F-47 vs F/A-XX: The AI Stakes Behind Fighter Funding

Congress just drew a bright line in the FY2026 defense policy bill: about $2.6B for the Air Force’s F-47 and about $74M for the Navy’s F/A-XX—a gap so wide it doesn’t just shape airframes, it shapes the pace and direction of AI integration in future combat aviation.

This matters in the “AI in Defense & National Security” series because next-gen fighters aren’t only aircraft. They’re flying compute nodes—platforms that fuse sensors, run onboard AI for targeting and electronic warfare, manage autonomy with collaborative drones, and plug into joint battle networks. When one service gets funded to sprint and another is told to jog in place, the U.S. risks building two different futures: one where AI-enabled air dominance matures quickly, and another where carrier aviation fights the next decade with a partial upgrade path.

I’m not arguing the Air Force shouldn’t move fast. I’m arguing that treating F/A-XX as “optional” is a strategic bet—and the AI/software realities of modern combat make that bet riskier than it looks on a budget table.

What Congress actually funded—and why it signals an AI priority

The direct answer: Congress is fully backing one sixth-generation fighter track (F-47) while keeping the other (F/A-XX) barely warm, citing industrial-base capacity and schedule risk.

The compromise FY2026 National Defense Authorization Act (NDAA) language described in the source article funds:

• ~$2.6B for the Air Force’s F-47
• ~$74M for the Navy’s F/A-XX development request
• $377M identified for a special access program associated with Navy next-gen fighter efforts (referred to as “Link Plumeria”)

Lawmakers also demanded more clarity on F-47: a report covering costs, schedule, and funding needs through 2034, plus basing, construction, training, and Guard/Reserve integration, due March 1, 2027.

The industrial base argument is real—but incomplete

The Pentagon’s stance (as described in the article) boils down to: the industrial base can only go fast on one sixth-gen program right now.

That’s plausible. The bottlenecks aren’t just final assembly lines. They’re:

• Mission-systems software talent (AI/ML engineers, safety engineers, systems integrators)
• Test infrastructure (ranges, digital testbeds, EW chambers)
• Advanced components (radar modules, processors, power/thermal management)
• Certification and airworthiness pipelines for autonomy-adjacent capabilities

But here’s the missing piece: AI-heavy aircraft programs don’t behave like traditional aircraft programs. The constraint isn’t only metal-bending capacity—it’s software throughput. If the U.S. concentrates funding and engineering momentum into one service’s ecosystem, the other service’s AI stack can fall behind even if the airframe eventually arrives.

Why next-gen fighters are really AI platforms (and why that changes the risk)

The direct answer: sixth-generation advantage is largely software-defined—AI-enabled sensor fusion, electronic warfare, and autonomy management matter as much as stealth and kinematics.

For decision-makers who don’t live in aviation minutiae, it helps to think of next-gen fighters as three systems:

1. The air vehicle (range, signature, propulsion, payload)
2. The mission system (sensors, EW, comms, data fusion)
3. The autonomy layer (human-machine teaming, collaborative aircraft control, decision aids)

Funding decisions that slow F/A-XX don’t just delay a jet. They slow down a software factory—and in modern defense, software factories create compounding advantages.

AI in the cockpit isn’t about replacing pilots

The most useful way to frame AI-enabled mission systems is decision compression: AI helps crews and commanders move from “I have data” to “I have an option I trust” faster.

In a contested environment—especially maritime—AI contributes to:

• Multi-sensor fusion (finding weak signals across radar, IR, EW, and offboard sources)
• Track management at scale (maintaining identity and intent across dense air/missile pictures)
• Electronic warfare orchestration (adaptive jamming, emissions control, deception)
• Weapons-employment decision aids (probability-of-kill estimation under uncertainty)
• Collaborative autonomy control for uncrewed teammates

If one program matures these capabilities faster, it can set the de facto standards for how joint airpower “does AI”—interfaces, data models, tactics, test methods, even training pipelines.

The Navy’s F/A-XX problem isn’t “a smaller F-47”—it’s a different AI fight

The direct answer: carrier aviation imposes unique constraints that make Navy AI integration a separate problem, not a derivative of the Air Force’s.

Some observers assume the Air Force can build the “real” sixth-gen fighter and the Navy can later adapt. That’s a comforting myth.

Carrier air wings operate in a different geometry:

• Shorter reaction times from sea-based positioning
• Higher reliance on organic EW and strike packages when land basing is politically constrained
• Different comms realities due to maritime dispersal and emissions control
• Sortie generation and maintenance cycles that punish fragile software/hardware integration

Maritime autonomy has to survive comms denial

Collaborative autonomy (manned-unmanned teaming) is widely discussed, but the sea fight raises the bar: you often can’t assume stable, high-bandwidth, always-on links.

That pushes autonomy design toward:

• Edge AI that runs onboard with intermittent updates
• Graceful degradation (mission continues when models or links fail)
• Policy-based autonomy (clear constraints and ROE baked into behaviors)
• Robust blue-force identification in cluttered maritime environments

If F/A-XX development stays underfunded, the Navy risks fielding a future air wing where the uncrewed pieces advance (because they’re cheaper and faster) but the crewed quarterback platform lags in AI integration and battle management.

The hidden consequence: AI fragmentation across the joint force

The direct answer: a one-program sprint can create two incompatible AI ecosystems—different data standards, toolchains, safety cases, and operator workflows—making joint operations harder and more expensive.

When one service gets the money and schedule priority, it tends to also get the power to define:

• Data schemas and labeling conventions
• Model validation and test protocols
• Interface standards for sensors, weapons, and autonomy controllers
• Human-machine interface patterns that pilots train on for years

If the Navy later tries to catch up, it may be forced into an awkward choice:

1. Adopt the Air Force’s stack even if it’s a poor fit for carrier operations, or
2. Build/maintain a separate stack, doubling sustainment and complicating interoperability

Neither is great. The joint force wins when the plumbing is shared and the tactics are tailored.

AI can help procurement decisions—but only if it’s used honestly

One bridge point from this campaign is often overlooked: AI can optimize resource allocation decisions in defense procurement, but only if leaders are willing to treat the outputs as decision support, not political cover.

Practical applications that can be implemented now (even before F/A-XX is fully funded) include:

• Industrial-base digital twins to model supplier constraints and workforce pipelines
• Schedule risk forecasting using historical program data and real-time supplier signals
• Portfolio optimization to compare marginal dollars: propulsion vs EW vs autonomy test
• Test-range allocation models to reduce queue time for autonomy and EW trials

These tools won’t eliminate hard tradeoffs, but they can expose when a “one program at a time” policy is truly necessary—and when it’s a convenient simplification.

What smart leaders should do in 2026: five moves that protect AI readiness

The direct answer: even if F/A-XX remains paced slowly, the Navy (and OSD/Congress) can protect AI progress by funding the right layers: software, data, and test infrastructure.

Here are five concrete steps I’d push for in 2026 planning and contracting cycles.

1) Fund the mission-system stack as a reusable product line

Treat mission systems like a product portfolio—shared components, shared pipelines, frequent releases.

• Build a common data backbone for sensors, EW, and autonomy interfaces
• Require modular open systems that actually ship with documentation and test harnesses

2) Buy test capacity like it’s a weapon system

Autonomy and EW are test-hungry. Delays often come from test scarcity, not design.

• Expand digital testbeds (hardware-in-the-loop, RF-in-the-loop)
• Prioritize model evaluation at the edge (thermal/power constraints matter)

3) Make “model updates under constraints” a requirement, not a feature

AI in operational aircraft must handle:

• disconnected operations
• patch windows that are rare
• cyber and supply-chain risk

So the requirement should read like: safe performance under degraded comms and stale models.

4) Standardize safety cases for autonomy

Certification for autonomy-adjacent features will bottleneck both services.

• Define common assurance patterns (fault detection, bounded behaviors, explainability thresholds)
• Use shared red-team evaluation and adversarial testing methods

5) Put joint interoperability into the contract incentives

If the Air Force is sprinting, fine—just don’t let it sprint into a cul-de-sac.

• Incentivize shared interfaces for collaborative aircraft control
• Require cross-service demos (carrier-relevant scenarios included)

Snippet-worthy reality: If autonomy can’t interoperate across services, it isn’t autonomy—it’s a boutique feature.

What to watch next: signals that F/A-XX is slipping (or recovering)

The direct answer: watch appropriations, classified add-ons, and whether software/test funding is protected even when airframe money isn’t.

Over the next few months, the most meaningful indicators won’t be glossy renderings. They’ll be:

• Whether appropriators align closer to the House and Senate marks discussed publicly (hundreds of millions to over a billion) versus the minimal request
• Whether Navy next-gen fighter funding shifts into classified lines (harder to track, but often visible via program “activity” signals)
• Whether the Navy issues software-focused solicitations and testbed expansions (a strong sign the AI layer is being protected)
• Whether joint standards emerge for collaborative autonomy and mission-system interfaces

Where this leaves the joint AI fight

The direct answer: fully funding F-47 while underfunding F/A-XX accelerates one AI aviation ecosystem and risks leaving carrier aviation with a slower, more fragile path to autonomy and decision advantage.

If you care about AI in defense and national security, this isn’t a niche Navy-Air Force rivalry. It’s a question about how the U.S. builds scalable, interoperable AI for high-end conflict—the data pipelines, the safety cases, the test capacity, and the operator trust.

There’s still time to correct course without pretending budgets are infinite. The practical move is to stop thinking in terms of “two jets” and start thinking in terms of two mission-system factories. If one factory gets shut down for a few years, you don’t restart it by writing a check later—you restart it by rebuilding people, processes, and confidence.

If your organization is working on mission autonomy, ISR fusion, electronic warfare AI, or AI-ready networks for contested environments, now’s the moment to pressure-test your approach: Can your system run at the edge, degrade safely, and interoperate across services—especially in maritime conditions?