Alarum 2025: AI Accelerated Revenue, but Cash Got Stuck and Margins Thinned

Getting to Know the Company

Alarum no longer looks like just another small proxy vendor. In 2025 it showed that it can sell a broader web-data stack: proxy access, data collection tools, and ready-to-use datasets. That is the part that is clearly working now. Revenue rose to $40.8 million, and the company proved it can move deeper into the value chain for customers that need web data at scale, including a large customer building foundational AI models.

But that is also where the problem starts. A superficial read may focus on the growth line and miss that the legacy engine, IPPN, actually declined from $30.3 million to $29.2 million, while almost all of the growth came from the newer layers. At the same time, gross margin fell from 75% to 58%, net revenue retention, or NRR, dropped to 0.83, and trade receivables jumped to $11.8 million. The company added volume, but it has not yet shown that this new volume is equally attractive for shareholders.

That makes the current bottleneck quite clear. Alarum’s bottleneck is not demand. It is demand quality: how much of that demand is recurring, how concentrated it is, and how much of it turns into cash rather than just reported revenue.

There is also a practical screen constraint. On the local TASE line, the last trading day showed only about NIS 56 thousand in turnover. That does not change the business, but it does affect how cleanly the local market can absorb good or bad news.

The geographic revenue map also reinforces the sense of a fast-moving mix shift. China and Hong Kong together contributed about $16.9 million, or more than 40% of total revenue, while revenue from the United Arab Emirates dropped sharply to $0.8 million from $6.5 million a year earlier. That is not enough on its own to build a sweeping geographic thesis, but it is another sign that 2025 was not a smooth organic expansion year. Demand moved around quickly.

In short, Alarum ends 2025 as a more interesting company, but not a cleaner one. That combination deserves attention, but it also demands much more proof than the top-line growth headline suggests.

Events and Triggers

A sharp shift from proxy access to higher-layer data products

The first trigger: the revenue mix changed materially. Data collection solutions and datasets contributed $11.1 million in 2025, versus just $0.6 million a year earlier. That matters because it means Alarum is no longer relying only on the breadth of its IP network. It is trying to sell higher-value layers on top of that network.

The other side of that move is less comfortable. The legacy IPPN engine declined by $1.1 million, so the newer products were not just incremental growth. They also had to offset weakness in the older core. This is not simply a case of a second engine being added on top of a healthy base. It is a case of a new engine being asked to replace part of the old one.

New demand is strong, but the older customer base is weakening

The second trigger: management explicitly describes a meaningful shift in customer mix, with strong growth from strategic customers in the AI vertical offset by declines elsewhere. That shows up clearly in NRR, which fell from 1.27 at the end of 2024 to 1.13 in the first quarter of 2025, then to 0.98, 0.92 and finally 0.83 by year-end.

That point is central. NRR below 1 means the older customer cohort, excluding new wins, is no longer expanding. It is shrinking. So 2025 growth is not evidence of broad-based deepening across the installed base. It is mostly evidence of the company’s ability to win large new accounts.

Customer concentration reinforces the same reading. Six customers purchasing more than $1 million each generated about 49% of web data collection revenue, and two customers alone represented 15% and 12% of total revenue. This is not theoretical concentration. It is concentration strong enough to explain both the growth burst and the fragility underneath it.

O.R.B. left the balance sheet, but not the capital question

The third trigger: the O.R.B. strategic funding was fully closed in October 2025, and there was no remaining long-term balance by year-end. That is a genuine improvement in balance-sheet simplicity. Across the life of the arrangement, the company received about $2.6 million and repaid about $2.7 million, of which roughly $1.6 million was cash and $1.1 million was paid in shares.

The dual message matters. Direct financing pressure is lower. Equity pressure is not gone. Part of the cleanup was done with shares, and the same year included $3.4 million of share-based compensation, while the ordinary share count rose to 71.8 million by year-end.

One overhang did come off

The fourth trigger: both the Israeli and the U.S. proceedings around alleged disclosure misstatements from 2024 were dismissed at the initiative of the plaintiffs, without settlement and without payment by the company. That is not a growth driver, but it does remove a legal overhang from the story.

Efficiency, Profitability and Competition

The key profitability insight from 2025 is simple: the company added activity, not earning power. Revenue increased by $8.9 million, but gross profit slipped slightly to $23.8 million from $23.9 million. In practice, almost every extra revenue dollar was absorbed by cost of revenue.

The cost build explains why. IP address costs actually fell to $4.7 million from $5.5 million. The real jump came from networks and servers, which rose to $6.7 million from $1.2 million, and from subcontractors and third-party services, which reached $3.9 million from zero. That is important because it shows the margin damage did not come mainly from the older core. It came from the newer layers, which at this stage require much heavier infrastructure and outside support.

That also frames the real competitive issue. Alarum operates in a market where players such as Bright Data, Similarweb, Oxylabs and SmartProxy compete for the same spending pools, and the company itself flags pricing pressure and aggressive discounting as real industry risks. If winning the new accounts requires more infrastructure, more outside services and more delivery effort, then the issue may not be temporary. The newer product layers look richer to customers, but for now they are not producing better unit economics for Alarum.

Operating expenses are not helping yet either. Research and development rose 67% to $7.5 million, sales and marketing rose 30% to $9.1 million, and general and administrative expense rose 23% to $7.0 million. Some of that is entirely understandable because the company is building a broader product stack. But when operating profit shrinks to just $0.2 million, the operating leverage argument is still waiting to be proven.

One more point that does not jump out on first read: share-based compensation reached $3.4 million in 2025, up from $2.0 million in 2024. That is a non-cash expense, but it still matters economically. The company reported $1.0 million of net profit, yet shareholders are still paying materially through dilution.

Cash Flow, Debt and Capital Structure

The balance-sheet story in 2025 is not really about heavy financial leverage. It is about cash conversion. Alarum ended the year with $12.3 million of cash and cash equivalents and another roughly $10.1 million of debt investments, while long-term loans were gone. That is a much cleaner starting position than it could have been.

But under an all-in cash flexibility lens, meaning after the period’s actual cash uses, 2025 was not a comfortable cash year. Operating cash flow was negative $2.0 million, purchases of property and intangible assets added about $0.4 million of cash use, and lease principal payments took another $0.6 million. The company can absorb that. It is just not a strong cash read.

The single number that explains the most is receivables. Net trade receivables rose to $11.8 million from $3.2 million, a jump of $8.6 million. The company did get a partial offset from a $4.6 million rise in other payables, but cash still moved much more slowly than revenue. This is exactly why it is too early to celebrate the AI angle without qualification. If the large customers are driving more volume but also demanding more credit, growth quality weakens.

The good news is that the balance sheet really is simpler. The strategic loan was fully repaid, part of it through the issuance of 914,856 shares, and there was no long-term loan balance left at year-end. The less comfortable part is that some of the financing clean-up was achieved through equity, not only through cash generation.

Leases are also worth watching. In June 2025 the company signed a new office lease that started in September 2025, with rent of about $58 thousand per month until the end of February 2027 and about $75 thousand per month until the end of August 2029. That is not a balance-sheet threat, but it does add fixed cost at a time when margin is already under pressure.

Outlook

Finding one: 2025 proved that Alarum has new revenue engines, but not yet that those engines are high quality.

Finding two: the older customer base weakened. An NRR of 0.83 means new wins are doing much more of the work than healthy expansion within the installed base.

Finding three: the direct financing problem fell away, but the capital question did not. O.R.B. is gone, yet dilution and share-based compensation remain material.

Finding four: the cash cushion exists, but collections are now the gating factor. Without better revenue-to-cash conversion, even fast growth remains incomplete.

That makes 2026 look like a proof year, not a breakout year. Management says current resources are sufficient for at least the next 12 months, and that clearly gives the company room. But the market is unlikely to reward another revenue beat by itself. It will want to see whether the new products can also deliver collections, margin and less dependence on a small number of very large customers.

The positive scenario is easy to sketch. If data collection solutions and datasets keep growing while infrastructure intensity, third-party costs and customer credit normalize, then 2025 will look like a messy transition year on the way to a deeper and more valuable data platform. NetNut’s Preferred Technological Enterprise status, with a 12% tax rate for 2024 through 2028, also helps if profitability scales.

The less comfortable scenario is just as clear. If the large AI-driven customers keep moving revenue but not NRR, if IPPN fails to stabilize, and if cash conversion remains weak, then Alarum will remain a company with a good growth narrative but unconvincing economic leverage. That is the difference between an attractive product story and a cleaner equity story.

Risks

The first risk is concentration disguised as growth. Two customers accounted for 27% of total revenue, and six customers generated 49% of web data collection revenue. Any pullback from one of them can hurt not only revenue, but also the market’s willingness to underwrite the story.

The second risk is erosion in the older customer base. An NRR move from 1.27 to 0.83 in one year means the company is running forward with one strong leg and one weakening leg. Until that turns, it is hard to call the growth engine truly stable.

The third risk is pricing pressure and the technology arms race. The company operates in a market where websites are layering more anti-bot protections and major platforms are becoming more aggressive toward automated data collection. If Alarum needs to keep adding infrastructure and outside services just to maintain success rates, margin pressure may persist even if revenue keeps growing.

The fourth risk is continued dilution. O.R.B. is gone, but the share count still increased, share-based compensation remains high, and a shelf registration of up to $100 million has been in place since late 2024. It is not proof that capital will be raised, but it does keep the equity question open.

The fifth risk is Israeli operating and currency exposure without hedging. Most of the workforce is in Israel, a meaningful portion of the cost base is in shekels, and the company states that it does not hedge foreign currency exposure. That was not the main problem in 2025, but it remains a real cross-business risk layer.

Conclusions

Alarum exits 2025 as a more interesting company, but not a cleaner one. It now has growth engines that barely existed a year earlier, enough liquidity, and a simpler balance sheet after O.R.B. On the other hand, margin thinned, NRR broke below 1, and cash did not keep up with revenue. This is exactly the kind of report the market can read in two directions, but in the near term it is likely to judge growth quality more than growth itself.

Current thesis in one line: Alarum is proving that there is real demand for its broader web-data stack, but it is not yet proving that this demand is diversified, profitable and backed by cash.

What changed versus the older read: the question is no longer whether the company can grow. It is whether the new growth engine is actually better than the core it is replacing.

The strongest counter-thesis: 2025 may have been only a messy installation year for the newer engines, which would mean the server costs, subcontractor load and receivable build are temporary rather than structural. If that is right, 2026 could show much sharper leverage.

What may change the market’s interpretation in the short to medium term: a combination of stabilizing NRR, a sharp recovery in operating cash flow, and gross margin that stops falling.

Why this matters: without those three things, Alarum risks remaining a growth company on paper with much less convincing economics underneath.

At the bottom line, what has to happen next is straightforward: the older base needs to stop shrinking, the newer customers need to start paying on time, and margin needs to show that the newer products create value for Alarum itself, not just for its customers. If that happens, 2025 will look like a messy but successful proof year. If it does not, the company may have bought expensive growth.