# Shield AI: The Autonomy Company Building Toward an IPO, Not a Sale

Shield AI reached a 12.7 billion dollar valuation with an AI pilot called Hivemind, and is raising its way toward a public listing rather than a sale, the venture path to staying independent.

Author: J.A. Watte
Published: July 11, 2026
Source: https://jwatte.com/blog/shield-ai-hivemind-autonomy/

---

This is one profile in a set on cleared government contractors and the different ways the people who build them turn engineering work into real wealth. Most of the set is about selling a company to a larger buyer, the pattern laid out in [Built to Be Bought](/blog/built-to-be-bought/). A few companies are built to do the opposite, and Shield AI is the second one in this series, after [Anduril](/blog/anduril-built-to-stay-independent/), built to stay independent on venture capital. It makes the artificial intelligence that flies aircraft with no pilot and no satellite link, it has raised its way to a valuation of 12.7 billion dollars, and it is aimed at an eventual public listing rather than a sale. It is also, as the strangest fact in this piece will show, becoming a supplier to the very rival it is racing.

## Two brothers and an AI pilot

Shield AI was founded in 2015 in San Diego by two brothers, Brandon and Ryan Tseng, along with Andrew Reiter. The mission came out of Brandon Tseng's service. He spent roughly seven years in the Navy, including as a SEAL, and the company's founding story ties its purpose to the reconnaissance gaps that get people killed, the problem of not knowing what is inside a building or over the next ridge, according to [Shield AI's own account](https://shield.ai/about/). Ryan Tseng, who ran the company as chief executive for its first decade, had earlier founded a wireless-charging startup that Qualcomm bought.

The product they built is called Hivemind, an artificial-intelligence pilot that can fly aircraft in places where GPS is jammed and no human is in the loop, now used across more than two dozen types of vehicle. The best known of those vehicles is the V-BAT, a vertical-takeoff surveillance drone that needs no runway. Shield AI filled out its line the way the companies in this series usually do, by buying. It acquired Heron Systems in 2021, the team whose AI had [beaten a human F-16 pilot five to nothing in a DARPA simulated dogfight](https://www.darpa.mil/news/2020/alphadogfight-trials-final), and Martin UAV the same year, the company that actually made the V-BAT, followed by an Australian sensor firm in 2024 and a simulation-software company in 2026.

## Raising its way to 12.7 billion dollars

Like Anduril, Shield AI is funded by venture and growth capital rather than a private equity buyout, and its valuation has climbed in long strides. It [raised a round at a 2.3 billion dollar valuation in 2022](https://shield.ai/shield-ai-raises-165m-series-e-to-accelerate-building-of-the-worlds-best-ai-pilot/), then [200 million dollars at 2.7 billion in October 2023](https://shield.ai/shield-ai-raises-200m-reaching-2-7b-valuation/), then [240 million dollars at 5.3 billion in March 2025](https://shield.ai/shield-ai-raises-240m-at-5-3b-valuation-to-scale-hivemind-enterprise-an-ai-powered-autonomy-developer-platform/) in a round whose strategic leads included the prime contractor L3Harris. Then, in March 2026, it raised about [2 billion dollars at a 12.7 billion dollar valuation](https://techcrunch.com/2026/03/26/defense-startup-shield-ai-lands-12-7b-valuation-up-140-after-u-s-air-force-deal/), more than doubling its worth in a year. Across its life it has taken in an estimated 3.5 billion dollars.

The composition of that last round is the important part, more than the headline number. It was co-led by the private equity crossover firm Advent International and by JPMorgan Chase's strategic investment arm, with 500 million dollars of preferred equity from Blackstone. That is the textbook late-stage capital stack a company assembles in the run-up to going public, not the kind that precedes a sale. Shield AI has not filed to go public and has named no date, so this is a direction rather than an announcement. But the direction is unmistakable, and it is the same one Anduril is traveling: raise enough private money, at a high enough valuation, to reach the public markets on your own terms instead of being absorbed by a larger prime.

## The tell: its software flies on its rival's jet

Here is the fact that captures Shield AI's odd position better than any valuation. In February 2026 the Air Force named the two companies it would rely on for the autonomy, the actual AI pilot, on its Collaborative Combat Aircraft program, the effort to field drone fighters that fly alongside crewed jets. The two companies were Shield AI and Anduril, [with Shield AI's Hivemind selected as a mission-autonomy provider](https://shield.ai/shield-ai-selected-as-mission-autonomy-provider-for-the-u-s-air-force-collaborative-combat-aircraft-program/). And Hivemind is now flying system-level tests aboard Anduril's own Fury fighter airframe. So the two most prominent independents in defense are at once each other's fiercest rivals and, on this program, supplier and platform. Shield AI is raising billions to stay independent of Anduril while its software becomes part of Anduril's airplane.

The other tell is who Shield AI hired to run it. In 2025 the founder chief executive, Ryan Tseng, stepped aside after a decade, and the company brought in [Gary Steele as chief executive](https://shield.ai/shield-ai-appoints-gary-steele-as-ceo-ryan-tseng-named-president/). Steele's entire prior track record is scaling software companies to enormous sales, Splunk to Cisco for 28 billion dollars and Proofpoint before it. A company hires that kind of operator when it intends to get very large and reach a public market, not when it intends to quietly sell itself. Brandon Tseng remains as president, and the founders kept their hands on the mission while handing the scaling to a specialist.

## What it builds, and where it flies

The contracts are real and increasingly large. The Coast Guard awarded Shield AI a [five-year deal worth up to about 198 million dollars](https://shield.ai/shield-ais-v-bat-selected-for-198-million-contract-to-provide-u-s-coast-guard-with-maritime-unmanned-aircraft-system-services/) for V-BAT surveillance services, and the Navy [selected it to compete for an 800 million dollar surveillance-services vehicle](https://shield.ai/shield-ai-selected-by-u-s-navy-to-compete-for-800m-in-isr-services-with-v-bat/). As with every contractor in this series, it is worth keeping the language honest: a vehicle a company is selected to compete on is a hunting license, not guaranteed revenue.

The most consequential proving ground has been Ukraine, where the V-BAT has flown since 2024. In one declassified mission a V-BAT flew deep into occupied territory under heavy jamming and [helped locate a Russian surface-to-air missile system](https://en.defence-ua.com/weapon_and_tech/new_v_bat_drones_in_ukraine_shield_ai_declassifies_their_use_in_the_unmanned_systems_forces-13639.html) that was then destroyed. More than half of the company's business in 2025 was international, spanning the Netherlands, Poland, India, Romania, and others. Its revenue is not audited or disclosed, because it is private, but figures reported through the business press put it around 300 million dollars in 2025, short of an internal target near 400 million, with the company projecting well over 500 million in 2026. Those numbers are estimates relayed through the press, not filed accounts, and belong in that category.

## The honest note

This series does not airbrush, and Shield AI has a genuine controversy on its record that a fair profile has to carry. Its V-BAT drone, whose propeller sits where people work, has been involved in a series of injuries. During a Navy demonstration in Texas in 2024 a service member's fingers were partially amputated by a V-BAT propeller, and in 2026 a Romanian Navy officer was injured in a similar incident during training. A [June 2026 Reuters investigation](https://www.carriermanagement.com/news/2026/06/08/288825.htm) reported, citing unnamed sources, more than 50 crashes across roughly 200 drones in the company's fleet over 18 months, a figure that traces to a single investigative chain and that Shield AI disputes. The company's response has been that the injuries stemmed from violations of established safety procedures rather than a product defect, and that operational mishaps are common for this class of aircraft. Separately, a former employee filed a lawsuit in 2026 alleging harassment and retaliation after raising safety concerns; Shield AI has said the claims lack merit and is contesting them, and the allegations are unproven. These belong in the record. A company moving this fast, putting spinning hardware into the hands of sailors, is going to be measured in part by how it handles exactly this.

## The ledger reading

Strip away the drones and Shield AI is a wager on the same idea as the rest of this series, held in an unusually pure form. The founders and employees own equity in a private company now valued at 12.7 billion dollars, and almost none of that wealth is real yet. It is paper, marked to the last funding round, and it turns into money only if the company reaches a public listing or is sold, and the whole strategy is built around the first outcome rather than the second. That is the venture bet in its starkest version. You take shares instead of a bigger salary, you decline the cash exit when a prime would happily buy you, and you carry the risk that the mark-to-market number on your equity is a mirage until the day the public actually prices it.

It is the same lesson underneath [The W-2 Trap](https://thew2trap.com), pushed to its edge. A paycheck is a claim on your hours, and hours do not compound, so the people building real wealth here took equity in the asset instead. What Shield AI adds is the reminder that owning the asset and cashing the asset are two different events, sometimes years apart, and that staying independent means living in the gap between them. Anduril is making the same bet with more money and a longer lead. Shield AI is making it while its software rides inside Anduril's jet, which is about the most vivid picture of this strange, crowded, well-funded corner of the map that you could ask for.

## Related reading

- [Built to Be Bought](/blog/built-to-be-bought/): the sell-and-roll playbook these venture-funded companies are refusing to run.
- [Anduril: The Defense Company Built Not to Be Bought](/blog/anduril-built-to-stay-independent/): the other independent in this race, its rival and, on the Air Force's drone-fighter program, its platform.
- [Sierra Nevada Corporation: The Family-Owned Prime That Refused to Sell](/blog/sierra-nevada-family-owned-defense/): staying independent the old way, through a family balance sheet instead of venture rounds.
- [York Space Systems: From a Billion-Dollar Buyout to a $4.75 Billion IPO](/blog/york-space-systems-ae-industrial/): what the public-listing exit actually looks like when a defense company reaches it.
- [Government Contract Vehicles, Explained](/blog/government-contract-vehicles-explained/): the surveillance-services vehicles and IDIQs a company like Shield AI competes on.

## Fact-check notes and sources

- **Founded in 2015 in San Diego by brothers Brandon and Ryan Tseng and Andrew Reiter, with Brandon Tseng's Navy SEAL service behind the mission, and Hivemind as the core AI-pilot product:** [Shield AI](https://shield.ai/about/) and [Contrary Research](https://research.contrary.com/company/shield-ai).
- **Acquisitions of Heron Systems (2021, the DARPA AlphaDogfight winner) and Martin UAV (2021, maker of the V-BAT), plus later purchases of an Australian sensor firm and a simulation company:** [DARPA on the AlphaDogfight Trials](https://www.darpa.mil/news/2020/alphadogfight-trials-final) and [Shield AI on the Heron acquisition](https://shield.ai/shield-ai-acquires-heron-systems/).
- **Funding trajectory: a 2.3 billion dollar valuation in 2022, 200 million dollars at 2.7 billion in October 2023, 240 million dollars at 5.3 billion in March 2025, and about 2 billion dollars at 12.7 billion in March 2026; roughly 3.5 billion dollars raised in all (an estimate):** [Shield AI, 2022](https://shield.ai/shield-ai-raises-165m-series-e-to-accelerate-building-of-the-worlds-best-ai-pilot/), [Shield AI, 2023](https://shield.ai/shield-ai-raises-200m-reaching-2-7b-valuation/), [Shield AI, 2025](https://shield.ai/shield-ai-raises-240m-at-5-3b-valuation-to-scale-hivemind-enterprise-an-ai-powered-autonomy-developer-platform/), and [TechCrunch, 2026](https://techcrunch.com/2026/03/26/defense-startup-shield-ai-lands-12-7b-valuation-up-140-after-u-s-air-force-deal/). The Series G was co-led by Advent International and JPMorgan's strategic investment group, with preferred equity from Blackstone. No registration statement to go public had been filed as of this writing.
- **Selected in February 2026 as one of two mission-autonomy providers, with Anduril, for the Air Force Collaborative Combat Aircraft program, with Hivemind flying on Anduril's Fury airframe:** [Shield AI](https://shield.ai/shield-ai-selected-as-mission-autonomy-provider-for-the-u-s-air-force-collaborative-combat-aircraft-program/). **Gary Steele (formerly of Splunk and Proofpoint) became chief executive in 2025 as founder Ryan Tseng stepped aside and Brandon Tseng stayed on as president:** [Shield AI](https://shield.ai/shield-ai-appoints-gary-steele-as-ceo-ryan-tseng-named-president/).
- **The Coast Guard contract worth up to about 198 million dollars and the Navy vehicle Shield AI was selected to compete for, worth up to 800 million dollars:** [Shield AI on the Coast Guard award](https://shield.ai/shield-ais-v-bat-selected-for-198-million-contract-to-provide-u-s-coast-guard-with-maritime-unmanned-aircraft-system-services/) and [Shield AI on the Navy vehicle](https://shield.ai/shield-ai-selected-by-u-s-navy-to-compete-for-800m-in-isr-services-with-v-bat/). Vehicle ceilings are not guaranteed revenue. **The V-BAT's use in Ukraine, including locating a Russian air-defense system:** [Defense Express](https://en.defence-ua.com/weapon_and_tech/new_v_bat_drones_in_ukraine_shield_ai_declassifies_their_use_in_the_unmanned_systems_forces-13639.html). Revenue figures near 300 million dollars for 2025 are press estimates, not audited disclosures.
- **The V-BAT propeller injuries (a 2024 Navy demonstration in Texas and a 2026 Romanian training incident), the June 2026 Reuters investigation reporting more than 50 crashes across the fleet, Shield AI's response that incidents stemmed from safety-procedure violations rather than defects, and a 2026 employee lawsuit the company is contesting:** [Reuters, via Carrier Management](https://www.carriermanagement.com/news/2026/06/08/288825.htm). The aggregate crash figure traces to a single investigative chain; the lawsuit's allegations are unproven. Both are presented here as contested.

*This post is informational and journalistic, describing a publicly reported company, its people, and public records. It is not investment, tax, legal, or M&A advice, and nothing here is a recommendation to buy or sell any security. All parties are discussed from public records and their own published statements as nominative fair use, with no affiliation implied and nothing endorsed by them.*


---

Canonical HTML: https://jwatte.com/blog/shield-ai-hivemind-autonomy/
RSS: https://jwatte.com/feed.xml
JSON Feed: https://jwatte.com/feed.json
Hero image: https://jwatte.com/images/shield-ai-hivemind-autonomy.webp