Enquirer Consulting Group

Reachable Buyer Map

Prepared for Jelena Erakovic · NextSilicon · United States · August 2026
In the United States, your market reads from the outside as a short list of very large doors: the national laboratories, the federal research programs around them, and the university computing groups that share their people. That list is real, it is defensible, and it is finite. This map is what sits next to it. The US commercial organizations running the same class of workload, who signs inside each one, and roughly how many of them there are. It describes the market rather than your business, and there is nothing to buy at the end of it.
National laboratories and federal research centers
The channel your public record already sits in, and the hardest one for anyone to take away. Worth stating the limit plainly rather than politely: it is a fixed number of doors, so growth inside it is depth in accounts already known rather than new names.
Who signs: laboratory computing director, HPC program lead, principal investigator, and the contracting officer on the acquisition side.
Roughly 40 to 50
national laboratories and federally funded research and development centers in the US; public bodies file nothing, so this group is named rather than counted
Doctoral research universities
Every university with a serious science program runs its own computing group, and the person who owns it holds a public title. Smaller budgets than the labs and a grant-shaped buying cycle, which makes this a landing motion rather than a flagship one.
Who signs: director of research computing, CIO, chair of the research computing committee, and the funded faculty investigator.
Roughly 250 to 300
US doctoral research universities, of which roughly 140 to 160 carry the heaviest research load; public institutions are described rather than counted
Scientific research and development companies
Contract research, computational biology, materials, applied physics. They buy compute the way a factory buys a machine, against one throughput problem with a date on it, and almost none of them keep a team on hand to rewrite working code.
Who signs: CTO, VP of research computing, head of scientific computing, director of infrastructure.
900 to 1,000
US employers registered in scientific research and development services; roughly 800 to 900 carry 50 or more people on the plan
Energy: subsurface and seismic
Seismic imaging and reservoir simulation are among the oldest commercial workloads of this kind, and the codes are decades old and load bearing. That makes portability the entire conversation before performance is even discussed.
Who signs: head of geoscience computing, VP of exploration technology, director of HPC, subsurface technology manager.
350 to 450
US employers registered in oil and gas extraction; roughly 300 to 350 carry 50 or more people on the plan
Quantitative finance and risk
Risk, valuation and backtesting runs overnight inside fixed windows, so the buying trigger is precise: a window that stopped fitting. The seat that owns it is technical, titled, and unusually easy to reach directly.
Who signs: head of quantitative development, CTO, director of compute or grid engineering, head of risk technology.
2,300 to 2,500
US employers registered in securities, investment and fund management; roughly 1,100 to 1,200 carry 50 or more people on the plan
Semiconductor and electronics design
Verification and simulation workloads that grow with every process node. This is also the one segment that understands your architecture without translation, which shortens a first conversation from an hour to ten minutes.
Who signs: VP of engineering, director of design infrastructure, head of CAD and EDA, IT director for silicon engineering.
380 to 450
US employers registered in semiconductor and electronic component production; roughly 320 to 380 carry 50 or more people on the plan

Where the openings are

1
The lab channel is a fixed number of doors. Forty to fifty nationally, all known to you and all known to everyone else selling into them. Every segment above them on this page is a list problem rather than a relationship problem, and lists get built. They do not arrive.
2
Your strongest public argument is the one least likely to reach the buyer who needs it most. Existing code in C, C plus plus, Fortran and OpenMP running without a rewrite matters most to the commercial buyer, because that buyer has no rewrite team and no appetite to hire one. From the outside, that argument travels mainly through conference talks and lab relationships, which is exactly where the people who already know you are standing.
3
The buyer is a title, not a company. Director of research computing, head of quantitative development, head of geoscience computing. Across the whole country these seats number in the low thousands, which is small enough to name every one of them and work them individually. Very few companies are in that position, and the position is wasted unless somebody works those seats deliberately.
4
Awareness and access are different problems. Being known in the HPC community is not the same as being in front of the handful of people inside a bank or a fab who can start a pilot. One is a reputation job, and the public record says it is largely done. The other is a distribution job, and it is the one that is open.
Built from public federal registry data covering US employers that file a full benefit plan return, which in practice means established organizations rather than the whole market. Counts are banded deliberately. Government bodies, national laboratories and public universities do not file, so those groups are described rather than counted. Sector codes are self-reported by the organizations themselves.
ENQUIRER CONSULTING GROUP
enquirerconsulting.com