Shield AI to acquire software simulation company Aechelon and raise $2B at $12.7B valuation
When the Pentagon needs to know how a new aircraft handles a specific combat scenario, it does not send a pilot into the sky. It sends the pilot into a simulator built by a company called Aechelon Technology. Aechelon creates the visual and sensor simulation environments that the U.S. military uses to train aircrews before they climb into a real cockpit. Shield AI just bought it.
The defense autonomy startup announced March 26 that it is acquiring Aechelon and raising a total of $2 billion: $1.5 billion in a Series G led by Advent International, with JPMorgan Chase Strategic Investment Group co-leading, at a $12.7 billion post-money valuation — plus $500 million in fixed-return preferred equity from Blackstone. The two instruments are distinct; the equity round and the preferred financing are different animals, even though they landed in the same press release. The company is not yet profitable.
The number is $2 billion. The actual story is what $2 billion buys.
That story starts with the Joint Simulation Environment — the Pentagon primary training platform where aircrews rehearse missions before they fly them for real. Aechelon builds the visual and sensor simulation that feeds JSE. Shield AI now owns that access. The Pentagon most sophisticated training environment is now aligned with the company accumulating the most operational flight data from actual combat: V-BATs running missions over Ukraine, YFQ-44As in CCA flight testing, L3Harris electronic warfare demonstrations run through hardware-in-the-loop simulation. If Shield AI can feed that real-world data back into the same simulation environment the military uses for training, it can train Hivemind on scenarios that no other AI pilot stack has ever encountered. That is the mechanism. That is the flywheel.
The flywheel is Shield AI Hivemind software: an AI pilot stack that has now flown 26 classes of vehicles, from F-16s and jet-powered UAVs to helicopters, drone boats, and ground vehicles, according to the company announcement. Hivemind was selected by the U.S. Air Force as a mission autonomy provider for the Collaborative Combat Aircraft (CCA) program — the service flagship effort to pair human controllers with autonomous wingmen. It is currently conducting flight tests aboard the Anduril YFQ-44A Fury, the CCA platform that Anduril moved into serial production at its Arsenal-1 factory in Ohio on March 23, three months ahead of schedule, the company said.
That selection matters. The CCA program is one of the most watched autonomy contracts in the U.S. defense ecosystem — Anduril and General Atomics are the two prime contractors building the aircraft themselves, and Shield AI is competing for the software layer that makes them autonomous. Winning mission autonomy provider status is a credential, not a deployment guarantee. The gap between selection and fielding is where defense tech stories routinely get ahead of themselves.
Shield AI existing hardware footprint gives the deal commercial logic beyond the software angle. The company V-BAT drone flew more than 200 missions in Ukraine in 2025 and completed over 130 more by March 2026, according to Brandon Tseng, Shield AI president and a former Navy SEAL, speaking to Business Insider. The drone is also the subject of a $198 million U.S. Coast Guard contract — the company says a single V-BAT runs approximately $1 million, against an $80 million MQ-9B SeaGuardian that the Coast Guard previously evaluated. Whether those cost claims hold up at scale, and whether the V-BAT operational record in Ukraine translates to U.S. military acceptance, are questions the Coast Guard contract will begin to answer.
The X-BAT tells a different story — one of ambition and timeline. Shield AI autonomous VTOL fighter jet is powered by a GE F110-GE-129 turbofan producing approximately 29,000 pounds of thrust. The engine fired for the first time on March 18, 2026, according to The Defense News. VTOL flight testing is planned for 2026, with initial operational capability targeted for 2028. That IOC date is a plan, not a fact. The defense industry has a long history of IOC dates that slip. Readers should treat 2028 as an ambition until the aircraft is actually flying.
There is a more immediate data point on Hivemind capabilities. In March 2026, Shield AI and L3Harris announced what they described as a breakthrough in autonomous electronic warfare: a Hivemind-powered drone that detected and responded to electromagnetic threats in real time during a hardware-in-the-loop simulation, without human intervention. The demonstration involved multiple unmanned aircraft systems feeding threat intelligence into a common operating picture. Whether this translates to actual combat conditions — where the electromagnetic environment is dirtier, more adversarial, and less controllable — is unknown. The word autonomously in a demo announcement is doing a lot of work there.
Shield AI is also training Ukrainian Armed Forces Unmanned Systems Forces on V-BAT operations, as part of a partnership with Ukraine Brave1 defense technology cluster, according to the Ukrainian Pravda. Ukraine has become the world most active test environment for autonomous military systems, and any company accumulating operational data from that theater has a significant advantage over competitors flying only test ranges in California or Ohio.
The Aechelon acquisition closes the loop between that real-world data and the simulation environment where it can be turned into training scenarios. Aechelon supports the Pentagon Joint Simulation Environment — the primary training platform where aircrews rehearse before they fly for real. If Shield AI can feed actual mission data from V-BATs in Ukraine and YFQ-44As in CCA testing back into the same simulation environment the military uses for pilot training, it can train Hivemind on scenarios that no other AI pilot stack has ever encountered. That the pitch. The question is whether the integration actually works, whether the military permits it, and whether any of this translates to a profitable business — because nine funding rounds in, Shield AI is still raising capital and not yet cash-flow positive.
Gary Steele, a former Cisco executive who took over as CEO, is the person tasked with building the commercial infrastructure around what Ryan and Brandon Tseng built technically. Brandon Tseng, the former Navy SEAL who co-founded the company with his brother Ryan, remains president. The Tseng brothers military background is not cosmetic — it is the product. Shield AI entire value proposition rests on the assumption that pilots who have actually flown combat missions understand what autonomous systems need to do in the air. Whether that instinct scales to a company managing $12.7 billion in valuation is the question Steele has to answer.
The funding terms — equity plus preferred from two different firms — reflect the stage Shield AI is at. It is big enough to need large checks and sophisticated structures, but still raising dilutive capital after nine rounds. That is not unusual for defense tech at this valuation, but it is worth noting: a $12.7 billion valuation on a company with no disclosed profitability, competing in a program (CCA) that has not yet produced a fielded system, is a bet on the timeline, not a reflection of current revenue.
Brandon Tseng, when asked about the valuation on a recent call with investors, noted that the company had grown from a $5.3 billion post-money valuation in March 2025 — more than doubling in twelve months. That growth is real. What it costs to sustain, and when (or whether) it produces actual profits, is the question the next funding round will answer.
The Aechelon acquisition is expected to close after customary regulatory approvals, with no specific close date specified in the company announcement.