There was zero talk of drones around NVIDIA’s closely watched Q2 earnings report this week. The focus was, as usual, a Data Center story: that single segment brought in $89 billion of the total and the conversation stayed focused on hyperscaler AI spending.

But two days earlier, The New York Times had reported something considerably darker: Ukrainian forensic examiners had found NVIDIA Jetson Orin computers — commercial edge-AI modules that sell for under $1,000 — inside the wreckage of Russian Molniya drones, including one that struck a gas station in Zaporizhzhia on July 6 and killed three civilians.

Investigators concluded the drone had selected its final target autonomously, using onboard machine vision, without a live human operator in the loop.

NVIDIA does not sell into Russia and its robotics business did not merit a mention on the earnings call. But that is precisely the point. The same week that NVIDIA's edge-AI hardware surfaced in a war-crimes-adjacent investigation, the business built on that hardware was too small to rate a sentence on an earnings call that otherwise ran for more than an hour. There is a multi-billion-dollar NVIDIA business hiding in the shadow of the data-center headline and this week gave two very different reasons to finally look at it.

NVIDIA calls it "physical AI": the extension of its data-center AI stack into robots, autonomous vehicles and autonomous aircraft, built around a family of embedded computing modules called Jetson. It is, by any normal company's standard, a large and fast-growing business. But it’s nearly invisible inside NVIDIA's own disclosures. The company has never broken out drone revenue and it does not even give robotics its own line item — the closest official figure is a broader "Automotive and Robotics" category, which generated about $2.3 billion in fiscal 2026, up 39% year over year. 

That number lived inside an "Edge Computing" segment that totaled $7.2 billion in NVIDIA's most recent full quarter of disclosure — against $75.2 billion from Data Center in that same quarter and $89 billion this past quarter alone.

CEO Jensen Huang has been more forthcoming in his own commentary than the financial statements are: he has described the physical AI business as running at roughly a $10 billion annualized rate today, with a stated ambition to reach $100 billion within a decade. Take that at face value and NVIDIA is describing a business that would rank as a Fortune 500 company in its own right — one that gets less disclosure than segments a fraction of its size at other companies, simply because it sits next to the largest wealth-creation story in corporate history.

In the world of drones, there are five commonly accepted classifications, each with increasingly complicated navigation technologies – and increasing computational requirement.s 

Each tier adds a new class of processing on top of the last rather than replacing it. 

Level 0 needs nothing beyond a basic MCU (microcontroller unit) running motor and stabilization loops. 

Level 1 adds simple range sensors for spatial awareness in enclosed spaces, still handled by the same MCU-class processor. 

Level 2 introduces a full autopilot stack and GPS (Global Positioning System) or the broader GNSS (Global Navigation Satellite System, the umbrella term covering GPS, Galileo, GLONASS and others) receiver for waypoint navigation, but the aircraft still operates VLOS (visual line of sight), meaning a human pilot can see it and intervene, so no onboard perception is required. 

Level 3 is where compute jumps materially: the aircraft now needs a companion system-on-chip running camera-based sensor fusion and obstacle avoidance in real time, alongside the existing flight controller rather than instead of it. 

Level 4 pushes that same perception workload further, handling landing and extended operation with minimal human oversight, which demands a dedicated edge-AI chip built for continuous vision processing. 

Level 5 adds an NPU (neural processing unit, a chip optimized specifically for running trained AI models efficiently) for onboard inference plus a secure element for authenticating and coordinating with other aircraft, since full autonomy at this tier includes swarm behavior rather than a single vehicle acting alone. The overall pattern is that each level keeps the compute burden of the level below it and stacks a new, more demanding workload on top — from motor control, to sensing, to full navigation, to real-time vision, to coordinated multi-vehicle inference.

The war in Ukraine has rapidly pushed drone development and apparently, in the last few weeks or months, a leap to all-AI driven drones used by Russian forces. 

Both sides had relied on GPS-blockers and radio communications blockers to identify and disable drones as they closed in on targets. 

Think of it: a drone that navigates by GPS and flies on an analog or RF video link is, functionally, a remote peripheral — it doesn't need much silicon onboard because the thinking happens at the ground station or via satellite. Two things broke that model simultaneously.

1. Electronic warfare made GPS and radio links unreliable. In Ukraine, FPV drones routinely lose GNSS within 5–10 km of jamming, and spoofing defeats the navigation stack without even triggering a warning. A $50 ground jammer neutralizes an aircraft that depends on satellite signal. The fix isn't a better antenna — it's replacing GPS with visual inertial odometry, which fuses camera-based feature tracking with high-frequency IMU data to hold position without any external signal. That fusion has to run in real time, onboard, because there's no link left to lean on. 

2. Once the drone has to "see" rather than just receive coordinates, it needs a GPU-class chip, not a flight-controller MCU. “Semantic SLAM” (Simultaneous Localization and Mapping), object detection, terminal-guidance lock-on, obstacle avoidance — all of that is vision-model inference running continuously in flight, which is exactly the workload NVIDIA Jetson-class modules are built for. Quantum Systems' Vector drone is a concrete example: day-and-night cameras feeding an onboard NVIDIA Jetson Orin computer that processes imagery in real time specifically because GPS-denied conditions took the ground link out of the loop.

The second half of the shift is the same story outside defense. Delivery drones (Wing), inspection drones, and swarm platforms all face a version of the same problem even without an adversary jamming them — beyond-visual-line-of-sight operation means the aircraft can't depend on constant human oversight or constant connectivity either, so the same onboard-perception architecture applies. Nvidia's own framing for this is "physical AI" — silicon and software stack (Jetson, Isaac, Cosmos) sold once for humanoids, industrial robots, and drones alike because they share the same requirement: real-time perception and decision-making with no round trip to the cloud.

NVIDIA’s Jetson Orin modules pair a GPU with a CPU and additional accelerators in a compact form factor with the memory, power circuitry and I/O needed for sensor fusion — combining camera, lidar, radar and other feeds into a single real-time picture. 

NVIDIA has been aggressively seeding this robotics market, writ large, with a three part strategy. 

That combination — cheap, commercial, off-the-shelf and backed by a mature software ecosystem — is exactly what has made Jetson popular with legitimate robotics builders: Boston Dynamics, Amazon Robotics (AMZN: NASDAQ) and FANUC in industrial and humanoid robotics; agricultural and delivery-drone developers building toward a future of routine autonomous flight. 

NVIDIA backs it with independent validation, not just marketing: Jetson modules have posted results on MLPerf, the industry-standard machine learning benchmark suite, with NVIDIA citing performance gains of up to 83% on the same hardware within a single year through software optimization alone.

We estimate NVIDIA’s annual drone business at $18.8 m, but admittedly, our Low and High range are wide enough to fly a lot of drones through it. 

NVIDIA does not disclose Jetson sales, drone shipments or even robotics revenue, making the drone business impossible to measure directly. A bottom-up estimate nevertheless suggests that if manufacturers sell roughly one million commercial and industrial drones annually, 15% require Jetson-class onboard AI, NVIDIA wins one-quarter of those systems and earns approximately $500 per module, annual drone-related revenue would be about $19 million. 

An aggressive case—two million units, 30% requiring advanced AI, a 40% NVIDIA share and nearly $1,000 of compute per aircraft—would approach $240 million.

That is meaningful for an emerging product category but immaterial beside NVIDIA’s $96.2 billion quarter. It also clarifies what Jensen Huang’s description of physical AI as an approximately $10 billion business does—and does not—mean: that figure encompasses autonomous vehicles and the broader robotics stack, including development and simulation infrastructure, rather than Jetson modules installed in drones. Drones may be a fast-growing edge of that opportunity, but the available evidence does not support calling them a multibillion-dollar NVIDIA business today.

The same qualities that made Jetson a commercial success — it is inexpensive, requires no export license as a civilian development kit and is sold openly to students and startups worldwide — are what Ukrainian investigators say made it attractive to Russian drone and missile designers. Beyond the Molniya drone involved in the Zaporizhzhia strike, officials have also reported finding a Jetson module in a Russian S-71 "Monochrome" cruise missile. The circuit boards recovered were reportedly stamped "Made in China," and the transshipment path from NVIDIA's authorized sales channels to a Russian weapons program remains unclear. 

The company says Jetson Orin computers are "consumer-grade products sold to students, developers and startups for a wide range of beneficial applications," that they "were not designed for military applications," and that while NVIDIA does not sell into Russia, it "cannot track" what happens to hardware once it reaches secondary resale markets.

Western export controls have raised Russia’s costs and disrupted its access to military technology, but third-country procurement networks remain active. A 2025 U.S. Government Accountability Office review found that China accounted for nearly 80% of reported exports to Russia of Common High Priority List goods following the February 2022 invasion. These 50 categories include semiconductors and radio-frequency components found in Russian missiles and drones.

Russia’s imports of these goods have since fallen by more than half in value, yet China still supplies approximately three-quarters of the remaining trade, while Hong Kong’s share of processor and integrated-circuit shipments is rising, according to a July 2026 KSE Institute analysis. That month, the Council of the European Union tightened restrictions on additional entities in China, Hong Kong and other jurisdictions accused of helping Russia obtain microelectronics, semiconductor-processing equipment and other controlled technologies.

But it’s clearly still happening. 

It would be a mistake to read the Zaporizhzhia story as the whole story, because the legitimate market this hardware serves is enormous and growing largely independent of any single bad actor. 

Aircraft registrations, operating approvals and defense procurement provide a more concrete measure of demand than consultancy forecasts. The FAA counted 424,516 active nonrecreational small drones in the United States in 2025 and projects approximately 541,000 by 2030. AUVSI—the Association for Uncrewed Vehicle Systems International – also counted 920 active FAA waivers for flights beyond the operator’s visual line of sight in 2026. These longer-range operations increasingly require onboard perception, navigation and collision-avoidance computing—the type of workload Jetson handles.

Defense adds another source of demand. AUVSI’s fiscal-2025 budget analysis identified $10.1 billion requested for uncrewed-vehicle acquisition and development, about $1 billion more than the preceding year. That figure includes aerial, ground and maritime systems—not just drones—but demonstrates the investment flowing into autonomous machines. Only a fraction require Jetson-class hardware, and NVIDIA competes with Qualcomm, Ambarella and custom processors for that onboard-compute slot.

The global drone market is at about $80 billion, growing at a 30% annual clip. But there may be big jumps ahead. Military spending, including Ukraine's own $7.6 billion in first-half-2026 drone contracts and a Pentagon budget request seeking more than $70 billion for drone and counter-drone systems, remains the largest single driver. But commercial regulation is the next catalyst investors are watching: the FAA's proposed Part 108 rule would finally replace the current visual-line-of-sight requirement with a standardized framework for routine beyond-visual-line-of-sight flight and companies like Zipline — which has passed two million commercial drone deliveries and raised over $600 million to expand into new U.S. markets — are building ahead of a final rule expected in late 2026 or early 2027. Every one of those legitimate use cases needs the same kind of onboard compute NVIDIA sells.

NVIDIA's edge-AI and robotics business is real, growing quickly and large enough on its own terms to matter — Huang's own numbers put it on a path toward a scale most companies would headline, not footnote. This week showed both halves of what that means: an earnings report where the business could not compete for attention against a $96 billion quarter and a New York Times investigation showing that the same commodity hardware, once it leaves NVIDIA's control, can end up determining life-and-death decisions on a battlefield NVIDIA never intended to supply. Both are true at once and neither is going away as this business keeps growing in NVIDIA's shadow.

On Schwab Network with Marley Kaden and Sam Vardas, talking about Synopsis important quarter, predicting the great success the company is having with it’s newly acquired Ansys business.

WATCH THE SCHWAB HIT HERE →

On NewsNation with Connell McShane I dissected Nvidia (NVDA: NASDAQ) earnings and the impact of the NIMBY data center fight — in short, I don’t see one.

WATCH THE NEWSNATION HIT HERE →

In San Francisco all week and available for interviews, background or client calls — and as we get closer to IPO filings from Anthropic and OpenAI, we’ll be ready!.

Written reports are available to clients, our IPO reports for $5,000. Video summaries on YouTube, and of course our popular summaries of the summaries on Instagram, TikTok, and YouTube Shorts.

^{We certify that (1) the views expressed in this report accurately reflect our views about all of the subject companies and securities and (2) no part of my compensation was, is or will be directly related to the specific recommendations or views expressed in this report.}

^{Important disclosures}

^{Important disclosures are available by calling (347) 619-2489 or writing to Epistrophy Capital Research, One Ferry Building, Suite 201, San Francisco, CA 94105.}

^{Epistrophy Capital Research is an independent research provider and does not operate as a financial institution. Epistrophy explicitly does not provide investment advice, stock recommendations, or solicitations to buy or sell any securities.}

^{The research reports provided by Epistrophy Capital Research contain opinions derived from publicly available information, issuer communications, recognized statistical services, and other reputable sources considered reliable. However, Epistrophy does not independently verify the accuracy or completeness of such information and explicitly disclaims responsibility for any errors or omissions.}

^{Opinions and analysis contained within Epistrophy's research reports are current only at the time of publication and are subject to change without notice. Readers must independently verify facts and conduct their own due diligence before making investment decisions.}

^{Epistrophy Capital Research does not consider or evaluate individual investor circumstances, including investment objectives, financial situations, or risk tolerance. Investing in securities, particularly small-cap and micro-cap stocks, involves substantial risks, including significant volatility and potential loss of principal. Readers are strongly advised to consult their financial advisor or another qualified professional before acting on any information provided. Readers should assume Epistrophy Capital Research, its principals or its contributors may have positions, long or short, any of the companies discussed and the Epistrophy Capital Research principals or contributors may have had or currently have business interests in the companies discussed.}

^{Past performance referenced in Epistrophy reports is not indicative of future results. Security prices can fluctuate widely, and investors should be aware that investments can result in significant financial losses. Epistrophy Capital Research or its Unless explicitly stated otherwise, prices quoted in reports reflect market closing prices from the previous trading day.}

^{Epistrophy Capital Research publications are intended solely for direct recipients and should not be redistributed or shared with third parties without explicit permission from Epistrophy Capital Research LLC.}

^{Epistrophy Capital Research reports are provided strictly for informational purposes and do not constitute a comprehensive analysis of any company, security, or industry. No content within these reports should be considered accounting, tax, legal, or professional advice.}

^{Links or references to third-party websites or external resources are provided solely for informational convenience. Epistrophy Capital Research expressly disclaims endorsement of, and responsibility for, the content, accuracy, or reliability of such external information. Accessing third-party information is done entirely at the user's own risk.}

^{For additional details, clarification, or specific inquiries regarding Epistrophy Capital Research reports, please contact Epistrophy Capital Research LLC directly}