One Soldier, 200 Drones: What Singapore Chose to Own

Analysis as of 5 August 2026.

On 3 July 2026, a team inside Singapore's Defence Science and Technology Agency received a public-sector innovation prize for something genuinely unusual: engineers had adapted the technology behind commercial drone light shows so that one soldier could direct as many as 200 aircraft. The programme carries the name LOCUST — Low-Cost Unmanned Sense-Strike Technology — and it was assembled by integrating commercial technologies into Robotics C3, a command backbone the agency writes in-house. The headline number is not the interesting part. The interesting part is what the state decided to keep for itself.

Start with the distance between claim and practice. At Exercise Wallaby 2025 — held from 7 September to 2 November at Shoalwater Bay, an overseas training area roughly five times Singapore's own size — a fifty-drone swarm flew, one item among twenty-plus technologies trialled across an event involving upwards of 5,000 troops and more than 500 platforms. Fifty flown; 200 claimed. Nothing published closes that gap. The fiscal setting makes the experimentation affordable: defence spending projected near S$23.4 billion for FY2025, a 12.4 per cent jump year on year, yet still held inside a decade-long band around 3 per cent of GDP.

The civilian substrate is documented and easy to miss. In July 2024 the city-state's aviation regulator rewrote its light-show rules: individual aircraft no longer required registration, and authorising a 1,000-drone display fell from S$25,120 to S$500 on first application, S$425 thereafter at the same site. A cheap, legal, high-volume civilian swarm practice existed before the military borrowed it. Purchasing transparency, meanwhile, cuts both ways. Tenders above S$6,000 are meant to appear on the GeBIZ portal, yet nothing LOCUST-specific — no award notice, no inventory declaration, no operational-capability statement — had surfaced by 17 July 2026. Direct sourcing for security-sensitive buys is expressly allowed, so the silence proves little either way.

What the record does establish is layering. An April 2026 demonstration at 6th Singapore Division put one operator over a reconnaissance swarm while machine vision converted video into targets and HIMARS launchers, fighter jets and AH-64D Apaches executed a simulated engagement; Prime Minister Lawrence Wong toured the formation's integrated training on 9 April. Already in the 2025 trials, a tactical mapping tool had stitched thousands of aerial frames into 3D terrain, rendering an area twenty-five times larger than before within an hour. And named foreign firms now sit inside the autonomy tier: Shield AI's Hivemind kit (February 2026), an algorithms agreement with Icarus Swarms (June 2025), vision-language navigation work with Mistral (June 2026). A CSIS brief dated 10 June 2026 observed that a first-person-view mission today still ties up three to six people per strike — which is precisely why orchestration, not airframes, is the prize.

Europe is running parallel experiments. ALTISS, finalised under the European Defence Fund in December 2025 by a consortium led by Magellium with Sagax, Radionor and M3S Belgium, targets single-operator swarm missions; NATO's DIANA accelerator lists coordinated unmanned swarms among its 2026 challenge priorities. The Singaporean difference is institutional rather than technical: the interface, the mapping and the integration authority stay inside a single agency, while the flying hardware is treated as consumable and swappable.

Three tests will tell whether the model is what it appears to be, and each one remains open on the public record:

— Can the 200-aircraft ceiling claimed by the developing agency be matched by formations flown in the field, when the disclosed evidence stops at fifty?

— If a purchase is eventually disclosed, will its value sit in the vehicles or in the control-and-fusion stack — and would it reach the open record at all, given the direct-sourcing carve-out?

— When Shield AI, Icarus Swarms and Mistral co-develop the autonomy tier itself, where exactly does the sovereign boundary run?