Razor 2025: The backlog is there, but the test shifts to ARR, cash and customer mix

Company Overview

Razor no longer looks like a small AI company selling mostly a story. By the end of 2025 it looks more like an industrial software company with a real commercial proof point in mining, and above all with one very large global customer. Its main product, DataMind AI, is built for predictive maintenance and equipment optimization. In practice, though, it is not sold as a clean, frictionless SaaS product. It comes as a broader package that includes software, hardware, installation and implementation.

That is exactly what a superficial first read can miss. The headline looks like a software scaling story. Revenue doubled, backlog is large, gross margin improved, and the new truck product has already produced a framework order of up to $78 million. Under that headline, however, there is still a clear gap between the scale of the wins and the economics already sitting inside the recurring business: ARR was only NIS 4.2 million, while 96% of 2025 revenue came from one customer.

What is working right now is the Land & Expand model. The company managed to take a mining contract signed in 2023, expand it in 2024 into the truck use case, launch the new truck product in November 2025, and sign a much larger multi year framework order a month later for heavy trucks and HME fleets. That is a real commercial achievement. What still blocks a cleaner thesis is conversion. Razor still has to prove that these wins become active licenses, ARR, ongoing collections and broader customer mix, rather than staying mostly as backlog, advances and installations.

That is also why the story matters now. There is no near term covenant or bank debt squeeze here. The company still has positive working capital of NIS 16.3 million, and management says it has enough cash for at least the next 12 months. But there is another actionability constraint: in 2025 cash burned quickly, equity turned negative, and the stock still trades with very weak liquidity, with the latest daily turnover at roughly NIS 77 thousand. Even if the operating story improves, this remains a small cap where execution has to be very sharp to change the market reading.

The quick economic map looks like this:

These charts make the core point immediately. Razor clearly knows how to generate revenue. It has not yet shown the same scale in the recurring layer that normally defines a cleaner software story. That is why the next year will be judged less by whether there are more orders, and more by what portion of the orders is actually active and renewable.

Events and Triggers

The mining engine still carries the story

The first large contract with the global customer, signed in late 2023, remained the anchor of the story in 2025. It covers licenses for 14 mining sites over 5 years, together with hardware and implementation, for total potential value of up to $31.2 million. As of the report date, five sites were active, six more licenses had been approved and were being installed, and four of those were delayed because of customer side constraints. That matters because it shows that demand is real, but timing is still not fully under Razor's control.

This leads directly to the first key insight of 2025: the global customer relationship is still expanding, but on the customer's timetable. The company says it is still working to complete all 14 sites by the end of the third contract year, while already in discussions about four more sites. What looks from the outside like one large signed contract is, in practice, an ongoing engine of approvals, installations and go lives.

Trucks moved from proof of concept to the next growth engine

In May 2024 Razor received another order from the same global customer, worth $7.4 million, to adapt DataMind AI to heavy trucks. That order included development work, a 9 month license period and connection of up to 25 trucks across three mining sites. By the report date the company had completed adaptation work on three truck models at two sites, while the third site had not yet started and management expected it to begin only in the second half of 2026. So the company has proved the use case, but not yet completed every milestone.

The more important step came at the end of 2025. The truck product was launched in November, and on December 31 the company signed a new framework order for software licenses covering the customer's vehicle fleets, with total potential value of up to $78 million. This is a very material deal, but it should not be read as a clean, locked in $78 million commitment. Under the order terms, after roughly seven months the customer can reduce the scope to a core subscription worth $17.5 million over 5 years for up to 200 vehicles. In addition, after three years the customer can reduce the full contract from 8 years to 5 years, taking the final total down to about $50 million.

That is not a legal footnote. It is the core economic read. The order proves that the customer believes in the new truck product. It does not yet prove that the full $78 million will flow into the business unchanged.

Smaller triggers that still matter

In December 2025 Razor received two more orders from new customers for commercial truck pilots. Those customer sites are in Australia and Spain, and the company says that, to the best of its knowledge, both customers have additional sites that could support further expansion. This does not change the 2025 financial picture because the current pilot amounts are not material. It does improve the quality of the story, because it is the first sign that the truck product may move outside the existing global customer relationship.

Another event worth noting is the governance change. On February 7, 2026 Raz Roditi ceased serving as chairman, while remaining CEO and a director. Formally, that slightly improves the separation between management and board oversight. Practically, founder concentration remains high.

Efficiency, Profitability and Competition

The growth is real, but it is still not clean SaaS growth

Revenue rose 101% to NIS 34.6 million. The second half of 2025 was also materially larger than the second half of 2024, at NIS 19.0 million versus NIS 12.2 million. Gross margin improved to 45% from 41%, mainly because of higher licensing revenue from DataMind AI together with complementary hardware sales and implementation work.

This is where the read has to slow down. ARR at the end of 2025 was only NIS 4.2 million, up from NIS 3.3 million at the end of 2024. That is 27% growth, not 101%. So the right economic conclusion is that revenue surged mainly because Razor is monetizing large contracts through hardware, implementations and staged project recognition, while the recurring layer is still relatively small. That is not a technology problem. It is simply a reminder that the model has not yet matured into a business living mostly off renewals.

Expenses still absorb almost all of the improvement

Gross profit rose to NIS 15.7 million, but R&D expense was NIS 17.6 million, sales and marketing was NIS 7.3 million, and G&A was NIS 7.3 million. As a result, loss from ordinary operations only improved to NIS 15.0 million from NIS 17.1 million in 2024. Even after doubling revenue, the company is still not close to operating break even.

That is not necessarily a negative on its own. Razor itself frames 2025 as a year in which it increased R&D spending in order to complete the truck product and close the larger license deal. If true, that can be read as an investment in market entry. But the report also forces the second read: for now the company is still producing growth at the cost of a very heavy expense layer.

What is more interesting is that one of the lines that helped the operating result did not come from the core product. Other operating income, net, was NIS 1.5 million. That includes sublease related income, and also a NIS 1.37 million gain tied to the decision not to extend part of the office lease. In other words, a meaningful part of the operating improvement came from office and lease adjustments, not just from stronger software economics.

Competition is less urgent right now than concentration

The company argues that the market it serves is still not crowded with specialized technology competitors, and that DataMind AI holds an advantage because it combines a broad set of data sources, including video and sensors, to diagnose the root cause of failures rather than just symptoms. That may well be true, but in 2025 competition is not the main issue. Concentration is.

The global customer accounted for 96% of 2025 revenue, up from 80% in 2024. Customer B, Elbit Systems, dropped from 14% of revenue to just 2%. So Razor did not become more diversified as it grew. It became less diversified. The same pattern shows up in backlog: NIS 106.3 million out of NIS 107.5 million came from the global customer relationship.

From a tech plus services perspective, that is the number that matters most. Not because the customer is weak, but because almost all commercial proof still sits inside one account, one sector and one expansion path. As long as that continues, the company can keep showing growth. As long as it does not spread, the market is still likely to apply a discount to the quality of that growth.

Cash Flow, Debt and Capital Structure

The right way to read Razor is through all-in cash flexibility. Not because the company is out of cash in the near term, but because the core question here is how much cash is left after real uses, not how much theoretical revenue still sits in backlog.

On that basis, 2025 was a heavy year. Cash and cash equivalents fell from NIS 55.3 million to NIS 35.2 million. Cash used in operating activity was NIS 16.9 million, versus only NIS 308 thousand in 2024. Investing activity was also negative at NIS 1.1 million, and financing activity negative at NIS 1.4 million. The total result was a NIS 20.1 million decline in cash and cash equivalents, plus another NIS 705 thousand from FX effects.

The critical point is not only the size of the burn, but its source. In 2024 the company benefited from a sharp increase in contract liabilities, meaning advances received up front that supported cash flow. In 2025 the direction reversed: contract liabilities fell by NIS 3.1 million, while the company recognized NIS 19.6 million of revenue during the year from amounts that had previously sat in contract liabilities. In plain English, Razor consumed part of the advance funding it had built up earlier.

At the same time, the company still has no financial debt to banks or lenders, which is an important distinction. The closest thing to debt is the lease liability, which fell to NIS 6.2 million from NIS 18.3 million. But that decline should be read carefully. It is not mainly the result of a suddenly much stronger financing profile. It came primarily from the decision not to extend part of the office lease, the reduction of right of use assets, and the expansion of subleasing.

Equity tells the same story. By the end of 2025 Razor had moved into an equity deficit of NIS 3.0 million, versus positive equity of NIS 11.2 million a year earlier. On the other hand, working capital remained positive at NIS 16.3 million, and the company states that based on its budget and 2026 to 2027 cash flow forecast, together with the existing backlog, it has enough cash for at least the next 12 months. So this is not an immediate liquidity crisis. It is a business that still needs to prove it can convert large wins into a steadier cash profile.

Outlook

Before looking at 2026, four findings need to stay on the table:

• First: NIS 107.5 million of backlog is not the same thing as a large recurring base. ARR of NIS 4.2 million makes clear that current economics still rely far more on implementations, hardware and project timing.
• Second: the $78 million truck framework order is a dramatic market validation, but its hard economic floor is far lower because the customer can step down to $17.5 million and later to $50 million.
• Third: concentration did not fall with growth. Razor ended 2025 with 96% of revenue and almost all backlog tied to the same customer.
• Fourth: 2025 was not yet a clean breakout year. It was a capability building year. The company invested heavily in development, cash burn was sharp, and the next proof point has to come through activations, collections and diversification.

The right word for 2026 is proof year. Not a quiet bridge year, because the company already has product, customer, backlog and a major framework order. But not a clean breakout year either, because too much still depends on one customer, on milestones and on how that customer chooses to deploy.

On one hand, management is targeting ARR of about NIS 18 million by the end of 2026, which implies growth of more than 300% from the end of 2025. If the company comes close to that, the market reading could change materially. On the other hand, management also says direct R&D investment in 2026 is expected to be about NIS 18.85 million. So the coming year is not supposed to be cheap.

The economic implication is clear. Razor does not only need continued growth. It needs better quality growth. For that to happen, several things have to move at the same time:

1. The 2025 truck order has to start showing up in the reports as active licenses and collections, not just as a large future headline.
2. The four delayed mining installations at the global customer need to be completed during 2026.
3. The third site in the 2024 truck program needs to finally start, otherwise the proof cycle will keep stretching.
4. New pilots need to convert into commercial agreements, otherwise concentration will remain intact even if revenue grows.

What can change the market interpretation in the short to medium term is less the next revenue line on its own, and more the quality of what sits beneath it. The market will want to see activations, visible ARR progression, evidence that expected collections are indeed coming in, and a first sign that the truck product can work beyond the existing customer relationship.

Risks

Customer concentration that is more than a percentage

Razor's main risk remains dependence on one customer. Not only because 96% of revenue came from that account, but also because most backlog, most commercial proof and most future expansion still sit inside the same relationship. Any delay in rollouts, any shift in deployment pace, or any use of the reduction rights in the truck framework order could affect the company materially.

Cash quality

The second risk is the gap between the income statement and the cash picture. 2025 showed that revenue can double while operating cash flow remains deeply negative at NIS 16.9 million. That happens because the business is still not built mainly on software renewals. It also depends on installations, advances, hardware and rollout timing at the customer. As long as that remains the model, a strong report can still look weaker once the analysis shifts to the cash balance.

FX, supply chain and hiring

The company has NIS 31.1 million of US dollar denominated monetary assets against NIS 2.86 million of liabilities, and it estimates that a 5% move in the shekel dollar rate affects finance income or expense by about NIS 1.4 million. That is not existential, but at this scale it is meaningful. In addition, the report itself flags difficulty hiring quality personnel and availability of electronic components in the supply chain as material risks. For a company that sells both software and hardware enabled deployments, that is not background noise.

Governance and dependence on the founders

Razor itself ranks dependence on a major customer and on its unique technology as high impact risks, and founder dependence as medium. In practice, the reliance on Raz Roditi and Michael Zolotov remains deep. The February 2026 governance change, in which Roditi ceased serving as chairman while remaining CEO and director, improves the formal structure but does not really remove key person risk. At the same time, around the Inspection AI activity with a defense customer there is an early stage document disclosure request connected to a possible derivative action, and the company still says it cannot assess the scope of any impact.

Conclusions

Razor ends 2025 with stronger commercial proof than it has ever had. What supports the thesis right now is a global customer still expanding usage, a truck product that has already reached a large framework order, and a gross margin that has started to improve. What blocks a cleaner thesis is the gap between large backlog and small ARR, between doubled revenue and negative operating cash flow, and between one major customer win and a broader customer base. In the short to medium term, the market will mainly focus on whether 2026 delivers activations and collections, or just more timing slippage inside the same customer.

Current thesis in one line: Razor has already proved that it has a product a large global mining customer wants, but it has not yet proved that it can turn that into a recurring, diversified and cash generative software business.

What changed versus the older version of the story is that the debate is no longer whether there is demand. The debate has moved to what part of that demand is actually sticky, recurring and cash converting.

The strongest counter thesis is that this read is too conservative, because the company already has NIS 107.5 million of backlog, a target of NIS 18 million ARR by the end of 2026, enough cash for the next 12 months according to management, and no bank debt creating immediate pressure.

What could change the market reading in the short to medium term would be confirmation in the next reports that the 2025 truck order is moving into activation and cash collection, further progress on the delayed mining site rollouts, and a first material commercial agreement with a new customer outside the existing relationship.

Why this matters is simple: in a small industrial technology company, the first big win is not the end of the story. It is the beginning. What ultimately drives the value of the business is not only whether the product works for one customer, but whether the model starts to replicate without consuming all of the cash along the way.

What has to happen over the next 2 to 4 quarters for the thesis to strengthen is visible ARR acceleration, actual collection from the 2025 order, and pilot conversion into commercial agreements. What would weaken it is customer use of the reduction mechanisms, further rollout delays, or another year in which backlog remains large while all-in cash flexibility keeps shrinking.

---