The most valuable dogs in the world are not pets

In Bengaluru, 16 dogs go to work…

Rose, Billy, Iggy, Whiskey, Phoebe, Cheetah, and Kaalu, an eight-month-old German Shepherd and Malinois mix who is the youngest on staff. The roster sounds like an obscure rock group.

They spend thirty minutes to an hour a day on a detection platform sniffing human breath samples, and their employer, a startup called Dognosis, designs the work as a game. Their compensation is treats.

Four thousand miles away, in a laboratory outside Tel Aviv, six Labrador Retrievers do a version of the same job. Three of them are Mars, Moon and Pluto. They were bred on site for the work, and between December 2021 and December 2023 they sniffed 1,386 breath samples in a double-blind clinical trial.

The people paying for what both groups of dogs produce are hospitals, port authorities and venture investors.

The pet industry knows working dogs exist. It sells them harnesses, goggles, boots and cooling vests. But the industry is built around the dog as a pet, an animal that consumes food, insurance, toys, grooming and vet care and produces companionship in return. Nearly every model this newsletter covers sits downstream of that.

There is a second economy where the dog produces the value instead of consuming it. It has federal contracts, real revenue and, in one narrow corner, venture capital.

The work spans a long list of jobs, most of which nobody thinks of as an industry. Some of it is famous and small, like truffle hunting and search and rescue. Some of it is invisible and large. Here are the corners where the money sits.

The first and biggest is security screening. When Allied Universal acquired MSA Security in September 2021, MSA had revenue approaching $200M and ran nearly 700 explosive-detection canine teams. Those two numbers do not divide into a per-team figure, because MSA sold plenty of non-canine security work alongside the dogs. Nobody in this business publishes a per-dog or per-team price anyway, which is most of the problem with sizing it.

Agriculture and biosecurity is the second, and it runs on appropriations rather than customers. Customs and Border Protection fields the teams that inspect what crosses the border. USDA breeds and trains the supply, graduating roughly 75% of the 65 to 80 dogs it takes in each year. Because it is a government program, it is also the only corner of this economy where headcount is published and reliable.

Conservation is the third, and it runs mostly on donations rather than fees. The field's pioneer is Conservation Canines, founded in 1997 by Samuel Wasser at the University of Washington, which trains rescue dogs to find scat from endangered species and has worked on everything from killer whales to pocket mice. Working Dogs for Conservation, one of the larger nonprofits in the field, reported $2,360,188 of revenue against $2,167,942 of expenses in its most recent Form 990. The whole sector is smaller than one mid-market pet brand.

Pest and structural inspection is the fourth and the most familiar, even if you have never thought about it as an industry. Anyone who lived in New York in the last fifteen years knows Roscoe, the beagle from Bell Environmental who ran cheerfully cheap local commercials and became, for a while, the most reassuring face attached to the second most feared creature in the city.

Roscoe started at Bell in 2008 and worked his way onto Good Morning America and into the New York Times. He was also a business. NESDCA, the main accrediting body, currently lists around 125 certified handler-and-dog teams across roughly 44 companies, though accreditation renews annually and other organizations certify teams too, so there is no fixed national count. We could not find a published price from any of them, which is unsurprising for work that gets scoped building by building. The closest verified figure is adjacent, a mold detection firm charging $899 for up to 2,000 square feet, roughly what a mid-range home inspection costs.

Medical detection is the newest corner and the only one venture capital has entered. It is also the only one where the dog's output ends up in a peer-reviewed journal rather than an invoice, which is why two small startups in India and Israel are the most closely watched operations in this entire economy.

Nobody can tell you how big any of it is, including the one corner that has been sized repeatedly. Four market research firms have tried to put a number on canine security services. They came back with $1.45B, $3.64B, $4.97B and $13.38B, all for roughly the same year, and none of them explains how they got there. When four estimates of the same thing disagree by an order of magnitude, the honest answer is that this economy has never been measured properly.

There is a version of this story where trained dogs are a charming stopgap and an electronic nose, a sensor array built to identify airborne chemical compounds the way an olfactory system does, eventually does the job cheaper and at scale. Thirty years into that expectation, no electronic nose has taken over a working-dog program anywhere we could find.

CBP's agriculture canine program went from over 160 active teams in 2024 to over 180 teams as of July 2026. Same program, two dates, the government's own numbers both times. TSA still deploys more than 1,000 canine teams after decades of investment in X-ray and trace detection hardware. The machines arrived and the dogs stayed.

Dognosis plans to double its canine workforce to 30 while it builds brain-computer interfaces to read those animals more precisely. It is adding technology and adding dogs at the same time.

The science explains why. A 2026 review of medical electronic noses found they correctly flag cancer in roughly 85% of cases across the published studies, against dogs that clear 90% or better in the best trials, and the machines still drift out of calibration and give different answers in different labs. A dog handles complex, messy odor mixtures and can be retrained onto a new target in weeks. A sensor has to have its chemistry designed in advance for the specific thing it is looking for.

We could not find a venture round into any company that owns and deploys working dogs, in security screening, pest inspection, biosecurity, agriculture or conservation, going back three years.

What happened to those companies instead was consolidation. Allied Universal bought MSA. Constellis bought American K-9 Detection Services. Global K9 Protection Group acquired the assets of a competitor. K2 Solutions divested its canine operations outright in January 2025. Ownership changing hands, private equity rather than venture, no new money arriving to build anything.

The venture money went to a different set of companies, all of them building hardware and software that captures what a dog's body and brain do at the moment it detects something, and turns that into a signal a lab can act on.

Canaery, now trading as General Sense, raised just over $4M in 2022 from Breakout Ventures, Dolby Family Ventures, KdT Ventures and SOSV to build a neural interface that reads olfactory activity directly off a working animal. It does not sell a replacement for detection dogs. It sells into the operators that already have them. SpotitEarly had raised $20.3M in total when it entered the US market in May 2025, from Hanaco VC, Menomedin VC, Wix chief executive Avishai Abrahami, former Timberland chief executive Jeff Swartz, and grants including the binational BIRD Foundation. It opened a Series A to accredited investors the same day, and has since opened a Regulation A round to everyone else. Dognosis raised $1.5M in 2024 from Boost VC, 1517 Fund, Accel India and Caffeinated Capital, plus a later round it has not sized.

Those are the three, and together they account for something in the region of $26M. That is the disclosed venture position in an economy that screens the air at every major US airport, guards the food supply and is now running cancer trials at Fox Chase and the University of Pennsylvania. Butternut Box, which sells fresh dog food, has raised more than €451M since inception, roughly twenty times everything above.

Both medical companies describe themselves accordingly. SpotitEarly's platform collects physiological and behavioral signals while the dogs work. Dognosis is adding computer vision and wearable EEG to capture brain activity before and during the sniff. Neither is trying to breed a better dog. Both are building the machinery that turns one animal's judgment into a repeatable, auditable, scalable output.

The dog is becoming a platform. Everyone is selling the wiring.

SpotitEarly is unusually visible for a company its size. Its Regulation A offering means it files audited financial statements with the SEC, and those filings are the only place anyone can see what a business built on dogs looks like on paper.

The trial that produced its headline numbers used six Labrador Retrievers. That count appears in the peer-reviewed literature. It does not appear in any of the five offering documents the company has filed. Neither does any other dog count, before or since.

The accounting explains why that is not an oversight. Research and development expense, the filing says, consists primarily of laboratory equipment and the costs of acquiring and training dogs that have no alternative use. Total company property and equipment at the end of 2025 came to $137K. The dogs are not an asset on the balance sheet. They are a running cost, in the same line as the equipment.

The trial reports 93.9% sensitivity for the combined system, the figure that appears on every marketing page. Table 6 of the same paper reports what the dogs achieve without the AI layer, which is 64.4%. The paper says performance dropped dramatically when AI was not used. The offering documents say only that the AI substantially improves upon the baseline performance of the sniffer dogs, and never say what the baseline is.

None of this is buried. It sits in a Nature paper anyone can read. But a company whose entire premise is canine olfaction has told the SEC almost nothing about the animals, and quite a lot about the software.

An uninstrumented dog is not a diagnostic, and the same 64.4% figure is the evidence.

Institut Curie ran a decade-long canine breast-cancer program and published its KDOG1 trial results in November 2025. Across 181 women, 80% of positive samples were identified by at least one of the two dogs and only 49% by both, which the team called insufficient for clinical use. The program is now working to identify the compounds the dogs were responding to, so an electronic nose can be built against them. The dogs were not being pushed aside by a better machine. They were running into the limits of what a dog alone can do.

The operating costs are heavy too. On the federal side, where the numbers are public, a green dog runs around $8,000 before training, TSA training costs $33K to $46K per team, and GAO put the fully loaded first-year cost of a deployed federal team at $85K to $223K, with $54K to $154K a year after that. Roughly a fifth to a quarter of candidate dogs wash out. Average working life is about five years.

Medical Detection Dogs, the UK charity running canine disease detection since 2007, shows the same shape from the other end. It reported £2,064,349 of income against £2,599,524 of expenditure in the year to March 2025, and it takes 57 employees and 655 volunteers to keep 18 dogs working. That ratio is the constraint in this whole category. Most of the cost sits in people rather than animals, and the human layer is exactly what an instrumentation company is built to reduce.

Both companies are betting that a dog plus good software reaches a working diagnostic faster than a sensor alone will. So far the published evidence supports them, and it is also why the software is where the money went.

A dog picking cancer out of a breath sample is the remarkable part of this story, and it is now documented in peer-reviewed journals rather than folklore. The part worth attention here is the direction the money moves.

Everything in this industry moves money from a human to an animal. Food, insurance, orthopedic beds, the whole shelf. This economy runs the other way. The animal produces something, a hospital or a port authority or a landlord pays for it, and the beneficiary at the end of the chain is a person.

That makes it a strange and possibly underrated place to put money if you like this category. An investor backing Dognosis is not buying pet exposure. They are buying an oncology screening business that happens to run on Labradors, and it will be priced against health tech and sensor comps rather than pet multiples. The dog is the mechanism, not the market.

There is also a simpler commercial idea sitting here that nobody has taken. These are elite working animals with public profiles, real performance records and, in Roscoe's case, a television career. Pet food and supplement brands already sponsor agility competitors and sled teams. A detection dog running trials at a major cancer center is a more interesting endorsement than most of what the category currently buys, and the sponsorship would cost a fraction of a media plan.

And for anyone building wearables or sensors, this is one of the few places a dog-attached device has an institutional buyer rather than an owner who churns in four months. We wrote about at-home diagnostics earlier this year and how hard it is to keep an owner paying for data about their own animal. The subscription problem looks different when the entity paying for the data is a health system running a clinical trial.

Roscoe retired years ago. Somebody replaced him, because the work did not go away and nothing else does it as well. That is the part the funding data keeps missing. The dogs are fine. It is everything built on top of them that is up for grabs.

If you operate anywhere near this, tell us what you are seeing.