Lockheed Martin puts an AI “agent” through live air intercepts, reviving investor debate over the company’s valuation

Lockheed Martin’s Skunk Works unit and the U.S. Air Force Test Pilot School have become the center of a new round of investor chatter after reports that the teams used an onboard artificial intelligence agent to carry out 27 sensor-driven air intercepts.

According to the market report, the exercise involved an AI system that made the decisions needed to execute intercepts using live sensor inputs, rather than relying entirely on ground-directed control during the runs. The key takeaway for defense technology watchers is that the demonstration was framed as an “onboard” capability, meaning the decision-making was carried out in-flight rather than after-the-fact analysis.

The article ties the demonstration to Lockheed Martin’s broader push to apply AI to defense missions, while also noting the role of the Air Force Test Pilot School in running flight-test-style validation. That matters because the Air Force’s test community is often viewed as a bridge between lab prototypes and fieldable systems.

Beyond the technology, the report turns to markets, asking whether Lockheed Martin’s stock is “still cheap.” That framing indicates that some investors see the AI test as incremental proof points for long-term demand, while others remain focused on near-term earnings power and how quickly programs convert demonstrations into contracted work. The article itself does not provide new valuation metrics, so readers are left with the question more than a specific answer.

Lockheed Martin did not accompany the report with additional public details in the material provided here, such as the aircraft types, the specific sensors used, the environment or threat models, or performance metrics beyond the count of intercepts. Those omissions limit what can be concluded about readiness level, integration complexity, and how closely the exercise matches operational scenarios.

For Lockheed, the practical business relevance of demonstrations like this is that they can shorten the gap between prototype and program office interest. In defense contracting, that pathway typically depends on requirements from customers such as the U.S. Air Force, integration into existing mission systems, and evidence that an approach works safely under realistic constraints.

The defense sector context is that AI is increasingly treated as a tool for autonomy, sensing, and faster decision cycles, especially where latency and communications constraints exist. But the market often distinguishes between technology that is promising in test conditions and technology that is mature enough for scalable procurement. Without more disclosure, it is not possible to determine where this particular effort sits on that spectrum.

What to watch next is whether Lockheed Martin and the Air Force provide additional transparency on the follow-on steps, including any transition plans, program milestones, or repeatable test results that would indicate the work is moving from a one-off demonstration toward something that can be implemented across platforms. Investors and defense technology stakeholders will likely look for more concrete reporting on contracts or customer evaluations rather than only headline test counts.