Drones—The booming "Physical AI," Rare Earth Elements Still Key

Wall Street Insights

The global drone market is expected to double to $40 billion in five years, with Barclays predicting it will expand to $250 billion by 2035. The real constraints of this revolution have gone beyond defense budgets, shifting towards AI computing power, energy grids, and critical minerals. China controls the primary supply of 52 out of 40 critical minerals, with rare earths concentration exceeding 90%. Beyond military demand, agricultural spraying has reduced costs by 70%, inventory efficiency in warehouses has improved by 50%, and drones can deliver in 19 minutes, pushing civilian applications deeper into this industrial transformation.

Drones are evolving from “battlefield tools” to a broader industrial chain.

According to Wind Trading Station, Zornitsa Todorova, a Barclays thematic FICC research analyst, stated in a recent report that the global drone market size has doubled from about $20 billion (in 2020) to over $40 billion (in 2025), with projections indicating it will reach $50 billion by 2026 and $250 billion by 2035.

Barclays pointed out the “true constraints” of this business that are often overlooked: “The constraints increasingly extend beyond the scope of defense budgets, reflected in capital expenditures for artificial intelligence, energy, and critical minerals.” The price of individual drones is indeed decreasing—disposable drones are commonly priced between $20,000 and $50,000—but the real costs and bottlenecks are increasingly shifting to computing power, electricity supply, and critical minerals, with rare earths remaining crucial.

Barclays views drones as a typical form of “AI landing in the physical world”: the expansion of defense budgets is just one line, while the other three lines—AI capital expenditures, energy and grids, and critical minerals—are linking defense capabilities to broader investment cycles, inflation, and industrial policies.

Furthermore, military demand acts as an accelerator, but it may not be the endpoint. Barclays expects that by 2035, the civilian share will rise from about 55% to approximately 65%, with agriculture, inventory checks, and last-mile delivery becoming new growth battlegrounds: drone spraying can reduce costs by about 70%, shorten operational time by over 90%, and decrease water usage by about 90%; in warehouses, inventory processing time can be reduced by 50%; in delivery, there are already cases of scaled operations with an average delivery time of 19 minutes.

Drones are entering a phase of “explosion in patents and revenue.”

Barclays defines “takeoff” with two sets of data: market revenue expansion and a surge in innovation density. The number of global drone patent grants has grown from fewer than 200 in 2014 to nearly 8,000 by 2024 (a 45-fold increase), mainly due to defense R&D investments and rapid commercial applications.

Estimating the market size itself is not easy—many companies do not disclose revenue by product line—Barclays’ approach is to aggregate disclosures from companies, industry research, demonstration materials, and other multi-source estimates, noting that predictions for 2024-2025 are highly concentrated, with a standard deviation of about $5 billion.

In a broader framework, Barclays categorizes drones as part of “Physical AI”: embodied intelligence encompasses humanoid robots, autonomous vehicles, advanced industrial automation, and drones, collectively viewed as heading toward trillion-dollar opportunities by 2035. Among these, drones are positioned as the second-largest segment, estimated between $150 billion and $350 billion, with a baseline scenario of $250 billion; autonomous vehicles are the largest segment (approximately $250 billion to $550 billion).

Growth is primarily a defense story: low-cost, disposable, swarm becoming the new standard.

The report’s conclusion is straightforward: defense demand is the dominant growth driver of the drone market, with related revenue accounting for about 40%-50%, contributing “up to half” of recent growth. It uses battlefield data to illustrate how “scalability” is becoming the norm: Ukraine’s drone production jumped from approximately 800,000 units in 2023 to nearly 5 million by 2025, with around 2 million being FPV small systems; during the initial weeks following the outbreak of conflict in the Middle East in 2026, recorded drone strikes approached 2,000.

In defense scenarios, Barclays roughly categorizes drones into two ends:

• High-value specialized platforms: unit prices can reach millions of dollars, with long endurance, high performance, and multi-tasking capabilities.

• Low-cost, highly scalable platforms: primarily short-range, limited endurance, and common disposable tasks, typically priced below $50,000.

The key change is that the ability of low-cost platforms to scale “more, faster, denser” increasingly relies on AI transferring capabilities from hardware to software—navigation, obstacle avoidance, and collaboration, allowing swarm actions to shift from “labor-intensive” to “model-driven.”

Cheap individual units do not equate to cheap systems: defense spending is beginning to “shift to computing power and software.”

The report emphasizes a structural migration: Physical AI is shifting defense spending from traditional “platforms and personnel” to R&D, computing power, data, and software, while also showing a tendency to reduce costs in operations and organization.

This places defense systems at the intersection of “four accounts”: defense budgets + AI capital expenditure + energy + minerals, with the latter three becoming new constraints.

Defense budgets themselves are still on the rise: SIPRI data shows that global military spending will reach $2.7 trillion in 2024; UN scenario modeling suggests that if trends continue, it could reach $3.5 trillion by 2030 and exceed $4.4 trillion by 2035; in a more aggressive scenario where military spending accounts for 5% of global GDP, it could reach $6.6 trillion by 2035.

Structurally, NATO member countries’ spending on “equipment and related R&D” has increased from 24% in 2014 to 30% in 2025 (with total NATO spending around $1.5 trillion in 2025), which Barclays views as an early signal of “shifting towards infrastructure and equipment.”

The report also mentions a surge in electricity demand, with the IEA estimating that data centers have consumed about 1.5%-2.0% of global electricity; by 2035, this proportion is expected to rise to 4.4%, approximately 1,600 terawatt-hours. More troubling are local constraints rather than national averages: AI data centers have electricity consumption intensity close to heavy industries like aluminum smelting, but demand is highly concentrated. For example, in the U.S., nearly half of data center capacity is concentrated in five regions; Barclays believes the issue is not whether “national power generation is sufficient,” but whether “electricity can be reliably delivered to the nodes where demand is emerging,” such as Northern Virginia (PJM), Texas, and parts of the U.S. Midwest.

Critical minerals are not just a cost item, but the boundaries of strategic autonomy.

The drone hardware stack’s dependence on critical minerals runs throughout the entire chain:

• Propulsion systems: high-performance motors require rare earths (such as Pr, Nd, Sm, Dy) as well as nickel and copper.

• Structural framework: aluminum, titanium, magnesium, and tantalum are used for lightweighting and strength.

• Communication and navigation: beryllium, gallium, germanium, and indium are used in fiber optics, sensors, radar, and imaging.

The report identifies 52 critical minerals, with China being the dominant supplier for 40 of them. The concentration of rare earths is even more extreme: rare earths include 17 elements, of which 14 are classified as critical minerals, with China’s supply share for these 14 exceeding 90%.

“Decentralization” of the supply chain is progressing, but Barclays’ conclusion is somewhat cold: in the next five years, non-China regions plan to add new rare earth mining capacity of over 50,000 tons/year, separation capacity of over 40,000 tons/year, and magnetic materials/alloys capacity of at least 70,000 tons/year, but truly reaching FID projects remain few; rare earth mining from exploration to commercialization often takes 10-20 years, with especially difficult gaps in mid-to-lower stream technology and experience.

After military incubation, civilian applications will complete the story.

The report suggests that the “ultimate demand” for drones is based on productivity improvement rather than simply defense cycles. It outlines three categories of implementation along the path of “military first—cost reduction—civilian diffusion”:

• Agriculture: By the end of 2024, DJI estimates its global active inventory of agricultural drones to be about 400,000 units (up from about 80,000 in 2020); if extrapolated at the same growth rate, the agricultural drone fleet may approach 3.5 million units by 2035. Cases in spraying scenarios show: compared to manual backpack spraying, costs can be reduced by about 70%, and compared to tractor spraying, costs can be reduced by about 50%; time is shortened by over 90%, and water usage is reduced by about 90%.

• Inventory checks: After introducing Verity autonomous drones at their “Velocity” warehouse in Kentucky, UPS Supply Chain Solutions saw inventory processing times drop by 50% within a few months.

• Last-mile delivery: Since partnering with Walmart in 2023, Alphabet’s Wing has achieved thousands of orders weekly, with an average delivery time of under 19 minutes, connecting 18 Walmart Supercenters in a “hub-spoke” model.

		Sina Statement: This information is reproduced from Sina's partner media, and Sina's publication of this article is intended to convey more information and does not imply approval of its views or confirmation of its descriptions. The content of the article is for reference only and does not constitute investment advice. Investors act on this basis at their own risk.

A wealth of information and precise interpretations can be found in the Sina Finance APP.

Editor: Ling Chen