NVIDIA Corporation (NVDA): Business Model Canvas [Dec-2025 Updated]

You're looking to map out exactly how the world's most valuable chip designer is printing money, and honestly, the answer is almost entirely in the Data Center, which pulled in a staggering $115.19 billion in FY2025 revenue. As an analyst who's seen a few cycles, I can tell you this model isn't just about selling GPUs; it's about owning the entire accelerated computing stack, cemented by the proprietary CUDA software that creates a massive competitive moat. Below, we break down the full nine-block framework-from key partnerships with hyperscalers to the high cost of advanced fabrication-so you can see the precise engine driving this market leader. Dive in to see the full picture.

NVIDIA Corporation (NVDA) - Canvas Business Model: Key Partnerships

You're mapping out the ecosystem that keeps NVIDIA Corporation at the center of the AI infrastructure buildout, so let's look at the hard numbers defining these critical relationships.

The fabless model hinges on manufacturing partners, where NVIDIA Corporation maintains a clear preference for technological leadership, though diversification efforts are visible.

The relationship with hyperscalers is the primary revenue driver, as they consume massive amounts of NVIDIA Corporation's compute for their cloud offerings.

• Hyperscalers like Microsoft Azure, AWS, and Google Cloud reinforce NVIDIA Corporation's software dominance via the CUDA ecosystem.
• Microsoft's Intelligent Cloud group reported $30.9 billion in sales for Q3 2025, with an annual revenue run rate of $123 billion.
• AWS reported $33 billion in sales for Q3 2025, translating to a $132 billion annual revenue run rate.
• Google Cloud demonstrated strong growth, with a 32% year-over-year growth rate in Q2 2025.
• Collectively, these three cloud giants planned to spend about $240 billion in 2025 on data centers and AI capabilities.

NVIDIA Corporation's Data Center segment revenue for the full Fiscal Year 2025 reached $115.19 billion, underscoring the importance of these cloud relationships. Still, the demand is so high that Blackwell GPUs are reported as sold out for 12 months.

Strategic investments solidify future compute demand and ecosystem lock-in, particularly with leading AI model developers.

The investment in Anthropic is part of a larger deal where Anthropic committed to purchasing $30 billion in compute capacity from Microsoft Azure running on NVIDIA AI systems. That's a serious commitment to the stack.

In the automotive sector, NVIDIA Corporation is embedding its compute platforms deep into future vehicle architectures, securing design wins that promise revenue streams years out.

• NVIDIA DRIVE Thor, built on the Blackwell architecture, delivers 1,000 teraflops of accelerated compute performance for inference tasks.
• Automotive safety pioneers like Mercedes-Benz are adopting the NVIDIA DRIVE platform and Drive OS.
• Toyota is building its next-generation vehicles on the predecessor, DRIVE Orin, which delivers 254 trillion operations per second.
• Continental and Aurora plan to mass-manufacture driverless trucks featuring DRIVE Thor running DriveOS in 2027.
• Hyundai Motor Group is actively collaborating with NVIDIA Corporation on AI innovation for manufacturing and mobility.

Finally, the push into telecommunications involves a significant capital allocation to accelerate the software layer for next-generation networks.

NVIDIA Corporation announced a strategic partnership with Nokia, which includes a $1 billion equity investment made at a subscription price of $6.01 per share, securing a 2.9% stake in Nokia. This is aimed at enabling AI-native 5G-Advanced and 6G networks, with new equipment expected to contribute to revenue starting in 2027. Finance: draft 13-week cash view by Friday.

NVIDIA Corporation (NVDA) - Canvas Business Model: Key Activities

You're managing a portfolio heavily weighted toward high-growth tech, so understanding the core engine-the Key Activities-that drives NVIDIA Corporation's value is crucial. It's not just about shipping silicon; it's about the continuous, high-intensity work that builds and defends their competitive moat. Here's a breakdown of what NVIDIA is actively doing to maintain its lead as of late 2025.

Designing next-generation GPU architectures (Blackwell, Rubin)

The cadence of architectural design is a primary activity, setting the pace for the entire industry. NVIDIA is executing on a tight, one-year cycle for major architectural releases. While the Blackwell generation is shipping, the focus is already heavily on the successor.

• Designing the Blackwell Ultra GPU, an incremental upgrade to the existing Blackwell architecture, slated for arrival in the second half of 2025.
• Designing the Rubin architecture, named after Vera Rubin, which is scheduled for mass production in Q4-2025, targeting early 2026 availability.
• The Rubin R100 GPU is expected to leverage TSMC's advanced 3 nm process, a step up from the Blackwell B100's TSMC-N4P node.

Developing and expanding the proprietary CUDA software ecosystem

This activity is arguably the most critical, as the software locks in customers regardless of minor hardware shifts. CUDA is the proprietary parallel computing platform and API that turns raw GPU power into accessible performance for AI and High-Performance Computing (HPC). It includes compilers, libraries like cuDNN and TensorRT, and runtime kernels.

The platform's maturity means that for many practitioners, dropping down to write custom CUDA kernels is becoming less necessary, with NVIDIA claiming custom kernels are only needed about 10% of the time as of GTC 2025 sessions. Still, the entire stack is built upon this foundation.

Manufacturing and managing a complex global supply chain

As a fabless designer, NVIDIA's key activity involves intense management of its manufacturing partners, primarily TSMC, and the complex logistics of getting high-demand components like HBM memory and advanced packaging (like CoWoS-L) to customers. The sheer scale of demand requires constant coordination.

Consider the output from just one segment: the Data Center segment generated $52 billion in revenue in Fiscal Year 2025, representing 59% of the total revenue for that period. Managing the supply chain to meet that level of demand is a massive undertaking.

Deep research and development (R&D)

This is the fuel for future product cycles, ensuring the next generation of chips is ready on schedule. NVIDIA's commitment here is substantial, translating directly into their ability to design architectures like Rubin and its successor, Feynman.

Here's the quick math on the investment for the fiscal year ending January 26, 2025:

What this estimate hides is that a significant portion of that R&D spend is dedicated to co-designing the CPU (like Vera) and GPU (like Rubin) together, which is a new level of system integration activity.

Cultivating the developer community

This activity reinforces the software moat by ensuring a large, proficient user base is ready for new hardware. The network effect is powerful: companies are more likely to adopt NVIDIA hardware because they can hire engineers already fluent in CUDA.

The scale of this community is significant as of early-to-mid 2025:

• The CUDA developer base is reported to be more than 4 million global developers as of March 2025.
• This community proficiency spans AI researchers, data scientists, and developers worldwide.
• Universities globally teach deep learning using NVIDIA GPUs and the CUDA toolkit, embedding the skill set early.

If onboarding takes 14+ days, churn risk rises, so keeping the developer experience smooth is defintely a top priority.

NVIDIA Corporation (NVDA) - Canvas Business Model: Key Resources

You're mapping out the core assets that make NVIDIA Corporation an infrastructure giant in late 2025. These aren't just products; they are the deep, hard-to-replicate foundations of their market dominance. Let's break down the numbers and the proprietary assets that anchor their position.

Proprietary CUDA Software Platform (a significant competitive moat)

The CUDA (Compute Unified Device Architecture) platform is arguably the single most important resource. It's a proprietary software platform and programming model developed over nearly two decades, which lets developers easily use NVIDIA GPUs for AI and other parallel computing tasks. This has created a massive developer base, estimated in the millions strong, and a vast library of optimized software. This ecosystem is the core of the so-called "CUDA moat," making it prohibitively difficult and expensive for customers to switch to competing hardware lacking native CUDA support. Honestly, almost every deep learning framework today relies on CUDA/GPU computing for acceleration in training and inference.

Advanced Intellectual Property (IP) in Parallel Computing

NVIDIA Corporation has secured patents covering the fundamental principles that make multi-GPU parallel computing viable. These patents control how computational tasks are allocated and managed across multiple GPUs, creating efficiency that competitors struggle to replicate without potential infringement. This IP portfolio, covering semiconductor circuit designs, data interconnects, and interfaces, extends protection beyond specific products to the core methodologies driving accelerated computing.

Financial Strength and Liquidity

The sheer financial backing allows NVIDIA Corporation to invest aggressively in R&D, secure supply chain capacity, and weather any near-term market fluctuations. You need to see the balance sheet strength here. Here's the quick math on their highly liquid assets:

What this estimate hides is the massive capital expenditure required to secure advanced packaging capacity, like contracting for over 70% of TSMC's CoWoS capacity for 2025.

Jensen Huang and the World-Class Engineering Talent

The leadership under CEO Jensen Huang is a critical, intangible resource. Huang, who has been CEO since the company's founding in 1993, maintains a relentless focus, reportedly working seven days a week in a constant state of vigilance, often describing the company as being "30 days from going out of business" for over three decades. This engineering-first culture underpins the company's ability to deliver successive architectural leaps. This talent base supports a dominant market position, with NVIDIA controlling an estimated 80% of the market for GPUs used in training and deploying AI models as of 2025.

High-Performance GPU Architectures (Hopper, Blackwell)

The pipeline of next-generation silicon is a tangible resource that locks in future revenue. NVIDIA Corporation's ability to deliver sequential performance gains across its architecture roadmap is unmatched. The current and next-gen platforms are:

• Hopper Architecture (H100, H200): Remained in high demand through early 2025.
• Blackwell Architecture (B200, GB200): Began shipping in Q4 Fiscal Year 2025, with early orders exceeding Hopper's peak volumes.
• Blackwell Ultra: Expected later in 2025, projected to boost AI factory output by up to 50x over Hopper.
• Rubin Platform: The next-generation architecture scheduled for release in 2026.
• Feynman GPU: Announced for a 2028 release.

Blackwell delivers performance improvements, such as 30x faster inference with 25x lower cost of ownership compared to Hopper. By late 2025, the Blackwell platform was expected to account for over 80% of NVIDIA's high-end GPU shipments.

NVIDIA Corporation (NVDA) - Canvas Business Model: Value Propositions

You're looking at the core reasons why customers are lining up for NVIDIA Corporation's gear, especially as we close out 2025. It really boils down to raw, demonstrable performance and a platform that covers the entire AI lifecycle, from the cloud to the car.

Unmatched compute performance for AI training and inference

The performance gains with the Blackwell architecture are not incremental; they are step-changes that redefine what's possible in large model deployment. For instance, the Blackwell series is showing up in MLPerf benchmarks as potentially outperforming the prior Hopper class by a factor of four on the biggest LLM workloads, like Llama 2 70B, driven by features like the second-generation Transformer Engine and FP4 Tensor Cores.

When you look at the hard numbers from the MLPerf v4.1 Training benchmarks, NVIDIA is reporting up to a 2.2x gain for Blackwell over Hopper. Honestly, the math on training time is staggering: achieving the same performance on the GPT-3 175B benchmark required only 64 Blackwell GPUs compared to 256 Hopper GPUs.

For inference, which is where most AI engines run in production, the performance advantage is also clear. The H200 delivered up to 27% more generative AI inference performance over previous benchmark tests. Furthermore, Blackwell systems are showing 10x throughput per megawatt compared to the previous generation in the SemiAnalysis InferenceMAX benchmarks.

The market demand reflects this: CEO Jensen Huang confirmed in the Q3 FY26 earnings call that Blackwell sales are 'off the charts,' and cloud GPUs are sold out. Management has stated they currently have visibility to $0.5 trillion in Blackwell and Rubin revenue from the start of 2025 through the end of calendar year 2026.

Here's a quick comparison of the training performance leap:

Full-stack accelerated computing platform (hardware, software, systems)

NVIDIA isn't just selling chips; they are selling the entire factory floor for AI. This full-stack approach integrates the chip architecture, the node and rack architecture (like the GB200 NVL72), and the necessary software layers. This is why the Data Center segment hit a record $51.2 billion in Q3 FY26 revenue, which is up 66% year-over-year. The total company revenue for that same quarter was $57.0 billion.

The platform's strength is evident across the stack:

• The networking business is now reported as the largest in the world.
• The non-GAAP gross margin for Q3 FY26 held strong at 73.6%.
• Systems are built with high-speed NVLink fabrics, HBM3e memory, and are designed for liquid cooling, which is table stakes for dense AI racks.

Lower Total Cost of Ownership (TCO) for AI infrastructure

While NVIDIA's performance is industry-leading, the competitive landscape means large hyperscalers are driving down the effective cost. For major customers, competitive pressure has reportedly led to concessions that reduce the Total Cost of Ownership (TCO) of their computing clusters by approximately 30%. This is seen when comparing the all-in cost per chip at rack scale for a GB200 or GB300 system versus alternatives like Google's TPUv7, which is cited as providing a more cost-effective alternative for certain performance levels.

Industry-leading AI-driven graphics and rendering for gamers

The gaming side still shows solid growth, even as Data Center dominates the narrative. For Q3 FY26, Gaming revenue came in at $4.3 billion, representing a 30% increase year-over-year. This is supported by the launch of technologies like NVIDIA DLSS 4 with Multi Frame Generation and NVIDIA Reflex.

End-to-end platforms for autonomous vehicles and robotics

NVIDIA Corporation's DRIVE platform provides a full 'cloud-to-car' stack, which is seeing significant commercial traction. The Automotive & Robotics segment reported $567 million in revenue for Q1 FY 2026, a 72% year-over-year jump. For the full fiscal year 2025, that segment generated $1.7 billion.

The company is targeting roughly $5 billion in automotive revenue for fiscal year 2026. This is being driven by major design wins:

• Toyota is building next-gen vehicles on DRIVE AGX Orin with DriveOS.
• Magna is deploying DRIVE Thor SoCs for L2-L4 ADAS.
• Continental plans to mass-produce NVIDIA-powered L4 self-driving trucks with Aurora.
• Partnerships include Volvo Cars, Mercedes-Benz, Lucid, BYD, and NIO using the DRIVE AGX platform.

Finance: review the Q4 FY26 automotive revenue forecast against the $5 billion FY2026 target by next Tuesday.

NVIDIA Corporation (NVDA) - Canvas Business Model: Customer Relationships

You're looking at how NVIDIA Corporation maintains its grip on the AI infrastructure market, and it all comes down to how they manage relationships across vastly different customer types. It's not a one-size-fits-all approach; it's highly segmented.

Dedicated, high-touch sales and engineering support for hyperscalers

For the largest cloud providers-the hyperscalers-the relationship is intensely collaborative. NVIDIA Corporation is enabling a scale and velocity in deploying one-and-a-half ton AI supercomputers the world has never seen before, according to their 2025 Annual Review. The Blackwell platform is powering AI infrastructure across these hyperscalers, enterprises, and sovereign clouds. This high-touch engagement is critical, as evidenced by the fact that NVIDIA's Data Centre revenue growth was reported at 17% in the second quarter of fiscal year 2025. This segment is about ensuring the entire stack, from the hardware to the networking like Spectrum-XGS Ethernet, is perfectly integrated for their massive AI factory buildouts.

Deep co-development with key enterprise and sovereign AI customers

The move from AI pilots to scaled deployment means deep integration with enterprise and government clients. NVIDIA Corporation is partnering with government and research institutions to build seven new supercomputers, with some systems utilizing more than 100,000 NVIDIA GPUs to support open science and national laboratories. This level of co-design extends to the enterprise side; for instance, Dell announced that it already had 2,000 customers within a year of announcing its NVIDIA AI stack. Furthermore, major enterprise SAS companies like ServiceNow, SAP, and Salesforce are adopting NVIDIA Inference Microservices (NIMs), which essentially require NVIDIA hardware to run effectively. Sovereign AI strategies are also a focus, with NVIDIA announcing GPU deployments with 12 global telcos to fuel these national infrastructure projects.

Large-scale, community-driven support for the developer ecosystem

The foundation of NVIDIA Corporation's long-term moat is its developer community, which is supported through extensive, scalable resources. The NVIDIA Developer Program provides free access to advanced tools and a dedicated community. This includes access to GPU-optimized software via the NGC Catalog and support for startups through the NVIDIA Inception accelerator, which provides access to the Deep Learning Institute (DLI). To democratize access, NVIDIA introduced Project Digits at CES 2025, a device priced at $3,000 that offers 1 PFLOPS of FP4 performance, tailored for developers to run large language models locally.

The key components of this developer engagement include:

• Access to the NGC Catalog for software and models.
• Support for startups via NVIDIA Inception.
• Training through the Deep Learning Institute (DLI).
• New hardware like Project Digits for local AI development.

Standardized, transactional relationship with retail consumers

For the consumer segment, primarily focused on gaming and creative workloads with GeForce GPUs, the relationship is largely transactional, driven by product availability and performance benchmarks. As of the first quarter of 2025, NVIDIA Corporation held a 92% share of the discrete desktop and laptop GPU market. This segment relies on the established brand and ecosystem, like DLSS 4 updates, but the direct, high-touch engineering support seen with hyperscalers is absent here.

GTC conference as the defintely central engagement point

The GPU Technology Conference (GTC) serves as the single most important event for aligning the entire ecosystem-from the largest customers to individual developers. It is the epicenter for showcasing AI opportunity, and every company wishing to play a role is in attendance. The March 2025 event solidified this role as the 'Super Bowl of AI.'

Here are the key engagement metrics from GTC 2025:

The conference is where NVIDIA Corporation unveils its next-generation platforms, such as Blackwell Ultra, which delivers 50x more AI factory output compared to the Hopper platform for large-scale reasoning workloads. Finance: draft 13-week cash view by Friday.

NVIDIA Corporation (NVDA) - Canvas Business Model: Channels

You're looking at how NVIDIA Corporation gets its massive revenue-which hit $130.5 billion in fiscal year 2025-into the hands of its customers. The channels are highly segmented, reflecting the dual nature of the business: powering the world's largest AI infrastructure and serving the consumer gaming market.

The Data Center segment is the engine, accounting for 88.27% of total revenue, or $115.19 billion in FY2025. This revenue flows through several critical, high-volume channels.

Direct sales to major Data Center customers and governments

This channel involves direct engagement for the highest-tier, largest-scale AI deployments. The concentration here is notable; in the most recent quarter, more than half of Data Center revenue came from just three unnamed clients. Here's the quick math on that concentration:

This direct channel also includes significant government contracts, such as the announced partnership for the $500 billion Stargate Project.

Cloud Service Providers (CSPs) offering GPU instances (e.g., DGX Cloud)

Cloud Service Providers are fundamental volume purchasers for the Data Center segment. NVIDIA revealed that major CSPs, including AWS, CoreWeave, Google Cloud Platform (GCP), Microsoft Azure, and Oracle Cloud Infrastructure (OCI), are deploying NVIDIA GB200 systems globally. The networking component supporting these massive clusters is also a key channel indicator; the combined networking segment delivered $8.19 billion in revenue in the third quarter of fiscal 2025, growing 162% year-over-year.

Original Equipment Manufacturers (OEMs) like Dell and HPE

OEMs take NVIDIA components, integrate them into servers and systems, and resell them. While the search results don't break out OEM revenue specifically, the 'OEM And Other' segment represented 0.30% of total FY2025 revenue, amounting to $389.00 million. This channel is crucial for distributing standard server platforms containing NVIDIA accelerators.

Global retail and e-commerce networks for Gaming GPUs

The Gaming segment generated $11.35 billion in FY2025, representing 8.7% of the total. This consumer-facing channel is dominated by NVIDIA's brand strength. In the first quarter of 2025, NVIDIA captured a staggering 92% share in the add-in board (AIB) GPU market, and generally holds over 80% market share in discrete GPUs used for gaming.

The launch of the GeForce RTX 50 Series drove this performance, with Blackwell architecture sales contributing billions of dollars in its first quarter, with one report citing $11 billion of Blackwell revenue delivered in the fourth quarter of fiscal 2025 alone.

Value-Added Resellers (VARs) for enterprise AI solutions

VARs are essential for deploying specialized, often smaller-scale or customized, enterprise AI solutions where direct CSP or OEM routes are less efficient. This channel helps distribute solutions built around platforms like the NVIDIA DGX Cloud and NIM microservices to a wider enterprise base.

The distribution of NVIDIA's massive Data Center revenue relies on a mix of direct hyperscaler deals and channel partners:

• Cloud Service Providers (CSPs) are the primary volume buyers for AI infrastructure.
• Direct sales capture the largest, most strategic national and government AI buildouts.
• OEMs and VARs handle the broader enterprise and system integrator market distribution.
• The Gaming channel maintains near-total dominance in the discrete GPU retail space.

Finance: draft 13-week cash view by Friday.

NVIDIA Corporation (NVDA) - Canvas Business Model: Customer Segments

You're looking at the core buyers driving NVIDIA Corporation's massive scale as of late 2025. Honestly, the customer base is heavily skewed, which is a key strategic point to watch.

Hyperscale Cloud Providers represent the undisputed largest segment. This group, which includes giants like AWS, Google Cloud Platform (GCP), Microsoft Azure, and Oracle Cloud Infrastructure (OCI), is responsible for the bulk of the company's success. In fiscal year 2025, the Data Center segment, which primarily serves these providers, generated $115.19 billion in revenue. That figure alone represents a staggering 88.27% of NVIDIA Corporation's total revenue for the year. These providers are deploying NVIDIA GB200 systems globally to meet the surging demand for AI training and inference workloads.

The next tier involves AI/ML Startups and Large Enterprises, including those in finance and healthcare. While often bundled into the Data Center reporting, this group is actively building sovereign AI capabilities and deploying AI infrastructure beyond the major cloud players. The growth here is fueled by the need for generative AI, moving from training to reasoning workloads.

For PC Gamers and Enthusiasts, this remains a foundational, though now smaller, customer group. Gaming and AI PC revenue was $11.35 billion in fiscal year 2025. That's about 8.7% of the total pie. They are the initial market for new consumer GPUs, like the recently announced GeForce RTX 50 Series cards.

The specialized segments round out the picture. Automotive OEMs and Tier 1 suppliers are buying in for AI-driven vehicle technologies. This segment brought in $1.69 billion in fiscal year 2025. Then you have Government and Academic High-Performance Computing (HPC) centers, which utilize the technology for research and national projects, such as powering the top machines on the Green500 list.

Here's the quick math on how the revenue broke down across these customer-facing areas for fiscal year 2025:

The core customer types driving the Data Center segment include:

• Cloud service providers (AWS, Azure, GCP, OCI)
• Enterprise customers building AI infrastructure
• Sovereign AI initiatives
• Consumer internet companies using generative AI

What this estimate hides is the intense focus on securing supply commitments; NVIDIA's purchase commitments and obligations for inventory and production capacity were $30.8 billion as of the end of FY2025, showing how much they are pre-paying to serve these top segments.

Finance: draft 13-week cash view by Friday.

NVIDIA Corporation (NVDA) - Canvas Business Model: Cost Structure

When you look at NVIDIA Corporation's cost structure, you're seeing the financial reality of leading the accelerated computing revolution. The sheer scale of their revenue in Fiscal Year 2025-a massive $130.50 billion-is what makes the absolute dollar costs for R&D and operations look so large, yet their efficiency, or operating leverage, is what really matters for your analysis.

The most significant component, the High cost of revenue due to advanced chip fabrication (CoR), reflects the expense of designing and outsourcing the manufacturing of their cutting-edge GPUs and networking gear. For FY2025, the Cost of Revenue was $32.639 billion. That translates to a CoR as a percentage of sales of about 24.99% for the full fiscal year, which is a key metric showing how efficiently they are managing the direct costs of their products, even with the complexity of advanced node fabrication.

Next, consider the engine for future growth: Research and Development (R&D). NVIDIA is pouring capital into staying ahead of the curve, especially with the Blackwell architecture now ramping. For FY2025, R&D expense hit $12.91 billion. The good news for your valuation model is that this investment, while large in absolute terms, represented only 9.89% of that year's revenue, showing significant operating leverage compared to prior years.

Here's a quick breakdown of the major expense categories from the close of FY2025, so you can map it against that $130.50 billion revenue base:

You'll notice the Sales, General, and Administrative (SG&A) expenses are relatively lean for a company of this size, coming in at $3.49 billion, or just 2.67% of revenue in FY2025. This low percentage is a direct result of the massive revenue growth outpacing the growth in overhead staff and administrative costs; that's the operating leverage you want to see.

The Costs associated with global supply chain and logistics are embedded within the Cost of Revenue and operating expenses, particularly in the SG&A for managing that global footprint. Since NVIDIA operates a fabless model, they avoid the multi-billion dollar capital expenditures of building foundries, but they still incur significant costs managing the complex logistics, inventory risk, and securing capacity with partners like TSMC. This is a variable cost that scales with production volume.

Looking ahead, the company's forward guidance gives you a sense of near-term cost control expectations. For instance, the Non-GAAP outlook for the first quarter of Fiscal Year 2026 projected operating expenses to be approximately $3.6 billion. Still, you should watch the full-year FY2026 operating expense growth projection, which management guided to be in the mid-30% range year-over-year, even as revenue growth forecasts moderated slightly due to export controls.

To summarize the expense profile you're dealing with:

• R&D spending is a strategic investment, not just a cost; it was $12.91 billion in FY2025.
• The company is managing overhead well, with SG&A at only 2.67% of FY2025 revenue.
• The Q1 FY2026 Non-GAAP operating expense projection was set at $3.6 billion.
• Cost of Revenue, at $32.639 billion in FY2025, is the largest single cost line item.

Finance: draft the Q2 FY2026 OpEx forecast based on the mid-30% full-year growth guidance by Friday.

NVIDIA Corporation (NVDA) - Canvas Business Model: Revenue Streams

You're looking at how NVIDIA Corporation actually brings in the money, and right now, it's all about the data center. It's a massive shift from where the company was even a few years ago, but the numbers tell the whole story for fiscal year 2025.

Data Center GPU and System Sales were the undisputed engine, pulling in a staggering $115.19 billion in FY2025. Honestly, this segment's growth is what defines the company's current valuation. This revenue comes from selling the core AI accelerators, like the H100s and the newer Blackwell systems, to hyperscalers and enterprise customers building out their AI infrastructure.

Gaming GPU Sales, while still a huge business, is now a smaller piece of the pie compared to the AI behemoth. For FY2025, this segment generated $11.35 billion. It's still a healthy business, driven by high-end GeForce GPUs for gamers and AI PC users, but the scale is dwarfed by the data center demand.

Software and Support Subscriptions are the recurring revenue layer that analysts love to see building out. The projected annual run rate is approaching $2 billion by the end of 2025. This is tied to things like the AI Enterprise software licenses and support contracts that lock customers into the NVIDIA ecosystem, which is a key part of their moat.

Automotive Platform and Licensing Fees brought in $1.69 billion in FY2025. This stream is about selling the DRIVE platform and related software for autonomous driving and in-vehicle infotainment systems. It shows NVIDIA is successfully monetizing its compute expertise beyond the server rack.

Professional Visualization Hardware and Software Sales also contributed significantly, hitting $1.88 billion in FY2025. This covers the RTX Ada Generation GPUs and related software for designers, engineers, and media professionals who need serious rendering power.

To give you a clearer picture of the entire revenue landscape for FY2025, here is the full breakdown of the key segments:

The growth in these streams is heavily concentrated, which is important to note for near-term risk assessment. The key revenue drivers for the Data Center segment, which is the lion's share, include:

• Hyperscale cloud provider demand for AI infrastructure.
• Enterprise adoption of sovereign AI capabilities.
• Sales of full AI racks, not just individual chips.

Also, remember that the software component is designed to reinforce the hardware sales. If onboarding takes 14+ days, churn risk rises, but the subscription model helps secure long-term revenue visibility. Finance: draft 13-week cash view by Friday.