Axon Vision 2025: The Backlog Swelled, but the Real Question Is How Much Turns Into Cash

Getting To Know The Company

Axon Vision looks stronger in the headline than in the day to day economics. Anyone stopping at NIS 29.3 million of binding orders in 2025, a NIS 15.3 million year-end backlog, and additional orders in January and March 2026 could easily conclude that commercialization is already proven. That is too fast a read. Revenue in 2025 grew only 9.6% to NIS 17.4 million, operating loss widened to NIS 8.8 million, and cash used in operating activity still stood at NIS 7.9 million. Orders are now arriving, but the conversion of those orders into revenue, cash, and operating leverage is still incomplete.

What is working now? The company is no longer at the stage where it only needs to prove technological relevance. Gross profit rose to NIS 8.9 million and gross margin improved to 51%, the non-Israel revenue share rose to 29% from 19% in 2024, and the business now has active commercial threads in Israel, Europe, Asia, and the United States. The balance sheet also looks very different: after the IPO, year-end cash and cash equivalents reached NIS 24.9 million, bank debt was fully repaid, and the company moved from financing stress to execution pressure.

What is still not clean? The sales engine expanded faster than revenue, trade receivables and unbilled revenue jumped to NIS 8.6 million from NIS 2.8 million, and some of the meaningful orders still rely on demos, prototypes, long milestone schedules, or credit offsets that are not equivalent to full cash consideration. That is exactly what a superficial reader can miss: Axon Vision’s backlog is real, but not every shekel of backlog is born equal.

This is also a market story with a practical actionability limit. As of April 6, 2026, market cap stood at roughly NIS 165 million, but daily trading volume in the stock that day was only about NIS 69 thousand. Short interest is negligible, so there is no aggressive market signal against the fundamentals. At the same time, this is not a stock with deep liquidity if the story slips.

The Economic Map

Axon Vision operates in one business line, developing, marketing, and selling AI systems for the battlefield. In practice, the economics rest on two product families:

The company operates with a flexible business model. Sometimes it sells a full system including hardware and software, sometimes software development against milestones, and sometimes software licenses. That matters because the quality of revenue changes with the contract type. A software license is usually cleaner revenue. A development, integration, or demo project can be strategically important, but it is not the same thing as a serial procurement order.

This chart makes one point quickly: Axon Vision is not only an armored-vehicle awareness story. As of 2025, the larger revenue engine already came from the aerial suite. That matters because it creates some diversification, but it also shows the company is not yet sitting on one dominant platform that defines the whole business.

The picture here is still mixed. International revenue rose to NIS 5.1 million, but 71% of revenue still came from Israel, and 37% of total revenue still came from the Ministry of Defense alone. In plain terms, the international story is now real, but not yet large enough to free the company from dependence on the domestic customer and its procurement cycles.

Events And Triggers

The first trigger: 2025 was the year the order book opened up. The company reported total orders of about NIS 30.1 million, including NIS 29.3 million of binding orders, versus only NIS 8 million in 2024. The sequence matters. In June 2025, it received an order of about NIS 5.5 million from the IDF AI and Autonomy Directorate for a situational awareness and perimeter protection project through 2026. In July 2025, it received an order of about USD 800 thousand from a European governmental entity for armored vehicle upgrades. In September 2025, it received an order of about USD 2 million from a domestic defense body to complete development and supply an upgrade kit. In October 2025, it received a software-license order of about NIS 2 million from a domestic defense company.

The second trigger: the post-balance-sheet flow of orders continued, but it also exposed backlog quality. On January 11, 2026, the company received an order of about USD 350 thousand under the Leonardo DRS cooperation framework for two demonstration systems intended for exercises and joint demonstrations in the United States. On March 16, 2026, it received another order from the IDF AI and Autonomy Directorate for about NIS 1.7 million, with delivery during 2026. These are positive signals because they show the business is no longer hanging on one isolated event. They also remind the reader that the story is still built around demonstrations, feasibility, and penetration, not broad serial procurement already locked in.

The third trigger: not every material order equals full cash. At the end of December 2025, the company received a USD 1.24 million order under its strategic agreement with Elbit, spread across 2026 through 2028. But USD 1.15 million of that amount is executed through a credit-offset mechanism rather than normal cash consideration, with only the remainder in cash. That does not erase the business value of the order. It does mean that reading the headline amount as though it is all straight cash collection is a mistake.

The fourth trigger: the August 2025 IPO changed the financing layer. In 2025 the company recorded net proceeds of NIS 37.547 million from the issuance of shares and warrants, repaid its bank loans by the end of September 2025, and reduced part of its related-party debt. That is a real inflection point. Before the IPO, the company lived on bank funding, shareholder loans, and a related-party facility from Razor. After the IPO, it got breathing room. The problem is that this breathing room bought time, not proof of commercialization.

The fifth trigger: at the start of 2026 the company clearly signaled that management now sees commercialization as the core challenge. Nery Tzin became CEO on February 15, 2026, alongside an approved equity package including 248 thousand options plus another 270 thousand options subject to shareholder approval. In January 2026, Ido Rozenberg, the company’s CTO and President, relocated to the United States as part of the expansion plan. This is not cosmetic. It is an admission that the central question for 2026 is no longer only engineering quality, but who turns the technology into larger commercial contracts in Europe and the US.

This chart shows why 2026 is the proof year. Roughly NIS 12.65 million of year-end backlog is slated for recognition in 2026, against full-year 2025 revenue of NIS 17.4 million. If delivery schedules hold and milestone billing follows, there is a real base for acceleration. If not, the issue will shift from backlog quantity to execution quality.

Efficiency, Profitability, And Competition

The critical 2025 number is not only that gross profit improved. The critical number is that the company is starting to look like a product company at the gross level, while still spending like a commercialization platform that has not finished penetrating its markets. Revenue rose to NIS 17.395 million from NIS 15.865 million. Gross profit rose to NIS 8.879 million from NIS 7.393 million, and gross margin improved to 51% from 46.6%. The company explains that the improvement came from better profitability in development projects and system sales, alongside first recognition of higher-margin software-license revenue. That is an important point. It means the economics of the product are starting to improve.

But this is also where the story flips. R&D expense rose to NIS 5.623 million, sales and marketing expense jumped to NIS 7.579 million, and G&A rose to NIS 4.512 million. Total net operating expenses reached NIS 17.714 million. In other words, all the gross improvement was swallowed by a higher cost base built in advance for future growth. That is why operating loss widened to NIS 8.835 million versus only NIS 2.310 million in 2024.

What Actually Improved

The chart shows that the problem is not at the gross layer. In fact, it is moving in the right direction. The strongest positive sign in the annual report is that the company no longer looks like a pure development shop billing hours. It is beginning to see software licenses and system sales at better profitability. Revenue mix also shows that the ground engine now sits alongside a meaningful aerial engine, so the business is not defined by a single platform.

Geography improved too. In 2025, 29% of revenue came from outside Israel versus 19% in 2024. Combined with the European, American, and Asian orders, this makes it clear that the company has already opened more than one commercial door. That is not a small achievement for a business that only became public in August 2025.

Why Operating Loss Widened

This is where the active bottleneck sits. Sales and marketing expense already reached 43.6% of revenue. That is not the spend profile of a company casually exploring opportunities. It is aggressive penetration investment. The headcount data supports that reading. At the end of 2025 the company had 34 employees and service providers, including 7 in sales and marketing, up from 4 at the end of 2024. Revenue per year-end employee was about NIS 0.5 million, down from about NIS 0.63 million at the end of 2024. That is not necessarily bad. It means the company is building ahead of revenue. But it also means the machine has already been loaded before all the revenue came through it.

There is another less visible point here. R&D is presented net. The company receives grants from the Israel Innovation Authority and from the Ministry of Economy, and those grants are offset against R&D expense. At the same time, some of the supported programs carry future royalties of 3% to 5% of sales until the grant is repaid with SOFR-linked terms. In other words, part of today’s expense relief can return as tomorrow’s economic cost if the products actually become sales lines.

Where The Moat Is Real, And Where It Is Still Thin

Axon Vision’s moat is not a portfolio of registered intangible assets. The company explicitly says it does not have registered ownership over intangible assets. Its moat sits elsewhere: access to operational data, live battlefield implementation, work with operational users, and the ability to run AI on constrained edge hardware. That is a real advantage, and the annual report also states that the company is already implementing systems on three types of armored platforms in the IDF, three armored platforms in NATO, four unmanned aerial vehicle types in the IDF, and one UAV model of a foreign army.

At the same time, the competitive field is not small. The company itself names Elbit, Rafael, Israel Aerospace Industries, large foreign players such as Lockheed Martin and Rheinmetall, and newer players such as Anduril. So technology alone is not enough. The company needs to prove that it can enter as a subsystem player, move through integrators, and close larger contracts without giving away pricing or getting trapped in endless demo cycles.

Cash Flow, Debt, And Capital Structure

In Axon Vision’s case, the right cash lens is all-in cash flexibility. The reason is simple. The company is at a stage where equity, debt, demo investment, and commercial expansion matter more than whether gross profit is moving in the right direction. So the key question is not how much cash the existing business can produce before growth uses, but how much was left after all the real cash uses of the year.

The Cash Picture

In 2025 the company used NIS 7.926 million in operating activity and another NIS 747 thousand in investing activity. At the same time, aside from the IPO, the financing picture was actually negative against debt and related parties. It received NIS 3.0 million from bank loans and NIS 1.923 million from related parties, but repaid NIS 4.499 million to banks, NIS 4.935 million to related parties, and paid NIS 252 thousand of interest. In other words, without the public-market raise, the year would not have ended with NIS 24.9 million of cash.

This is the cash story of 2025. The company bought itself time through equity, not through self-funded operating strength. That is not automatically negative. For an early commercialization business, it can be exactly the right move. But the distinction matters. The year-end cash pile is runway for the 2026 proof year, not proof that the commercialization model is already self-sustaining.

What Was Solved In The Balance Sheet, And What Was Not

The good news is that bank debt disappeared from the year-end balance sheet after all bank loans were repaid by September 2025. Related-party loans also came down, and equity moved from a deficit of NIS 3.834 million at the end of 2024 to positive equity of NIS 26.329 million at the end of 2025. In pure balance-sheet optics, that is a major change.

But the balance sheet did not become simple. First, receivables and unbilled revenue rose to NIS 8.585 million from NIS 2.790 million at the end of 2024, mainly because projects were completed or near completion but not yet billed under milestone terms. Second, net fixed assets rose to NIS 2.197 million, mainly because of demo kits. In other words, part of commercialization is already sitting inside capitalized commercial tools, not only in operating expense. Third, by year-end 2025 the company had entered IFRS 16 accounting for its expanded office lease and recognized a right-of-use asset of NIS 2.121 million and a lease liability of NIS 2.280 million.

There is also a more subtle financing signal in the annual report. The controlling shareholders committed to support the company through the end of 2026, but only after the company first exhausts the IPO proceeds, then its other cash balances, and only then its ability to obtain funding from third-party lenders. This is not a soft backstop. It is a clear statement that the market already provided the fuel, and management is expected to drive.

Outlook And Forward View

First finding: 2026 is a backlog-conversion proof year, not merely another order-announcement year. Out of the NIS 15.34 million backlog at the end of 2025, NIS 12.65 million is slated for recognition in 2026, NIS 2.049 million in 2027, and NIS 641 thousand in 2028. If schedules hold, acceleration is visible. If timing slips, the issue will move from backlog size to execution quality.

Second finding: the order book is broad, but not homogeneous in quality. Some orders are demo systems, others are milestone-based Ministry of Defense work, some include hardware and integration, and one of the most important orders relies mainly on a credit-offset mechanism. So the right way to read the backlog is not “how much was ordered,” but “how much gets recognized, how much gets collected, and how much returns as repeat business.”

Third finding: the company has already built the expense base for international penetration. The US entity, the relocation of the company’s President to the US, and the sharp rise in sales and marketing expense all show that 2026 now has to produce commercial output. The spend is already there. Revenue now needs to follow it.

Fourth finding: the management reset is not just a compensation event. The option package for the new CEO, including the additional 270 thousand options with performance conditions, is clearly trying to align the technology story with measurable commercial delivery. That also means the board is effectively acknowledging that the next test is commercial, not engineering.

What Must Happen In The Next 2 To 4 Quarters

The strategic direction is clear enough. The company is trying to move from a technology company proving capability to a defense-product company proving commercialization. For that to work, at least part of the 2025 and early 2026 project flow has to turn into follow-on business of higher quality. A Leonardo DRS demo order opens a door. It is not yet a serial supply route. An order from the IDF AI and Autonomy Directorate is a valuable signal. It is still not the same thing as recurring multi-year revenue visibility.

From an industry perspective, the backdrop is clearly supportive. The company describes rising defense budgets, stronger demand for edge AI on operational platforms, and growing openness to AI deployment on the battlefield. But this is exactly where discipline matters. A strong industry backdrop explains why orders started to arrive. It does not guarantee that this particular company will turn that backdrop into revenue, cash, and cleaner margins.

Risks

First Risk, Order Quality Is Not Always Cash Quality

The company has already shown that it can receive orders, but some of those orders depend on milestones, delivery terms, demos, or credit offsets. That is especially true for the Elbit-linked order received at the end of 2025, where USD 1.15 million out of USD 1.24 million is executed through a credit mechanism. If investors confuse economic value with accessible cash, they can overstate the current speed of commercialization.

Second Risk, Domestic Customer Concentration Runs Deeper Than The Percentage Line

37% of 2025 revenue came from the Ministry of Defense and another 10% from Elbit. But the risk is not only quantitative. The Ministry of Defense procurement framework gives the customer broad power, including the right to demand changes, increase quantity by up to 20% without changing price before the midpoint of the supply period, and cancel orders with compensation tied to reasonable direct costs and defined terms. So Ministry dependence is not only about customer concentration. It is also about a contract structure tilted toward the customer.

Third Risk, The Technological Advantage Is Not Yet Wrapped In Formal IP Protection

The company relies on distinctive technology, but it has no registered ownership over intangible assets. That means the moat today is built more on secrecy, contracts, operational data, customer relationships, and deployment know-how than on a formal registered IP layer. That can be enough at this stage, but it also makes the story more sensitive to employee turnover, replication risk, and faster competitive response from larger players.

Fourth Risk, Expansion Has Already Changed The Cost Base

The company enlarged its office footprint, moved into IFRS 16 lease accounting, scaled sales and marketing, built a US entity, and relocated its President to the United States. If 2026 does not deliver acceleration in revenue and collections, that fixed cost step-up will turn quickly from a growth support layer into a burden.

Fifth Risk, Market Liquidity Is Weak And Dilution Has Not Disappeared

The stock trades with low liquidity, and shallow liquidity absorbs bad news poorly. On top of that, 248 thousand options were already approved for the new CEO and another 270 thousand options are proposed subject to shareholder approval, alongside the public warrants issued in the IPO. The dilution alone is not a crisis point, but it is a reminder that equity is still an active financing tool in this story.

Conclusions

Axon Vision exits 2025 with a stronger balance sheet, broader commercial reach, and more proven technology. The central bottleneck has shifted from survival financing to converting orders into revenue and cash. In the near term, the market may focus on the order flow, the new CEO, and the possibility that 2026 marks a step change. Over a slightly wider horizon, the more important question is whether the 2025 backlog actually turns into cash flow rather than remaining only a documented promise.

Current thesis in one line: Axon Vision has already proved it has a product and a market, but 2026 will determine whether it can also turn that into repeat revenue, collections, and operating leverage.

What changed versus the older understanding of the company is fairly clear. The order book now exists, the IPO removed the immediate financing squeeze, and the company built a broader international commercialization platform. What remains unresolved is conversion quality. Orders still have not lined up with cash flow, receivables expanded sharply, and the go-to-market cost base already ran ahead.

The strongest counter-thesis is that this article may be too cautious. There is a fair case that 2025 was exactly the right transition year, with penetration spending coming before revenue. If the NIS 12.65 million slated for 2026 is recognized on time, and if Europe, the US, and Asia begin generating follow-on orders, 2025 may later look like the necessary bridge year rather than a warning year.

What can change the market’s reading over the short to medium term is not another cooperation headline, but a combination of three things: backlog recognition pace, receivables and collection behavior, and whether demo orders actually evolve into serial procurement. That matters because for a small listed defense technology company, the line between “proving capability” and “building a company” is measured less by the number of announcements and more by the quality of the cash that comes in.

Over the next 2 to 4 quarters, the thesis strengthens if 2025 and early 2026 orders begin to flow into recognized revenue at a reasonable pace, if collection starts to align with recognition, and if foreign growth engines begin to produce follow-on business rather than only pilots. It weakens if revenue remains slow, receivables keep running ahead of sales, or cooperation frameworks stay trapped in the demo stage.