion; Groq accounts for most of that $15.6 billion increase for the year, all of which was allocated to Compute & Networking.

On capital return: the Board authorized an additional $60.0 billion of repurchases on August 26, 2025; NVIDIA repurchased 282 million shares valued at $40.4 billion in fiscal 2026, $40.1 billion of cash paid for repurchases on the statement of cash flows, and had $58.5 billion of authorization remaining at year end, plus 8 million shares for $1.5 billion between January 26 and February 20, 2026. Dividends paid were $974 million.

Ecosystem investment was a stated fiscal 2026 priority: $17.5 billion invested in private companies and infrastructure funds, primarily early-stage startups including AI model makers that buy NVIDIA products directly or through cloud providers, plus investments in public equities and $3.5 billion of land, power and shell guarantees provided to early-stage companies over multi-year periods.

Management flagged as forward challenges the availability of data centers, energy and capital for customers' buildouts; the complexity of one-year-cadence product transitions (Blackwell Ultra including GB300 began shipping in the second quarter of fiscal 2026; Rubin was next); the rise of high-quality open-source models; supply-chain concentration in Asia and expansion into the U.S. and Latin America; and macroeconomic factors including tariffs.

Current quarter, second quarter of fiscal 2027

From the Form 10-Q for the quarter ended July 26, 2026, accession 0001045810-26-000075.

Results. Revenue was $96.2 billion, up 106% year over year and up 18% sequentially. Gross margin was 75.0%, up 2.6 points from a year ago on improved mix from Blackwell Ultra and roughly flat sequentially. Operating expenses were $8.4 billion, up 55% year over year and 10% sequentially on higher compute infrastructure and compensation costs. Operating income was $63.7 billion (up 124%), net income $59.7 billion (up 126%) and diluted EPS $2.46 (up 128%). For the first half, revenue was $177.8 billion (up 96%), operating income $117.3 billion and net income $118.0 billion.

Other income, net was $7.8 billion for the quarter and $24.1 billion for the half, dominated by net gains from equity securities of $7.8 billion and $23.7 billion respectively, the single largest swing factor in reported earnings. Interest income was $496 million and interest expense $227 million for the quarter. The effective tax rate was 16.5% for both the quarter and the half, up from 15.3% and 14.9%.

Mix. Under the fiscal 2027 market-platform presentation, Data Center revenue was $89.0 billion, up 117% year over year and 18% sequentially, driven by the Blackwell Ultra ramp. Within it, Hyperscale was $48.7 billion (up 102%, up 13% sequentially) and AI Clouds, Industrial & Enterprise was $40.3 billion (up 138%, up 25% sequentially) on demand from AI natives, enterprises and sovereign customers plus hyperscalers using AI clouds. Edge Computing was $7.2 billion, up 27% and up 13% sequentially, on strong Blackwell workstation sales partly offset by slower consumer PC sales tempered by elevated memory and system prices. Gaming, Professional Visualization and Automotive are no longer disclosed as separate revenue lines under this presentation; the Graphics reportable segment is the closest remaining proxy for the gaming and visualization businesses. During the quarter one company was reclassified from AI Clouds, Industrial & Enterprise to Hyperscale and prior periods were recast. Data Center Hopper shipments to China were less than 1% of Data Center revenue. By reportable segment, Compute & Networking revenue was $88.3 billion (up 114%) with operating income of $62.7 billion, and Graphics revenue was $7.9 billion (up 46%) with operating income of $3.9 billion. Revenue from customers headquartered outside the United States was 38% of the total for the quarter; Taiwan-headquartered customers accounted for $27.0 billion and China including Hong Kong $7.9 billion.

Balance sheet and cash flow. Total assets grew to $320.3 billion from $206.8 billion at fiscal year end. Accounts receivable rose to $63.1 billion from $38.5 billion, days sales outstanding of 60, up from 45 sequentially, because of extended payment terms on large multi-quarter agreements with investment-grade customers. Inventories rose to $31.6 billion from $21.4 billion ahead of the Vera Rubin introduction, with raw materials nearly tripling to $11.3 billion. Non-marketable securities rose to $51.2 billion from $22.3 billion and marketable equity securities to $42.8 billion from $12.9 billion. Cash, cash equivalents and marketable debt securities were $56.6 billion. Operating cash flow was $24.1 billion for the quarter and $74.4 billion for the half. Less $2.7 billion of purchases of property, equipment and intangible assets, that leaves free cash flow of $21.4 billion for the quarter; NVIDIA's own measure deducts a further $59 million of principal payments on such assets financed by third parties and is therefore $21.3 billion. Purchases of equity securities were $42.4 billion in the first half against $7.2 billion of proceeds from sales.

Financing and capital return. In June 2026 NVIDIA issued $25.0 billion of senior unsecured notes across seven tranches for general corporate purposes, lifting the net carrying amount of debt to $33.4 billion from $8.5 billion, of which $32.4 billion is long term. The commercial paper program has $25.0 billion of capacity with nothing outstanding. The Board approved an additional $80.0 billion of repurchase authorization on May 18, 2026 and raised the quarterly dividend from $0.01 to $0.25 per share the same day. NVIDIA repurchased 94 million shares in the quarter and 203 million in the half; the value of shares repurchased, as the company reports it, was $19.7 billion and $39.8 billion respectively, while cash paid for repurchases on the statement of cash flows was $39.0 billion for the half. The company paid $6.0 billion of dividends in the quarter and had $99.3 billion of repurchase authorization remaining at quarter end. Shares outstanding fell to 24,147 million from 24,304 million at fiscal year end.

Commitments, the central development of the quarter. Supply and capacity commitments more than doubled to $279 billion from $119 billion a quarter earlier, primarily memory procurement and manufacturing facilities for data center infrastructure systems across current and future architectures. Total future commitments were $366 billion, comprising supply and capacity $279 billion, cloud service agreements $29 billion, data center leases not yet commenced $25 billion, equity investments $25 billion and capital expenditures $8 billion; $120 billion falls in the remainder of fiscal 2027. Separately disclosed additional commitments total $56 billion: $36 billion of AI cloud agreements and $20 billion of data center leases not commenced that NVIDIA expects to reassign to third parties. Equity investments held were $99 billion with $25 billion of further commitments.

New AI cloud business model. In the second quarter NVIDIA introduced an arrangement with select AI cloud partners under which the partners procure NVIDIA data center infrastructure and NVIDIA commits to cloud service agreements, typically six years, $36 billion in aggregate at quarter end, which the AI clouds may unilaterally stop providing to NVIDIA and instead sell to third-party customers at better rates. NVIDIA's commitment declines as capacity is taken up by third parties or used for its own research and development, and if certain criteria are met NVIDIA participates in revenue share from the AI clouds' third-party customers. Management's rationale is that AI clouds and model makers have strong demand but lack long-term infrastructure contracts and investment-grade financing capacity.

Other items. Goodwill rose $293 million in the first half from an acquisition allocated to Compute & Networking; on May 15, 2026 NVIDIA acquired a company (unnamed in the filing) and issued 926,447 shares to key employees valued at approximately $209 million. Payments of $2.9 billion relating to the Groq, Inc. non-exclusive license appear in first-half financing activities, with $986 million of accrued purchase consideration remaining. New derivative positions appeared this quarter: public company warrants with a $4.8 billion notional and $824 million fair value, and a $1.0 billion equity forward contract. Land, power and shell guarantees for AI clouds carried a $3.5 billion maximum gross exposure at quarter end, partly mitigated by $712 million of partner escrow. Fiscal 2027 is a 53-week year, with a 14-week fourth quarter. A phased ERP upgrade is under way; disclosure controls were concluded effective with no material change in internal control over financial reporting during the quarter.

Guidance. For the third quarter of fiscal 2027 NVIDIA guided to revenue of $108.0 billion plus or minus 2%, assuming no Data Center compute revenue from China; GAAP and non-GAAP gross margins of 74.0% plus or minus 50 basis points; and GAAP and non-GAAP operating expenses of approximately $9.2 billion and $9.0 billion. For the full fiscal year, GAAP and non-GAAP tax rates are expected between 16.0% and 18.0%, excluding discrete items and material changes in the tax environment. (Form 8-K, accession 0001045810-26-000073.)

Subsequent events

Disclosed in the Form 10-Q for the quarter ended July 26, 2026, accession 0001045810-26-000075, covering the period after July 26, 2026.

SB Energy Corp. guarantees, August 2026. NVIDIA entered into guarantees capped at $105 billion in aggregate providing credit support on a land, power and shell buildout with affiliates of SB Energy Corp. on behalf of a customer, an affiliate of OpenAI Group PBC, relating to leases for approximately 4.25 gigawatts of IT load at SB Energy's PORTS Technology Campus in Pike County, Ohio. Each guarantee generally becomes effective on commencement of the applicable lease, with guarantee amounts stepping up as each of nine phases of data center construction is completed, the first expected in fiscal year 2029. Payment obligations trigger on certain tenant defaults and are expected to decline over each phase's 20-year lease term; they are limited to defined portions of lease and power payments rather than the full cost of the site or all of the tenant's obligations, and terminate on certain events including OpenAI achieving a satisfactory credit rating or completion of each lease term. In exchange, the site will exclusively host NVIDIA AI infrastructure, subject to limited exceptions. NVIDIA also holds an option, exercisable at its sole discretion, to provide additional credit support in phases for approximately 3.8 further gigawatts as the site scales. Including these guarantees, NVIDIA's maximum gross guarantee exposure is $108.5 billion, $105.0 billion for SB Energy plus $3.5 billion of existing land, power and shell guarantees for AI clouds.

Financing-platform memoranda of understanding, August 2026. NVIDIA entered into memoranda of understanding with several large capital providers to establish independent financing platforms designed to mobilize more than $500 billion of third-party capital over time for AI infrastructure deployment. The funding structures would be independently underwritten and provided by the capital providers; at its option NVIDIA may provide limited residual-value support for a portion of specific projects. The filing states these and other preliminary arrangements may not lead to definitive agreements.

Vera Rubin production shipments. NVIDIA's next-generation Data Center architecture, Vera Rubin, began production shipments in the third quarter of fiscal 2027, after the quarter closed. The company expects to ship both Blackwell and Rubin systems going forward and states it is currently experiencing certain supply constraints.

No acquisitions, divestitures or litigation resolutions were disclosed as occurring after July 26, 2026 in this Quarterly Report.