Inside the Dien Bien drone sandbox, the Ca Mau rice pilots and the 60 regulatory exemptions that decide whether 6,000 aircraft change a 43-million-tonne harvest
By: Logan Pierce – SeaPRwire – The number that decides whether Vietnam’s agricultural technology story is real is not 6,000. It is 20.
Six thousand is the size of the country’s agricultural drone fleet, a figure that appears in every provincial report and every investor deck. Twenty is the number of locally trained pilots that Ca Mau province requires each of its two rice pilot zones to produce. One is a fleet count. The other is a labour count, and only one of them tells you whether a hectare of the Mekong Delta will be sprayed by a machine next season or by a man with a backpack sprayer who has done it that way for thirty years.
Ca Mau registered 55,000 hectares into the one-million-hectare low-emission rice programme. Inside that area it carved two pilot zones, each between 100 and 500 hectares, and wrote a rule that each zone must train at least 20 local operators. Run the arithmetic: 200 to 1,000 hectares of active pilot inside 55,000 hectares of registration, which is 0.4 to 1.8 percent of the province’s commitment. The 55,000 hectares is a signature on a document. The 200 to 1,000 hectares is what a drone will actually fly over this season.
That gap is not a scandal. It is how agricultural programmes work everywhere, and it is the most useful thing to understand about Vietnam’s agritech moment. The policy layer has moved faster than any other part of the system — faster than the data layer, faster than the training layer, and much faster than the financing layer. What follows is an attempt to walk the chain from the sandbox in Dien Bien down to the paddy in Ca Mau and see where it holds.
Six thousand aircraft and a seventy percent import bill
Start with the hardware, because it is the part everyone can see.
Vietnam’s agricultural drone fleet reached roughly 6,000 units in 2025. More than 70 percent of them carry Chinese brands. The aircraft are cheap, they work, and the service network that supports them is built around spare parts shipped from Shenzhen rather than from a factory in Can Tho.
That ratio gets discussed as a dependency problem. It is better understood as a price discovery outcome. A crop-spraying drone in the Mekong Delta is a tool with a two-to-three-season payback at prevailing spraying rates, and the operator buying one is a farmer or a small cooperative with limited capital and no ability to wait for a domestic manufacturer to reach scale. Chinese manufacturers reached scale first, so they set the price, so they won the fleet. There is no mystery in it and no policy failure either.
What the ratio does not tell you is where the value sits. An airframe is a commodity. The application map that tells the drone where to spray, the soil sensor network that tells it how much, and the traceability record that tells a buyer in Manila or Tokyo what happened to that field — those are not commodities, and Vietnam’s position in them is much weaker.
The state has noticed. Resolution 57, issued by the Politburo in December 2024, put science and technology at the centre of national development strategy. Resolution 85, in April 2026, set a target for Vietnamese agriculture to rank among the global top 15 by 2030. A priority high-tech investment list now names about 70 categories eligible for incentives. These are the documents that fund the next layer, and they are why the data centre and the soil sensor companies exist at all.
The sandbox is a scorecard, not a deregulation
Dien Bien province, in the far northwest, is not where you would expect a national experiment in low-altitude economics. It is mountainous, sparsely populated, and nowhere near the delta. That is partly why it was chosen: a crash there damages a hillside instead of a village.
The sandbox approved 6,000 test flights and exempted operators from about 60 certification and testing requirements. Ten provinces have expressed interest in replicating it. The programme runs until May 2027.
Read the exemption number alone and it sounds like Vietnam took the brakes off. Read the targets attached to it and it sounds like something else entirely.
| Sandbox parameter | Value | Type |
| Approved test flights | 6,000 | actual |
| Certification requirements waived | 60 | actual |
| Provinces seeking to replicate | 10 | actual |
| Flight safety target | 99% | target |
| Business models to be validated | 3 to 5 | target |
| Cost and time improvement target | 20% to 30% | target |
| Programme end | May 2027 | actual |
Six thousand flights against a 99 percent safety target is a test with a pass mark. Three to five business models is a shortlist, not an open field. The 20 to 30 percent cost-and-time improvement is the number that will decide whether any of the models survives contact with a commercial market after 2027. This is a structured experiment with a defined exit, and the exemptions exist to make the experiment possible, not to make the industry permanent.
That distinction is the one most foreign coverage misses. A sandbox that exempts you from 60 rules for 22 months is not a deregulated market. It is a market where the rules have been suspended long enough to find out what the rules should be — and where everybody operating inside it knows the suspension has a date.
Why the pilot zone is measured in hundreds of hectares
The Mekong Delta’s structural problem is not water, and it is not yield. It is that the average rice plot is small, fragmented across several parcels, and held by a household that makes its decisions independently of the household next door.
A drone spraying operation has fixed costs per flight: travel, battery, calibration, the pilot’s time. Spread those over 0.3 hectares and the unit economics die. Spread them over 300 hectares and they work. This is the entire reason the pilot zone exists as a concept, and it is the reason Ca Mau’s 100-to-500-hectare zones matter more than its 55,000-hectare registration.
The zone is the unit of aggregation. It is the smallest land area over which a single operator can amortise a flight, and it is the smallest area over which a single traceability record can be issued that a buyer will accept. Below that threshold you have a farmer with a drone. Above it you have an agricultural service business.
Now the personnel arithmetic. Twenty trained pilots per zone, two zones, means at least 40 people with a certification that did not exist in Ca Mau three years ago. If the model scales to the full 55,000 hectares at the same pilot-to-area ratio implied by the zones, the province needs somewhere between 2,200 and 5,500 trained operators. Nobody has announced a training programme of that size. The 20-person requirement is a pilot-scale number, and it is honest about being one.
The million-hectare programme is a credit story
Here is where the standard reading of Vietnam’s rice programme goes wrong.
The one-million-hectare low-emission rice programme is usually described as an export play: better rice, premium prices, access to stricter markets. The export numbers look like they support that. Vietnam produces about 43 million tonnes of rice a year at roughly 6 tonnes per hectare. In January 2026 it shipped 331,000 tonnes to the Philippines, which took 51 percent of the month’s exports. Shipments to Ghana rose 21.3 percent, to Cote d’Ivoire 67.5 percent, and to Bangladesh by a factor of 125 from a base small enough that the multiple is close to meaningless.
But the programme’s own targets point somewhere else. A 20 to 40 percent cut in methane emissions. A 30 percent increase in net profit for participating farmers. Those two numbers are not export metrics. They are collateral.
Methane reduction is measurable, verifiable and increasingly sellable as a carbon credit. A 30 percent profit increase is what makes a farmer creditworthy enough to borrow against the next season. Put them together and the programme’s real product is a data record that a bank or a carbon market will accept — not a premium-priced bag of rice.
That is why the AI-MRV layer is the part to watch. Remote sensing across 80,000 hectares at TTC AgriS, with a claimed 95 percent yield forecast accuracy that the company itself labels a forecast rather than a result. Five hundred soil sensors deployed by enFarm. A $95 million data centre agreement between Viettel and KT. A $70 million Series A for Techcoop, which is by a wide margin the largest venture round in the sector and larger than most of the public programme budgets.
Trace the money and the priority is plain. Vietnam is not funding rice. It is funding the measurement of rice, because the measurement is what gets securitised.
| Programme metric | Value | Basis |
| Registered area, one-million-hectare programme | 1,000,000 ha | target |
| Ca Mau share of programme | 55,000 ha | actual |
| Ca Mau share as percentage of programme | 5.5% | derived |
| Rice yield, national average | 6 t/ha | actual |
| Annual paddy output | 43,000,000 t | actual |
| Methane reduction per hectare | 20% to 40% | actual, programme reporting |
| Net profit change for participating farmers | +30% | actual, programme reporting |
| Rice exports to the Philippines, January 2026 | 331,000 t | actual |
| Philippines share of January exports | 51% | actual |
Read the last two rows against the fifth. Vietnam exports a large share of a 43-million-tonne crop, and about half of a given month’s shipments go to one buyer. That concentration is the reason the programme exists and the reason the premium-price story is weak: a buyer who takes half your exports does not pay a premium for a certificate, because the buyer knows you have nowhere else to put the tonnage in a harvest month.
The carbon and credit reading is stronger. A verified 20 to 40 percent methane reduction on one million hectares is a very large number of credits, and it is the kind of asset that Japanese, Korean and Singaporean buyers have been contracting for since 2024. The 30 percent profit increase does something subtler: it changes the farmer from a credit risk into a credit customer, which is the precondition for every input-financing model that agritech companies actually want to sell.
The three provincial models are not variations on one theme
It is tempting to treat Dien Bien, Ca Mau and Vinh Long as three data points on a single adoption curve. They are three different experiments with three different success criteria, and the differences matter more than the similarities.
| Province | What was built | Scale | Success criterion |
| Dien Bien | Low-altitude aviation sandbox with certification waivers | 6,000 flights, 60 exemptions | Safety rate and commercial viability by May 2027 |
| Ca Mau | Aggregated rice pilot zones inside a registered area | 2 zones, 100 to 500 ha each, 20 trained pilots per zone | A repeatable administrative unit a certifier will accept |
| Vinh Long | Drone spraying at service-provider scale | Province-wide, backed by 5 resolutions | Chemical reduction and yield gain per hectare |
Dien Bien is a regulatory experiment. Ca Mau is an institutional experiment. Vinh Long is an agronomic experiment. Only Vinh Long has produced numbers that a farmer can act on, and only Dien Bien has produced a framework that a ministry can extend nationally. Ca Mau, the youngest of the three, is the one that will determine whether the other two compound, because it is building the unit — the zone — that both of them need in order to scale.
No provincial agriculture department contacted for this article would discuss the sequencing on the record. The programme is under review, the review has not concluded, and the officials who administer it are civil servants in a system where speaking ahead of a ministry statement is not a career move. What is public is the design itself, and the design says plainly that the aircraft arrived before the paperwork, and that the paperwork is what banks read.
Vinh Long already ran the experiment
Two provinces south of Ca Mau, Vinh Long published the results that the rest of the delta is now trying to reproduce.
Drone spraying cut pesticide and fertiliser use by 20 to 30 percent. In some models, yields rose 15 to 20 percent. The province backed it with five provincial resolutions and an AI-MRV platform to record what happened.
Those are good numbers, and they are also the numbers most likely to disappoint at scale. A 20 to 30 percent reduction in chemical use is easiest to achieve on a field that was over-applying to begin with, which is exactly the field a pilot programme selects for. A 15 to 20 percent yield increase is easiest to achieve on a plot that had a pest problem, in a season with favourable weather, with a technician present. The phrase “in some models” in the province’s own reporting is doing a lot of work, and it is the phrase that should travel with the number.
The honest version is this: the direction is established and the magnitude is not. What Vinh Long proved is that the mechanism works when somebody is paying attention. What it did not prove is that the mechanism survives when nobody is.
The date that decides whether any of this is permanent
May 2027 is not a soft deadline. It is the date on which 60 exemptions stop applying and every operator flying commercially in Dien Bien becomes, in the eyes of the law, either licensed or illegal.
Nobody has published the post-2027 rulebook. That is not unusual — sandboxes everywhere produce their permanent rules at the end of the pilot rather than the beginning — but it creates a specific and predictable behaviour in the meantime. An operator who expects to need a licence in 22 months does not buy a second aircraft in month six. A cooperative that expects its certification regime to change does not sign a three-year service contract. The rational move is to wait, and waiting is what the sector will do unless the Ministry of Transport publishes a draft before the end of 2026.
Compare the structure to the rice programme, which has the opposite problem. There, the rules are permanent and the enforcement is soft: a province can register 55,000 hectares without any mechanism to verify that a hectare is being farmed to the low-emission protocol. The sandbox in Dien Bien is strict and temporary. The rice programme is permissive and indefinite. Vietnam’s agritech sector is being governed by two instruments pulling in opposite directions, and no document reconciles them.
The reconciliation, when it comes, will look like the Ca Mau zone. A defined area, a membership list, a measurement protocol, an audit trail. That is a design that satisfies both a safety regulator and a carbon certifier, and it is the only design that can be extended to a million hectares without a million inspectors.
The dependency that matters is not the airframe
Return to the 70 percent import share and ask what Vietnam would actually lose if it could not buy another Chinese drone.
Not much, in the short run. Airframes are the most substitutable component in the stack, and a fleet of 6,000 units has a replacement cycle measured in years. A domestic assembler could serve the replacement market at a modest premium within two or three seasons.
The irreplaceable layer sits underneath: the flight-control firmware that decides where the aircraft goes, the application maps that decide what it releases, and the cloud platform that stores the record. Those are licensed, updated remotely, and governed by terms of service that a Vietnamese cooperative has never read. If the dependency question matters — and it does — it matters there, at the layer where an update pushed from another country can change what happens over a Vietnamese field.
Which is why the $95 million Viettel-KT data centre agreement is a more consequential piece of industrial policy than any drone subsidy. Data sovereignty in agriculture means the field record stays in the country. Everything downstream of the field record — the carbon credit, the bank loan, the export certificate — depends on who holds it.
What the field shows that the reports do not
Three things show up on the ground that no provincial scorecard records.
The first is that the operator is usually not the farmer. Drone spraying in the delta is a service business run by a cooperative or a small entrepreneur who owns two or three aircraft and serves thirty to fifty households. The farmer’s decision is not whether to buy a drone. It is whether to hire one, and that decision turns on a single comparison between the per-hectare spraying fee and the labour cost of doing it by hand in a province where agricultural labour is getting older and scarcer every season.
The second is that the training requirement is the binding constraint, not the capital. A drone can be financed. A person who can calibrate a spray pattern, read a weather window, and fill in a digital record that satisfies an auditor cannot be financed as easily, and cannot be imported at scale in a country where the language of the certification is Vietnamese.
The third is that the pilot zones are producing a template, not a product. What Ca Mau is building in those 100-to-500-hectare blocks is a repeatable administrative unit: a boundary that a satellite can see, a membership list that a cooperative can verify, a measurement protocol that a certifier will accept. If that template works, the number of hectares inside it is a rounding detail. If it does not, 55,000 hectares of registration is a press release.
Count the trained pilots, not the aircraft
The most reliable indicator of Vietnam’s agricultural technology transition is not the fleet size, the export volume, or the venture funding. It is how many people hold a certification that lets them operate a spray drone commercially, and how quickly that number compounds.
Twelve months from now, if the count in Ca Mau has moved from 40 toward several hundred without a new decree, the model is working and the hectare count will follow. If the count is still in the dozens and the sandbox has been extended past May 2027, then the constraint was never technology. It was the supply of people who can be trusted with a machine over somebody else’s field, and no amount of procurement solves that.
Six thousand aircraft is a fleet. Twenty pilots is a system. The system is the harder number to move, and it is the only one that survives a bad season.
Sources
- VietNamNet reporting on the Dien Bien low-altitude economy sandbox, July 2026
- Ca Mau Department of Agriculture and Environment, rice pilot zone designations, June 2026, via vietnam.vn
- Vinh Long Department of Agriculture and Environment, drone spraying pilot results, February 2026, via vietnam.vn
- Nation Thailand citing Vietnam’s Department of Trade Promotion, rice output and export series, 2026
- Politburo Resolution 57 on science, technology and innovation, December 2024
- Resolution 85 on agricultural development targets, April 2026
- Digital in Asia, Techcoop Series A coverage, 2025
- Track B agritech dataset, Vietnam cell, knowledge base compiled October 2026