Selling to NVIDIA

NVIDIA reported 215.9 billion dollars of revenue in fiscal 2026, the year ended 25 January 2026, up 65 percent. It ended that year with approximately 42,000 employees in 38 countries. Its 10-K breaks that down: 31,000 in research and development, 11,000 in sales, marketing, operations and administrative positions.

Read those together before you plan this account. Eleven thousand people cover every commercial and back office function at a company turning over more than 215 billion dollars, roughly five million dollars of revenue per employee. There is no thick middle layer, no procurement bureaucracy to work and no army of programme managers to sponsor your pilot. The people who could buy what you sell are already doing three other jobs.

The second structural fact is that NVIDIA is fabless. It does not own the plants. That reshapes what a supplier relationship here means, and it is where most sellers start researching the wrong thing.

How the buying process actually runs

There is no published supplier registration route. As at the verification date on this page, nvidia.com carries no general supplier portal, purchasing guidelines document or vendor registration form. JPMorgan publishes a supplier network you can apply to, Microsoft publishes SupplierWeb, Apple publishes Supplier Connect. NVIDIA publishes a partner network instead. If your plan was to register and wait, there is nothing to register with.

The supplier machinery that is published points at manufacturing, and it is not about you. The 10-K states that NVIDIA uses a fabless and contract manufacturing strategy, naming Taiwan Semiconductor Manufacturing Company and Samsung for wafers, SK Hynix, Micron and Samsung for memory, CoWoS technology for packaging, and Hon Hai, Wistron and Fabrinet for assembly, testing and packaging. The assurance apparatus on top, built on the Responsible Business Alliance Code of Conduct with self assessment questionnaires and on site Validated Assessment Program audits for strategic suppliers, is aimed squarely at those firms. A software or services vendor passes through none of it.

The documented route in is the partner network, and its shape tells you what NVIDIA now buys. The NVIDIA Partner Network sorts members into four families: AI Infrastructure, Cloud, Influencer and Seller/Builder. Inside those sit Global Systems Integrator, Independent Software Vendor, Solution Advisor, Advanced Technology Partner, Storage Partner, Distributor, Original Equipment Manufacturer and Solution Provider, at Registered, Preferred and Elite levels. It also carries three types no software company's partner programme has: Data Center, Power and Cooling, and Architecture, Engineering and Construction partners. When a chip company certifies firms that distribute power and remove heat from buildings, the list of teams inside with a budget has changed.

There are two free front doors. NVIDIA Inception is free for AI startups, with no application fees, membership fees or equity requirement, and NVIDIA Connect is the equivalent for software development companies and service providers. Neither is a sales channel, but both make you visible internally.

Field operations changed hands in 2026. Ajay Puri, who led Worldwide Field Operations for 21 years, told the company in June 2026 he intended to retire, and Nick Parker joined from Microsoft into that role with a start date of 24 August 2026. Route to market, partner economics and coverage models are all under review at once.

Who owns what, so you know where your deal belongs

You will not be selling to these people. Their remits tell you which function your deal sits inside.

Notice what is missing. NVIDIA's newsroom lists four company officers alongside its two founders, Jensen Huang and Chris Malachowsky, a Founder and NVIDIA Fellow. There is no published chief information officer, no chief product officer and no divisional chief executive. At Microsoft you can read a leadership page and work out which organisation owns your problem. Here the page runs out after five names and the company reports in only two segments, Compute and Networking, and Graphics.

Names and titles taken from NVIDIA's own newsroom executive bios, its fiscal 2026 Form 10-K and its 2026 proxy statement, verified 11 September 2026. Senior roles change. Verify at source before using any name.

You cannot sell into this account from the bottom

NVIDIA spent 18.5 billion dollars on research and development in fiscal 2026 and states it has invested more than 76.7 billion dollars since inception. Its Data Center business alone produced 193.7 billion dollars of revenue. Against that, a deal has to be large to be worth anyone's calendar.

The build option is also unusually live. More than 80 percent of the workforce holds a technical role, more than half of employees hold an advanced degree, and the 10-K states that more than half of NVIDIA's engineers work on software. This is not a hardware company that buys its software outside. The competing bid is usually a team that already writes in your category, and that argument is settled in a room you are not in. Only a sponsor senior enough to overrule it keeps the deal alive, which sets the altitude your entry point has to reach.

The layer that matters is below the public names, and here it moves

Every page like this says the real buyers sit under the executives. At NVIDIA there is a specific reason it is hard to map. Revenue grew 65 percent in a single year against a 42,000 person workforce, while movement out stayed low: the company reports a fiscal 2026 turnover rate of 3.7 percent. The people are stable and their remits are not. Someone who ran a function two quarters ago now runs something several times larger under a different name. A buying committee assembled from a data provider's org chart will be wrong in ways you cannot see.

The board is a published second-degree map, and here it is a thin one

NVIDIA's board has ten members and, unusually, no chairperson. The company states that it gives each member an equal voice by having an independent Lead Director instead, a role held by Stephen C. Neal. Three of the ten, Tench Coxe, Harvey C. Jones and Jensen Huang, have served since 1993, the founding year.

• Stephen C. Neal, Lead Director, is Chairman Emeritus and Senior Counsel of Cooley LLP, where he was Chief Executive Officer from 2001 to 2008.
• Tench Coxe was a Managing Director of Sutter Hill Ventures from 1989 to 2020. He holds no current public company board seats.
• Harvey C. Jones has been Managing Partner of Square Wave Ventures since 2004, and previously led Synopsys as Chief Executive Officer and then Executive Chairman.
• Mark A. Stevens has been Managing Partner of S-Cubed Capital since 2012, was a Managing Partner of Sequoia Capital from 1993 to 2011, and is a trustee of the University of Southern California.
• A. Brooke Seawell has been a Venture Partner at New Enterprise Associates since 2005 and currently sits on the board of Tenable Holdings.
• Melissa B. Lora, a former President of Taco Bell International, currently sits on the board of Conagra Brands.
• Aarti Shah, former Senior Vice President and Chief Information and Digital Officer of Eli Lilly and Company, currently sits on the board of Sandoz Group AG and is a trustee of Northwestern Mutual.
• John O. Dabiri is Centennial Professor of Aeronautics and Mechanical Engineering at Caltech and a trustee of the Gordon and Betty Moore Foundation.
• Dawn Hudson is the former Chief Marketing Officer of the National Football League and former Chief Executive Officer of Pepsi-Cola North America, and holds no current public company board seats.

Read that as a seller. Across the entire board there are only three current outside public company directorships: Conagra Brands, Tenable Holdings and Sandoz Group AG. At Apple or Microsoft the board opens doors to a dozen large corporates. Here the shared-board-seat play returns almost nothing.

What it returns instead is venture capital and institutions: Sutter Hill Ventures, New Enterprise Associates, Square Wave Ventures, S-Cubed Capital and the Sequoia network, plus Cooley LLP, Caltech, the University of Southern California and Northwestern Mutual. If your company is venture backed, that is a far more useful map than a corporate board list. Cooley is worth checking first, because it is counsel to a large share of the venture backed technology market and may already be counsel to you.

Other connector routes, and which of them actually exist

There is no company-run alumni network to work. NVIDIA operates nothing equivalent to the independent nonprofit that Microsoft alumni run. The 10-K says something more useful: over 40 percent of its fiscal 2026 new hires came from employee referrals. This company already hires through personal networks, so a referral from a current employee carries weight.

Mellanox alumni are the significant acquired-company pool. The 2020 acquisition brought in the networking business that now sits at the centre of the data center platform, so anyone who worked there before the deal has former colleagues in the fastest growing part of NVIDIA.

Partners and integrators are already inside. Every systems integrator, ISV and cloud partner in the NVIDIA Partner Network has a named NVIDIA contact and a reason to keep it busy.

Customers on both sides of the transaction. NVIDIA's direct customers include OEMs, ODMs, system integrators, distributors and the major cloud providers, and the 10-K discloses that in fiscal 2026 one direct customer represented 22 percent of total revenue and another 14 percent. If you sell to any of them, you are one hop from people who talk to NVIDIA weekly.

Working out which of those paths exists inside your own company's address book is what a relationship intelligence platform is for. At an account with almost no public org surface, it is not a convenience, it is the method.

And then you have to get to your own connector

This step fails more introductions than the external ask does. Suppose the best route runs through a partner at one of your investors. You still have to get that person to prioritise it and spend a piece of a relationship built over years, on you. Ask twice in a quarter and you will not be asked again. Two of your own reps approaching the same connector about NVIDIA in the same month is common, and at most companies invisible.

Three rules for the introduction itself

Ask for the team, not the company. "An introduction to whoever owns data center facilities partnerships" is actionable. "An introduction to NVIDIA" is not, at a company that publishes five executive names and two segments.

Write the blurb your connector will forward. They will not write your pitch. Give them a paragraph they can send without editing. The forwardable blurb is the whole craft.

A warm path starts the clock, it does not skip it. You will still negotiate contracting and, depending on what you sell and where, export control review. What it buys is the build versus buy conversation months earlier, and a second person in the room when an internal team says it could write this itself. That is why multithreading matters more here than the size of the company suggests.

See the warm paths into NVIDIA

Everything above describes the problem: eleven thousand people running the entire commercial surface. What it does not tell you is which of your own relationships already reaches inside. That is a question about your network, not theirs, and it is the one most teams answer from memory.

Boomerang maps the warm paths your company already has into accounts like NVIDIA, across your team, your customers, your board and investors, and your partners. It then drafts the ask, routes it through the right connector and tracks it to a booked meeting. Book a 15-minute walkthrough and see it run against your own target accounts.