Shield AI raises $2 billion at a $12.7 billion valuation and buys Aechelon, the simulation house behind the Pentagon's combat test range

Shield AI, the San Diego maker of the Hivemind autonomy software and the V-BAT drone, said on March 26, 2026 that it is raising $2 billion and buying Aechelon Technology, whose software supplies the simulation inside the Pentagon's Joint Simulation Environment. The package comprises $1.5 billion of Series G equity at a $12.7 billion post-money valuation and $500 million of fixed-return preferred equity from funds managed by Blackstone, according to the company's release.

The new valuation is 2.4 times the $5.3 billion Shield AI disclosed when it raised $240 million in an F-1 round in March 2025, and 4.7 times the $2.7 billion attached to its $200 million Series F in October 2023. Blackstone has also committed a further $250 million delayed draw facility for future growth, which sits outside the $2 billion headline. Advent International leads the Series G, with the Strategic Investment Group of JPMorganChase's Security and Resiliency Initiative as co-lead. Existing investors Snowpoint Ventures, Innovation X Advisors, Riot Ventures, Disruptive and Apandion also took part.

Advent chairman David Mussafer joins Shield AI's board, and JPMorganChase's Todd Combs becomes a board observer. Neither side disclosed what Shield AI is paying for Aechelon. Sagewind Capital, the private equity firm selling the business, said in its own announcement that the deal is expected to close after customary regulatory approvals. Morgan Stanley was Shield AI's financial adviser, and J.P. Morgan Securities advised Aechelon.

Aechelon builds geo-specific flight simulators, physics-based sensor simulation and what it calls synthetic reality environments, used for pilot training, aircraft testing and the validation of autonomous systems, according to Sagewind. The company is the simulation technology provider for the Joint Simulation Environment, or JSE, the government environment the US military uses to test next-generation combat aircraft and autonomy against threats that cannot be replicated on an open range. After closing, Aechelon will run as an independent unit, with co-founder and chief executive Nacho Sanz-Pastor reporting to Shield AI chief executive Gary Steele.

Steele said the acquisition speeds up simulation work for Hivemind, including its support for the JSE, and advances what Shield AI calls its Hivemind Foundation Model for Defense, a domain-specific model trained in simulation and combined with real operational data. The company said the proceeds will pay for the acquisition, Hivemind's simulation work and the next development phases of X-BAT, its autonomous combat jet.

Founded in 2015, Shield AI says Hivemind now runs across 26 vehicle classes, from F-16 fighters and helicopters to drone boats and ground vehicles. On February 13, 2026 the company said the US Air Force had selected Hivemind for mission autonomy risk-reduction work on Anduril's YFQ-44A Fury, one of the Collaborative Combat Aircraft prototypes, with Collins Aerospace and RTX also supplying autonomy software in that phase. On February 26 it reported a first Hivemind flight test aboard the YFQ-44A over the Mojave Desert, completing all required test points. X-BAT, unveiled in October 2025, is a vertical takeoff and landing combat jet flown by Hivemind that Shield AI says can carry a full mission payload more than 2,000 nautical miles.

The round lands in a crowded field of venture-backed defence autonomy companies. Anduril raised $1.5 billion in a Series F in August 2024 to build its Arsenal-1 factory. Saronic, which makes autonomous surface vessels, raised $600 million at a $4 billion valuation in February 2025. In Europe, Munich-based Helsing raised 600 million euros in June 2025 in a round led by Prima Materia, the investment firm of its chairman Daniel Ek. Shield AI's position in the Fury programme is unusual: it is flying its software on a rival's airframe, a test of whether mission autonomy can be bought separately from the aircraft that carries it.

Simulation is central to that pitch because an AI pilot cannot be trained or cleared by flight hours alone. Reinforcement learning agents need millions of simulated engagements, and test authorities need repeatable scenarios with accurate radar, infrared and terrain. Aechelon's databases and sensor models give Hivemind a synthetic world close enough to the real one that behaviours learned there can carry over to the aircraft. It is the same sim-to-real problem civilian robotics companies face when they train humanoids and warehouse arms in software before deploying them.

The financing structure carries its own message. Fixed-return preferred equity pays its holders a set return ahead of common shareholders, which makes it more expensive than ordinary equity, and companies typically use it to raise cash without selling more common stock at the headline price. Shield AI did not disclose revenue, profit or headcount. The CCA work is a risk-reduction phase rather than a production award, Collins remains a competitor in the same phase, and X-BAT had not flown at the time of the announcement.

The next markers are regulatory clearance and closing of the Aechelon deal, further YFQ-44A flights and the Air Force's choice of mission autonomy for CCA production, X-BAT propulsion and first flight milestones, and whether Shield AI draws on Blackstone's $250 million facility.