ByTAL PINKASOVICH

JULY 28, 2026 08:48

Updated: JULY 28, 2026 11:46

In May, a firm launched only two years earlier reached unicorn status. Kela, started in summer 2024, secured $200 million at a $1.2 billion valuation led by Stripes and D1, with Bill Ackman and Eric Schmidt participating, following the $100 million already raised from Sequoia, Lux Capital, and In-Q-Tel, the CIA investment arm. Roughly $20 million in revenue. About 150 staff. An operating system for the battlefield.

That same month, the global count of tech layoffs in 2026 surpassed 185,000, with over half the announcements citing AI as the cause.

Two labor markets share one economy. One informs engineers that machines now handle their tasks. The other cannot locate enough engineers to construct those machines.

The defense-tech narrative often appears as a stock quote. Kela reaches a billion. Anduril doubles valuation to $61 billion in under a year. Palantir boosts revenue 85 percent and integrates into the US Army next command system. The account holds true yet overlooks the layers beneath.

Defense tech is not a single firm or even one sector. It forms a stack of distinct technologies, from cutting-edge AI to fabricated metal, each layer an industry with unique economics and hiring needs. Consolidation occurs on one level only. All other levels are splitting, expanding, and recruiting.

The View from the Top Floor

The uppermost level draws all attention. Palantir reported $1.63 billion in first-quarter revenue and gained a key position in the data layer of the US Army next-generation command system. Anduril raised $5 billion in May at $61 billion, with secondary trades by mid-June suggesting near $100 billion. Kela expanded from one site to double-digit border positions in Israel within a year.

This platform layer sees real consolidation. One contender may become the battlefield equivalent of Android, with few rivals. Focus solely here and the outcome appears settled: the contest ends, large players prevail, and little remains to support or join.

The view misses the mark. Descend the levels, as each operates its own expansion.

One level below platforms lies compute. The defense semiconductor market stood at $13.4 billion in 2025 and is forecast to nearly triple in ten years. In December the Pentagon created a contracting mechanism with a $25 billion limit to accelerate US-made military chips into systems.

Below compute rests navigation and sensing, where quantum technology has moved beyond theory. SandboxAQ AQNav reads Earth magnetic fields for navigation without satellites and has recorded over 450 flight hours on four aircraft types toward a Pentagon quantum-sensing effort. GPS disruption in the eastern Mediterranean displaced 1,650 ships from routes in one week. Quantum navigation addresses the layer where GPS signals fail.

Below sensing lies firmware and edge autonomy, where a key 2026 military component costs under a pair of running shoes. The TFL-1 vision module from Ukrainian startup The Fourth Law, priced below $100, guides strike drones through the final half kilometer using machine vision alone, without signals or cables. It holds certification from Ukraine Defense Ministry and operates with about 20 brigades. It does not surpass million-dollar jammers but renders them ineffective over the last 500 meters.

At ground level sits bent metal. LUCAS, the American loitering munition first used in combat over Iran in January, costs $35,000 and was created by SpektreWorks, a 14-person Arizona firm now aiming for 10,000 units yearly. The Pentagon initial drone program this year purchased 30,000 drones at roughly $5,000 each from 12 separate vendors, not two large contractors.

One additional level exists, rarely shown: the base where items are produced. Software updates in seconds. Production advances at the pace of poured concrete.

Anduril invests a billion dollars in Arsenal-1, an automated factory in Ohio, recognizing the market will be settled not by elegant code but by capacity to produce 10,000 airframes yearly without motor or battery suppliers halting output. The rarest role in defense tech today is not a machine-learning researcher but a manufacturing engineer.

Software at the top, chips and sensors in the middle, firmware and hardware lower, and the production floor at the base. The top level consolidates. All below it expands.

Addressing the Bubble Concern Directly

The objection arises plainly: this resembles a bubble. A two-year-old firm valued at $1.2 billion on $20 million revenue, another at $61 billion on sales far below legacy primes, investors embracing defense as software returns slowed. In consumer markets these signs would warn of excess.

Bubbles burst when buyers shift, yet this buyer is the state, rearming on timelines driven by security concerns. At the Ankara summit on July 7, NATO approved over $40 billion across five years for counter-drone systems alone, as current costs favor cheap drones over expensive interceptors.

Israeli figures illustrate the point sharply. Defense exports hit $19.2 billion in 2025, up from $14.8 billion, marking a fifth straight record. Israeli startups collaborating with the Defense Ministry R&D unit raised more than $1 billion in 2025 versus about $150 million the prior year. This reflects a sector mobilized, not merely expanding.

Some valuations may exceed current results, especially at the top level where excess appears. When adjustment occurs it will stem from procurement timing rather than demand. A request for 300,000 drones remains a query, not a firm order.

The Pentagon fast-track office still requires about 19 months from request to first delivery, and many defense startups fail in that interval between demonstration and funded order. Losses will concentrate where pricing sits highest: at the top. That is not the core of the stack.

Entry Points Lie at Lower Levels

The top-focused view overlooks low barriers at deeper stack layers. A $100 module bypasses million-dollar jamming. Fourteen people created the drone used over Iran.

Since October 2023 more than 130 Israeli startups joined the war effort, roughly half in autonomy and AI, a quarter in sensors. Few required a fabrication plant, prime contract, or large funding round. They needed a layer, a scenario, and pace.

This presents dual opportunity. Investors find the top priced for success while lower levels remain valued as specialist areas. Engineers facing AI-related layoffs encounter even clearer contrast.

In 2018 thousands of Google staff opposed a Pentagon AI effort and the firm withdrew. That reflected prior industry consensus against weapons work. Seven years later the same sector cuts staff by the hundred thousand, and the fastest-growing technical employer in the West is the former taboo field.

Human life ranks as the costliest item in current conflict economics. The labor market tests this at scale. Talent and capital move toward valued areas.

The layoff notice and funding announcement represent the same facts viewed differently. One economy treats people as costs to reduce. The other sees them as essential capacity. The stack rises high, demand comes from states, and every level recruits.

The issue is not whether entry exists but how long it remains available.

Tal Pinkasovich serves as a defense technology analyst and reserve major in the Israeli navy, with over a decade in the Israeli startup sector. His coverage of drone operations and defense economics has appeared in The Jerusalem Post.