The robot that saves the bees.

An almond orchard in February, and a box that does not move.

It is bloom season in California's Central Valley, and a flatbed truck has just dropped a Beewise BeeHome into a row of almond trees. The box is roughly the size of a chest freezer. It is yellow. It hums. On its roof, a solar panel; inside, twenty-four honeybee colonies, a fleet of cameras, environmental sensors and a robotic arm that does the work a human beekeeper used to do with a smoker and a hive tool. The grower never sees a beekeeper this season. He sees an app.

This is, depending on your tolerance for marketing copy, either "the iPhone of beehives" - Inc.'s phrase, not ours - or a freight-rated piece of climate infrastructure aimed at a problem most of the food industry has decided to ignore. Beewise prefers the second framing. The first one sells better.

One in three bites of food. One in three colonies, gone.

Pollinators are responsible for somewhere between a third and a fourth of global crop output - almonds, apples, blueberries, cherries, the things people put on Instagram. The honeybee, Apis mellifera, is the species most of that work is outsourced to. Which would be fine, except commercial bee colonies in the United States have been collapsing at about 40% per year for the better part of a decade.

The reasons are a Russian novel: Varroa mites, pesticide drift, monocropped diets, queens that don't last, weather that won't behave. The result is simpler. Beekeepers spend their time driving from yard to yard, opening wooden boxes one at a time, inspecting frames they can barely keep up with, and absorbing losses no other industry would tolerate. The cost of pollination has gone up. The number of beekeepers has gone down. The bees, for their part, have continued dying on schedule.

A beekeeper walked into a tech founder's office.

In 2018, Eliyah Radzyner - a fourth-generation commercial beekeeper in Israel - sat down with Saar Safra, a serial software founder who had never, by his own admission, kept a bee. Radzyner explained, in detail, why beekeeping at scale was an impossible job. Safra listened, then asked the obvious software-person question: which parts of this could a machine do.

The answer, it turned out, was most of them. Inspection is visual. Pest treatment is mechanical. Climate control is thermodynamic. Feeding is logistics. The cofounding team - Safra, Radzyner, Hallel Schreier, Yossi Sorin and Boaz Petersil - bet that the right combination of computer vision, edge AI and a robot arm could turn beekeeping from a craft into a service. Investors agreed. The first cheques came from lool Ventures. Insight Partners and Fortissimo Capital followed, repeatedly.

The bet was not subtle. It assumed the world's pollination crisis would get worse before it got better, that growers would pay more for reliable bees, and that hardware-and-software combined could, eventually, become the lower-cost option. Seven years later, that assumption is starting to look prescient and slightly dull. Which is how good bets tend to look in hindsight.

What a robot beekeeper actually does at 3am.

The BeeHome 4 is the current generation. It draws its power from a roof-mounted solar panel and a battery. Inside, computer vision cameras watch each frame; sensors track weight, temperature, humidity and acoustic signature. The robotic arm dispenses food, redistributes brood between colonies, applies treatments and identifies queens. Entrances close on command, which matters more than it sounds: when a neighboring grower sprays pesticide, a beekeeper with a phone can seal twenty-four colonies in a single tap.

The hardware bit that journalists keep writing about is the Heat Chamber. Varroa mites - the parasite responsible for most US winter losses - die at a temperature that bees can survive. Beewise's chamber exploits the gap, reportedly killing 99% of mites without a single drop of miticide. Bees emerge slightly warmer, considerably less parasitised, and, as far as anyone can tell, mildly annoyed.

None of this is theoretical. As of mid-2025, Beewise reports 1,240 BeeHomes in operation, pollinating more than 300,000 acres a year for hundreds of growers. NVIDIA's Jetson platform runs the on-device inference. The pitch to a commercial beekeeper is mathematical, not poetic: same number of colonies, 90% less labor, 70% lower losses, one app.

The number that matters is the one that drops.

If you want to evaluate a company that claims to save bees, you check, first, whether the bees actually live. Beewise's published field numbers put annual colony loss in BeeHomes at around 8 to 10%, against an industry baseline above 40%. Independent reporting from Phys.org, Axios and VentureBeat has covered the same range. It is not a rounding error.

The customer list reads like a who-paid-the-bee-bill of American agriculture: large California almond and tree-nut growers contracting Beewise for managed pollination. The investor list - Fortissimo, Insight, APG Asset Management, lool, Badiya Capital, Marav Mazon Group, Austin Hearst - has put in roughly $170 million between them. None of those people are particularly known for sentimentality about bees.

Save the bees to save the food supply.

The mission statement is a single sentence and Beewise repeats it the way other companies repeat quarterly earnings. It is unfashionable in two directions at once. Climate purists find it too commercial - bees are a service, not a charity. Agribusiness pragmatists find it too mission-y - just sell the boxes. Beewise has answered by doing both.

The company is not coy about being a business. Hardware is sold and leased. Pollination is contracted by the acre. Software runs underneath. The argument is that a pollinator-as-a-service company that needs bees to live is, structurally, more aligned with bees than an industry that has been quietly losing 40% of them per year and absorbing the cost.

Pollination has been free for ten thousand years. That is ending.

For most of human agriculture, pollination was a gift from the ecosystem - something farmers received for the cost of not actively destroying it. That contract is being renegotiated. Climate volatility, monoculture, parasites and pesticide load have turned an ambient service into a managed one. Almond growers already pay rental fees per hive. Apple growers will. Berry growers will.

In that world, the question is not whether pollination becomes infrastructure. It is who runs the infrastructure. Beewise's bet is that the answer looks a lot like a yellow box, a robot arm, and a dashboard - boring, repeatable, software-defined. If they are right, the BeeHome stops being a curiosity and starts being plumbing. Which is to say, you will mostly stop hearing about it. Which is, in agritech, the highest form of success.

The almond bloom, three months later.

The Central Valley almond bloom has ended. The flatbed comes back. The yellow box is loaded, slightly heavier than when it arrived, and trucked to its next contract. No one opens it. No one needs to. A beekeeper somewhere in Modesto checks twenty-four colonies on her phone between cups of coffee. The grower looks at his yield numbers and signs next year's pollination contract before lunch.

The bees, the part that always mattered, are mostly still alive. That is the only metric Beewise has ever really been measured on. So far, they are passing it.

Interviews and demos

Links, press and signals