Phinergy: The LOI Is Large, but 2026 Is Still a Proof Year

Introduction to the company

At first glance, Phinergy looks like a dream story around clean backup power for data centers. There is aluminum-air technology, there are big names such as Google and Microsoft inside the Net Zero framework, there is an LOI for up to 20 GW, and there is also a first deployment agreement with Swisscom. But economically, this is still not a growth company already selling into the data-center market at scale. It is a development and commercialization platform with a very small real commercial base in telecom and critical infrastructure, trying to use that base to fund and validate a move into a much larger market.

What is working now is clearer than it used to be. During 2025 the company completed its pivot toward the data-center market, finished the engineering design for the Net Zero system, signed an LOI with a hyperscaler, signed an initial agreement with Swisscom, and ended the year with NIS 18.8 million of backlog. On top of that, in January 2026 it completed a roughly NIS 30 million gross equity raise and received an additional order for 76 systems, completing a broader program of about NIS 22 million with Israel National Roads contractors.

The problem is that the first screen is misleading. In practice, 2025 ended with only NIS 354 thousand of revenue, down 92.7% from 2024, and with a net loss of NIS 54.3 million. This is not a company already being carried by data-center revenue. It is a company that still burns cash, still relies on equity raises and warrant exercises, and still has to clear several proof points before an LOI or a validation program turns into binding large-scale orders.

That is the active bottleneck: not only technology, but conversion. Conversion of pilots into validation, validation into orders, orders into serial production, and production into a cash story that does not keep leaning on the capital markets. Anyone looking only at the size of the LOI is likely to miss that management itself says the path depends on successful validation, production readiness, supply-chain readiness, and a binding supply agreement.

There is also a practical market screen here. This is still a small R&D equity story, and three traded warrant series remain on screen at the same time. So in Phinergy's case, funding and dilution are not background noise. They are part of the everyday equity story, almost as much as the commercial milestones themselves.

The quick economic map at the end of 2025 looks like this:

Events and triggers

The LOI is large, but it is still not revenue

The event that caught the market's eye at the start of 2026 was the February 19 LOI with a hyperscaler inside the Net Zero framework. The headline is large: potential purchases of up to 20 GW through 2030, implying potential sales in the billions of dollars. For a company of Phinergy's size, that is a very unusual demand signal.

But this is exactly where the analysis has to slow down. The LOI is subject to successful validation, production readiness, supply-chain readiness, and a binding supply agreement. That means the market can get carried away by the size of the headline, while the company itself is already telling investors what still blocks the path. The real trigger here is not 20 GW. The real trigger is validation and the ability to produce.

Net Zero is the core of the proof path

To understand the LOI, the prior project matters more than the press headline. Under Net Zero, the company is developing and installing a system of about 500 kW and 10 MWh, while the validation work itself is funded by the framework in an estimated amount of about NIS 2.3 million. On December 30, 2025 the engineering design was completed, and from that point the company moved into execution.

The path itself is staged. First comes an intermediate module of about 70 kW for testing and demonstration, and only then the full system. The company says the validation process is expected to finish in the fourth quarter of 2026. That leads to a simple conclusion: even if the strategic story is real, 2026 is still a proof year in the product. Not yet a year of large-scale revenue from this market.

Swisscom and BIRD deepen the story, but they also underline the transition phase

Two additional moves add depth to the thesis, but they do not change its nature. On March 27, 2026 the company signed an agreement with Swisscom Broadcast AG. At the first stage, Swisscom is buying only two systems, for consideration that is not material to Phinergy, in order to run demonstration and validation during the second half of 2026. Here too the company points to an addressable market of about 2,000 sites over five years, but here too future purchases depend on successful validation.

At the same time, in a separate path, the BIRD and NYPA project gives the company about NIS 4.8 million of grant support for a 27 month joint project to develop and demonstrate a diesel alternative backup system. That project includes a system of about 30 kW and 1.5 MWh, while the company itself is expected to fund about $1 million of the program. This is an important external validation that credible institutions are willing to fund and test the technology. But it is also a reminder that the first layer of proof is still partly wrapped in grant and validation frameworks, not yet in large end-customer purchase orders.

For now, the older business still produces the real numbers

The actual hard commercial progress came from the older market. On January 9, 2025 the company completed a successful pilot for Israel National Roads, where it had to demonstrate 48 hours of uninterrupted supply. After that, between November 2025 and January 2026, it signed binding agreements for 300 backup systems with four contractors of Israel National Roads, for an overall program of about NIS 22 million. In the January 8, 2026 additional order for 76 systems, expected consideration is about NIS 4.5 million for the systems plus about NIS 1 million for related services, with about NIS 5 million expected to be collected during the first two years.

That matters because it shows the company is not living only on intent. It does have real signed business. But it matters just as much to understand what this is not. This is still the telecom and critical-infrastructure market, based on Phinergy's existing product, not yet the large new engine the market tends to associate with the data-center narrative.

The pivot toward data centers came with the shutdown of an older path

Another non-obvious point is that the new focus did not only open a door. It also closed one. On December 1, 2025 the company said it was orderly reducing the India JV with Indian Oil as part of its focus on data centers. Following that decision, it wrote down the remaining investment and translation reserve in a total amount of about NIS 1.6 million, and the receivable tied to the sale of 40 systems to IOP was reclassified in order to pursue near-term collection.

That matters because the pivot is not only a strategic expansion. It also comes with an admission that an older commercialization route did not mature. So 2025 is not just a year of building a new story. It is also a year of acknowledging that an earlier story did not get where it was supposed to go.

Efficiency, profitability and competition

The accounting story of 2025 may look simple at first glance: revenue collapsed, but the operating loss narrowed slightly. That is not evidence of a healthier business model. It mainly shows that the company cut cost around a business that barely sold anything.

The revenue mix looks more mature only because product sales collapsed

In 2024 the company sold NIS 4.546 million of products and NIS 297 thousand of services. In 2025 product revenue fell to only NIS 198 thousand, while services fell to NIS 156 thousand. That is why services rose to 44.1% of revenue from only 6.2% a year earlier.

Anyone reading the mix without reading the absolute numbers could conclude the company is building a more recurring service base. That would be the wrong read. The mix changed mainly because product sales collapsed. The recurring base itself remained very small.

The customer base is also extremely concentrated. In 2025, 86% of revenue came from Cellcom, 12% from Israel National Roads, and the rest was almost negligible. That means the company still has not built a broad commercial layer even in the older market. Before thinking about data centers, it is worth remembering that the current commercial base still sits almost entirely on two names.

The gross-loss improvement is largely a favorable comparison effect

The gross loss narrowed to NIS 2.342 million from NIS 8.767 million. That sounds like a sharp improvement, but the underlying reason matters. The company itself says that 2024 cost of sales included about NIS 2.8 million of inventory write-downs, versus only about NIS 180 thousand in 2025. So part of the improvement comes from comparing against a year loaded with inventory impairments.

Even after that, the picture is still weak. On only NIS 354 thousand of revenue, the company still recorded NIS 2.696 million of cost of sales, including NIS 1.255 million of depreciation and amortization. In other words, the industrial platform and the production layer are already visible in the accounts, but the revenue base is not yet moving with them.

The company cut overhead, not really development

Phinergy did manage to reduce G&A to NIS 13.775 million from NIS 16.340 million, mainly through lower professional fees. That is real and it is part of the efficiency plan. But R&D barely moved, NIS 23.878 million versus NIS 24.199 million, and sales and marketing also barely changed, NIS 5.972 million versus NIS 6.027 million.

That means the company did not really shift into the model of a mature operating business optimizing around growth. It shifted into the model of a commercialization-stage company still investing heavily in development, while trimming the corporate layer and hoping that investment starts to pay off. That is consistent with the stage of the company. It is simply still far from a self-supported revenue engine.

Even in the second half, revenue did not begin to carry the cost structure

The second half of 2025 looked slightly better on revenue, NIS 290 thousand versus NIS 64 thousand in the first half. But the operating loss actually widened to NIS 28.0 million from NIS 22.4 million, partly because of higher R&D and a NIS 1.7 million other-expense line that included the India JV write-down.

That matters because it prevents an overly comfortable read of the year-end trend. Even when some service revenue showed up, it still did not change the underlying economics of the business.

The theoretical production capacity is there, but the service proof is still ahead

The air-cathode production line already has meaningful theoretical capacity relative to the company's size, 20 MW on one shift, or 40 MW to 60 MW on two to three shifts. In the company's own translation, that means about 5,000 backup systems of 4 kW or about 25 data-center backup systems of 800 kW on one shift.

But this is exactly where the gap between industrial capability and a ready commercial business sits. The company itself classifies production scale-up, market penetration, and the need for very high service standards as high-impact risks. So even if the product works, it still has to be delivered, installed and supported at the level demanded by telecom, critical infrastructure and data-center customers. Those are conservative markets with strict reliability requirements.

Cash flow, debt and capital structure

Anyone trying to understand Phinergy has to read 2025 through cash, not only through the income statement. The picture here is very clear: the company bought time, but it still has not built an internal cash engine.

The all-in cash flexibility picture is still deeply negative

Under an all-in cash flexibility framing, meaning cash left after actual period cash uses, 2025 ended with about NIS 35.7 million of cash burn. Cash flow from operations was negative NIS 32.063 million. On top of that came reported CAPEX of NIS 459 thousand, lease principal cash of NIS 3.040 million, and NIS 119 thousand of payments tied to the Innovation Authority obligation.

There is not a more reassuring normalized cash-generation picture underneath that. In some commercialization businesses, it is possible to argue that the existing business generates cash before growth CAPEX. That is not the case here, because even before growth investment the business itself still consumed more than NIS 32 million of operating cash.

The January 2026 raise is a bridge, not a resolution

At the end of 2025 the company had NIS 21.5 million of cash and another NIS 3.2 million of short deposits. That is meaningful for a company of this size, but it is still less than the operating cash burn of 2025. That is why the roughly NIS 30 million gross raise in January 2026 was necessary, not optional.

Management effectively says that itself. The 18 month cash forecast relies on the January 2026 raise, on warrant exercises that already happened, and on the ability to cut expenses further if needed. It also says that if additional funding does not arrive through Series 3 exercises by April 15, 2026, the company believes it can make additional cost adjustments until the required financing is completed.

That is not the language of a comfortable balance sheet. It is the language of a financing bridge.

There is no bank debt, but there are real economic obligations

The company emphasizes that it has no financial debt in the banking sense. Formally that is correct. But economically it is impossible to ignore NIS 21.434 million of government-grant liabilities and NIS 9.069 million of lease liabilities. Together they already amount to NIS 30.5 million, versus only NIS 24.7 million of cash and short deposits.

And that is before the deeper meaning of some of the grants. Under the oil-alternative program, the grants the company received were converted into an obligation to pay the Innovation Authority royalties of 3% or 4% of all company revenue. So if commercialization does arrive, a slice of revenue is already pre-committed.

The company is also subject to knowledge-transfer restrictions. In some cases, transferring knowledge or rights to use it outside Israel can require prior approval and even payment of up to six times the funding received, plus interest, net of royalties already paid. This is not an immediate balance sheet threat. But it does matter when thinking about global commercialization. Not all value created in the technology is automatically and freely available to common shareholders.

The most important outside warning signal comes from the auditors

There is also a clear external warning signal here. The auditors did not qualify the statements, but they did include an emphasis-of-matter paragraph on the company's financial condition. They also identified both the company's financial condition and the measurement of government-grant liabilities as key audit matters. This is not technical noise. It means the two questions that matter most for the equity thesis, funding and future obligations, are also the two questions the auditor saw as especially sensitive.

Outlook and guidance

Before getting into 2026, four points matter more than they may seem on a first read:

• The NIS 18.8 million backlog is real, but it is smaller than the NIS 32.1 million operating cash burn of 2025.
• The existing revenue base is still extremely small, so even several small or medium contracts are not enough by themselves to change the whole picture.
• Management presents large theoretical inflow potential from the warrant series, but the cash forecast does not include warrants that have not yet been exercised.
• Most of the critical Net Zero and Swisscom milestones sit in 2026, which means product proof still comes before revenue proof.

2026 is a proof year in the product

The company says expected R&D investment for the next 12 months, excluding Net Zero, is about NIS 15 million. That is a very important number. It means the company will keep spending meaningfully even after the cost cuts, before the new revenue engine has been proven.

In practice, 2026 has to deliver three technological and commercial milestones together:

1. Completion of the roughly 70 kW intermediate module inside Net Zero.
2. Meaningful progress toward the full 500 kW and 10 MWh system by the fourth quarter of 2026.
3. Successful demonstration with Swisscom during the second half of 2026.

If one of these slips, even the largest LOI will remain a headline rather than a commercial path.

2026 is also a transition year in revenue

Even under a positive operating scenario, it is still hard to see 2026 turning immediately into a breakout accounting year. The backlog scheduled for 2026 stands at NIS 10.148 million, while 2025 revenue was only NIS 354 thousand. So the company can certainly show strong top-line growth from here, but even that would still be starting from a very low base.

In other words, even if 2026 ends up being a good operating year, it may still look like a transition year in the reported numbers. The reason is simple: most of the strategic value would still be sitting on building the next step, not on already-converted large revenue.

The third test is still funding

This is probably where the market will focus most in the near term. Beyond the capital already raised in January 2026, the company still operates in a world where warrants are both a funding mechanism and a dilution mechanism. Series 3 can be exercised until April 15, 2026 at NIS 2 per warrant, while the last trading day is April 10, 2026. By the time the financial statements were approved, about 2.307 million Series 3 warrants had already been exercised for about NIS 4.7 million, but management still does not build its forecast around warrant exercises that have not yet happened.

That distinction matters. Anyone looking only at the theoretical warrant inflow could conclude the funding question is already closed. Management itself is much less comfortable than that.

The market is not leaning on short sellers here, but on execution

One more useful market-layer point is that short interest does not currently signal an aggressive bearish stance. As of March 27, 2026 short interest stood at only 0.29% of float, with SIR of 0.1. There was a temporary spike to 3.0% of float and SIR of 3.9 in January, but by late March most of that pressure had faded.

That matters because it suggests 2026 will probably not be decided by a battle with short sellers. It will be decided by reports, milestones and the funding path.

Risks

Most of the data-center layer is still not contractually locked

The central risk is that the gap between commercial interest and binding orders is still large. The LOI is subject to explicit conditions. Swisscom at the first stage is buying only two systems. Net Zero is still in the demonstration stage. So the company has already proved that it can generate interest, but it has not yet proved that this interest reliably turns into large binding orders.

Production and service can become the next bottleneck

The company itself classifies production scale-up, market penetration and the need for high service standards as high-impact risks. That is an important admission. Even if the product works, it still has to be delivered, operated and supported at the level expected by telecom, critical-infrastructure and data-center customers. These are conservative markets with very strict reliability demands.

Funding remains a core risk, not a footnote

The company explicitly classifies funding sources as a unique high-impact risk, and that is justified. 2025 ended with negative operating cash flow, 2026 started with another equity raise, and the cash forecast still relies on the ability to cut expenses again if warrant exercises are not enough. This is not a hypothetical risk. It is the current operating model.

The Innovation Authority is both support and friction

Government funding helped finance the technology over the years. But it also creates royalties, a NIS 21.4 million accounting liability, and restrictions on knowledge transfer. So if the positive story does work, it will not flow one-for-one into freely accessible value. Part of it is already committed through obligations and royalties.

Conclusions

Phinergy enters 2026 with a much sharper strategic direction, several genuine demand signals, and a target market far larger than the one it serves today. That supports the thesis. What still blocks a cleaner read is that the company barely sells today, still burns cash, and still has to clear the validation and production path before it can be read as a true commercialization story. In the near term, the market is likely to react less to the size of the LOI and more to whether funding holds and whether one of the key technical milestones is actually delivered on time.

Current thesis in one line: Phinergy has already proved that large players are willing to engage, but 2026 will decide whether that engagement becomes a commercial path or remains mainly a funded validation layer.

What changed relative to the superficial reading of the company is fairly clear. A few quarters ago it was easy to read Phinergy as an interesting technology dream without enough motion around it. Now there is a sharper data-center pivot, a Net Zero project, an LOI, Swisscom, and signed backlog. But it is now also much clearer that this path still sits on a cash bridge, dilution, and a fairly long proof timeline.

The strongest counter thesis is that the caution here is excessive. One can argue that it should already be enough that the company has a first production line, real business with Israel National Roads, completed engineering design for Net Zero, and an unusually large demand signal from a hyperscaler. In that reading, commercialization has already started and only needs time. That is a serious argument. The problem is that the 2025 numbers still do not confirm it.

What could change the market's interpretation in the short to medium term is a combination of three things: real Net Zero progress on schedule, successful validation with Swisscom or a similar anchor, and a funding layer that looks less dependent on another round of market financing or warrant exercise. If those three happen together, the way Phinergy is read can change quickly.

Why does this matter? Because in Phinergy's case the large value, if it arrives, will not be created by one more quarter of telecom revenue. It will be created when the company proves it can move from a differentiated solution to a large-scale backup platform that can actually be sold and delivered. Until then, this remains a proof story.

What has to happen over the next 2 to 4 quarters for the thesis to strengthen is fairly clear: validation has to keep moving on schedule, the full Net Zero system has to progress, one of the proof paths has to start converting into a broader commercial commitment, and the funding layer has to keep holding without another major round of cuts or highly dilutive financing. What would weaken the thesis is a technical delay, a supply-chain problem, failed warrant funding, or another year in which backlog remains too small relative to the pace of cash burn.