Which TASE-listed companies are exposed to data centers? And which of them are actually mature?

Why this matters now

In the first months of 2026 a wave of filings created the impression that data centers had suddenly become the hottest theme on the TASE. AZRIELI GROUP expanded the Frankfurt project to 54 MW, MEGA OR showed an active facility and a 314 MW IT pipeline under construction, KEYSTONE INFRA laid the cornerstone for a 20 MW IT facility adjacent to the IPM power plant, KARDAN ISRAEL received Form 4 for Kfar Saba and advanced the shell in Shoham, MIVNE closed a partnership with Digital Realty in Petah Tikva, LEVINSTEIN ENG set off with ALFA in Netanya, and DALIA ENERGY, AMPA, DORAL ENERGY, NOFAR ENERGY, SOLAER, and PHINERGY all tried to stake out their own angle.

The problem is that the market tends to pack fundamentally different business models under one headline. A company running an occupied campus, a company holding land with a grid connection, a partner in a single project, a supplier of power or control equipment, and a company that just laid a cornerstone without yet signing a customer — none of these rely on the same economics. A mention of data centers in a filing does not turn a company into direct exposure, and laying a cornerstone is not a substitute for occupancy and NOI.

The real question, then, is not who used the right buzzwords, but who controls a strategic asset in the value chain, who can fund the heavy CAPEX without crushing the balance sheet, and who sits on the right side of the gap between a physical site and shareholder value creation. The most important new layer in the updated map is the IPM cluster in Be'er Tuvia: an operating power plant, adjacent spare land, an explicit plan, and three listed companies on the same site. But even there, each one captures a different slice.

Map of the arena

In practice the arena splits into four layers:

1. Mature direct platform: a company already running campuses, selling capacity to large customers, and showing visible contracted NOI.
2. Single concrete project: a company with a real site, a partner, and agreements, but without a portfolio or broad platform.
3. Power, execution and equipment: a company benefiting from demand without owning the campus — through power contracts, control systems, or construction services.
4. Early optionality: a memorandum of understanding, negotiation, or supply-chain story that has not yet landed a binding contract.

The market cap distribution across this list paints a sharp picture of which names are financially equipped to carry heavy CAPEX, and which are still holding the headline without the capital.

AZRIELI GROUP and MEGA OR already sell capacity rather than just land

AZRIELI GROUP is already at the point where its data-center arm should be read like an operating business rather than a real-estate option. At the reporting date the group operated four data centers in Norway and one active campus in East London, with average occupancy in active assets at about 99%. At the same time it had roughly 148 MW signed in active assets in Norway and the UK, while the 2025 presentation already showed 257 MW contracted and contracted NOI of about ILS 1.04 billion. On February 16, the Frankfurt option for another 18 MW was exercised, taking that project to 54 MW. On February 25, non-binding financing principles were signed for an 80 MW built-to-suit project in Norway.

What matters in AZRIELI GROUP is not only scale, but the type of progress. The TikTok project in Norway already completed delivery of 90 MW in 2024, while the customer also exercised an option to expand to 120 MW, even if the final terms are still under negotiation. In Frankfurt, data-center service agreements were signed for 36 MW, expanded to 54 MW, and the project is backed by non-recourse (no recourse to the parent) financing. That is the exact difference between a site story and a platform story. The customer, the contract, the funding, and the build-out already sit on the same track.

MEGA OR is the closest Israeli version of that model, but with a heavier execution profile. As of the reporting date, through Mega D.C it had one active facility in Modi'in at about 9.5 MW IT, plus seven facilities under construction with total capacity of 314 MW IT. Within that framework it had signed agreements for 174 MW IT of data-center services. On May 14 it signed an 8 MW IT stage A agreement in Modi'in with a subsidiary of a Nasdaq-listed international cloud and AI infrastructure group. On December 31 it signed another contract package for 19 MW IT in Modi'in and Yoav, with an option to expand by another 50 MW IT. On January 6 it added another 80 MW IT across Yoav and Beit Shemesh, including 22 MW IT in Masmiya and 58 MW IT in Beit Shemesh.

In MEGA OR, the thesis already looks real, but it still sits on heavy CAPEX. The company points to a cost of roughly $11 million per MW IT including land. That means the market should not ask only whether demand exists. It should ask how quickly those agreements turn into occupancy and NOI, and how much capital has to be carried along the way. This is no longer optionality, but it is still an execution year.

KARDAN ISRAEL is already close to revenue but is still stuck in a bridge year

KARDAN ISRAEL sits almost exactly between those two groups. It is no longer a land-and-zoning story, but it still does not look like a fully matured platform in the style of AZRIELI GROUP or MEGA OR. In Kfar Saba, the company has an underground project of about 12,500 square meters with expected power allocation of 22.2 MVA and maximum capacity of roughly 16 MW IT. In March 2025 it signed a lease with Serverz at roughly ILS 6.7 million of annual rent, and after the balance-sheet date it received Form 4 (occupancy permit). By the reporting date, the site already had initial telecom and financial-sector customers at around 0.5 MW IT.

That matters because Kfar Saba has already crossed the line between planning and actual use. But Shoham still shows how capital-heavy the path remains. In Shoham, three data centers are planned with 48 MVA of power allocation plus 48 MVA of backup, and up to 24 MW IT. The permit for data-center use was received in January 2025, the first two facilities are already under construction, and the shell has been completed. Management frames those first two sites as carrying annual revenue potential of about ILS 135 million at full occupancy. Against that, the cost of those two sites alone is estimated at more than ILS 700 million, and the company had already provided about ILS 161 million of funding to the Shoham joint venture by the end of 2025.

The point in KARDAN ISRAEL is not that there is no project. There is a project, and there is even a second site that is already touching occupancy. The point is that the company is financing a bridge year right now. In both Kfar Saba and Shoham it guarantees part of the partner obligations to the banks. That is why KARDAN ISRAEL is much closer than optionality names, but still not yet at the stage where the story can be measured through stable NOI rather than capital, guarantees, and construction progress.

One site in Be'er Tuvia created three different equity stories

The IPM cluster is the most important addition to the updated map because it offers something the local market barely has: an existing power plant, adjacent spare land, and an explicit plan for server farms on the same site. But that story does not belong to one company. It belongs to three. The KEYSTONE INFRA presentation shows a 100 MW IT development pipeline in data centers, including 40 MW IT adjacent to the IPM power plant. That is enough to put KEYSTONE INFRA firmly into the discussion, but not enough to place it alongside platforms that already sell occupied capacity.

ALMA YESODOT exposes most of the economic depth of the site. The company owns 18.14% of Triple, and Triple owns about 84% of IPM. The Be'er Tuvia power station operates at 451 MW, and on Triple's spare land the company is advancing two server farms of 20 MW each, together 40 MW. Building-permit applications were filed on December 29, 2024, the local committee approved them on November 19, 2025 subject to conditions, and an excavation and shoring permit for the northern lot was received on January 22, 2026. ALMA YESODOT also says the first 20 MW server farm is expected to finish during 2028, and that after the balance-sheet date Triple's board approved decisions for the project with investment of about ILS 900 million.

The exposure of G.P. GLOBAL-M is different. In its materials, the story runs through the IPM power station itself, its value uplift, and its economic cash flows, not through a separate disclosed line of server-farm revenue. That changes the read materially. G.P. GLOBAL-M benefits if the power station and the surrounding site create value, but its filings do not yet present a model of customers, occupancy, or direct NOI from a data-center venture. That is why anyone who places KEYSTONE INFRA, G.P. GLOBAL-M, and ALMA YESODOT in the same basket misses the fact that each one sits on a different layer of the same site.

There is also structural friction here. ALMA YESODOT details legal proceedings that emerged around Triple board decisions from February 16 regarding the server-farm venture. That does not cancel the site's potential, but it does remind the reader that the path to value runs not only through power and land, but also through governance, partner alignment, and the ability to fund construction without turning the site into a shareholder conflict.

MIVNE and LEVINSTEIN ENG hold concrete projects, while DORAL ENERGY, UNITRONICS and ORMAT TECHNO. hold secondary angles

MIVNE and LEVINSTEIN ENG belong in the concrete-project bucket rather than the broad-platform bucket. In MIVNE, that is especially clear. On May 8, 2025 the company, through a joint venture with Digital Realty, signed with MedOne to add it as a partner in an 18 MW data-center project on about 10 dunams in Petah Tikva. Competition approval was received on June 25, and the deal closed on November 5. This is already more than theoretical optionality, but it is still ownership in one project with strong partners, not a stand-alone platform that starts to look scaled.

In LEVINSTEIN ENG, the story is ALFA Data Center in Netanya. The company and an international partner are building, on a 50-50 basis, a server farm with capacity of roughly 18 to 21 MW IT and built area of about 20,000 square meters. Total stage-one investment is estimated at ILS 780 million, of which the company's share is about ILS 390 million, and a similar stage B remains possible. Here too there is a site, a partner, and better visibility than in most option stories, but the exposure still sits in a single asset rather than a portfolio.

DORAL ENERGY, ORMAT TECHNO., and UNITRONICS offer a very different angle. On February 16, DORAL ENERGY signed a cooperation agreement with Ampa for the development, planning, construction, financing, and operation of data centers in Israel through a 50-50 special-purpose vehicle. The company argues that its edge comes from access to grid connections, land near those connections, relevant customers, and projects already being advanced in solar and storage. That logic is understandable, but the agreement is still a framework that awaits approvals, project selection, and translation into an actual signed site.

ORMAT TECHNO. does not operate an Israeli campus. It supplies the power layer. On January 12 it signed a 20-year PPA (long-term power purchase agreement) with Switch for approximately 13 MW of geothermal power to support Switch's Nevada data centers, with an option to add roughly 7 MW of solar. On March 23, ORMAT TECHNO. also closed a $1 billion convertible issuance at low or zero coupon, reinforcing its funding flexibility. That is real exposure to data-center demand, but it is measured through power contracts, not through rented server space.

UNITRONICS is the clearest example of a secondary technology name with disclosed proof. The company says that in 2025 data-center activity grew from negligible to more than 10% of total sales after targeted marketing and product development for controllers used in cooling and energy-management systems. It specifically highlights RedFish and IPv6 support and alignment with hyperscale (very large cloud customers) requirements. That does not make UNITRONICS a real-estate or infrastructure company, but it does make it a name where the exposure has already moved from headline to revenue line.

By contrast, in ELMOR ELECTRIC and parent RAPAC, the local materials still do not disclose a separate data-center activity line, either in backlog or in revenue. That does not mean the angle is irrelevant. ELMOR ELECTRIC is clearly active in electrical works and infrastructure across low, high, and extra-high voltage, and the materials show experience in control systems, electrical components, substations, and grid connection. But at this stage the reader still cannot underwrite a separate data-center economics line in ELMOR ELECTRIC or RAPAC. For now, they remain indirect execution exposure, not the core of the theme.

The big money arrives only after grid, funding and occupancy line up

Demand is not the main bottleneck here. Coordination is. In AZRIELI GROUP, the chain is already relatively aligned: a large customer, funded projects, active assets, and high occupancy. In MEGA OR, customers and agreements already exist, but most of the economics still have to pass through execution years and occupancy ramps. In KARDAN ISRAEL, part of the story is already touching revenue, while another part still consumes capital and guarantees. In the IPM cluster, the story is even earlier. A strong site and clear power advantage still do not replace signed funding, occupancy, and a clean path for value capture by shareholders.

That is also why MW should not be compared on a one-for-one basis. AZRIELI GROUP's 257 MW contracted are not equivalent to KEYSTONE INFRA's 40 MW next to IPM, or to MIVNE's 18 MW in Petah Tikva. These sit at very different stages of the value chain. Even within Israel, the 174 MW IT signed by MEGA OR are worth more than a general statement of intent, because they already come with contracts, delivery schedules, and a first active facility.

The funding layer matters just as much as the power layer. MEGA OR speaks about roughly $11 million per MW IT including land. KARDAN ISRAEL carries more than ILS 700 million of cost for the first two Shoham facilities. Through Triple, ALMA YESODOT points to around ILS 900 million of investment. Those figures explain why the market should stop getting excited by every land story and start asking who closes the capital, who meets the timetable, and who reaches cash generation that actually flows back to the listed shareholder.

What will decide the map over the next year

For AZRIELI GROUP, the near-term test is whether the extra 18 MW in Frankfurt and the 80 MW dedicated Norway project keep moving forward with the same funding and execution discipline already seen in prior projects. For MEGA OR, the question is less how many MW are signed and more how many turn into delivery, customer fit-out, and visible NOI in 2026 and 2027. For KARDAN ISRAEL, the real metric will be the pace of occupancy in Kfar Saba and the move in Shoham from shell and construction into active capacity.

In the IPM cluster, the questions are harder. KEYSTONE INFRA, G.P. GLOBAL-M, and ALMA YESODOT still need to show that permits, funding, partner coordination, and first works are converging into a clear development path. Without that, the site will remain strategically important but still far from value that can be rolled into a cash-flow model. For MIVNE and LEVINSTEIN ENG, the test is whether those discrete projects progress into occupancy rather than remaining mostly valuation stories. DORAL ENERGY needs to move from framework agreement to defined sites and customers. ELMOR ELECTRIC and RAPAC need the first disclosed data point, not just theoretical relevance.

Even in the names that already show proof, the next question remains open. UNITRONICS has already shown that the field passed the 10% sales threshold, but it now has to prove that the contribution is recurring rather than one-off. ORMAT TECHNO. already sits on a real power contract into the data-center industry, but it will be judged by the depth of the pipeline that follows Switch, not by the existence of the first agreement alone.

Conclusions

The picture that emerges from this review leaves no room for illusions. Not every company that floats the phrase "data centers" provides real exposure to the theme, and not every MW is worth the same.

AZRIELI GROUP is the most mature name in the field: 148 MW active and leased, 257 MW in total commitments, contracted NOI of about ILS 1.04 billion, and non-recourse financing in place. It is the only complete equation in the local arena.

MEGA OR is the most interesting Israeli platform, with an active facility, 174 MW IT of signed contracts, and 314 MW IT under construction. But the gap between today's representative NOI of ILS 38 million and the full target of about ILS 1 billion is a reminder that this is still an execution year, not a delivery year.

KARDAN ISRAEL is closer to revenue than the market tends to assume: initial occupancy in Kfar Saba, shell going up in Shoham, permits in hand. But anyone buying today is buying mostly the investment stage, the guarantees, and the execution risk.

The IPM cluster is the most important addition: KEYSTONE INFRA holds the plan and the cornerstone, ALMA YESODOT carries the full economic picture through Triple and IPM, and G.P. GLOBAL-M holds the power plant itself. But the structural risks — legal proceedings between partners, unclosed financing, and the absence of separate disclosure in some filings — are a reminder that the physical site is still some distance from shareholder value creation.

MIVNE and LEVINSTEIN ENG each hold one concrete project with a strong partner and a known investment framework. Real exposure, but narrow.

DORAL ENERGY, ORMAT TECHNO., and UNITRONICS show tangible secondary activity, each in its own layer: a framework agreement, a signed PPA, or more than 10% of sales from data-center infrastructure controls.

RAPAC and ELMOR ELECTRIC still do not present filing-level disclosure strong enough to classify them as core names, despite the mentions. The capability exists; the disclosure has not caught up.

DALIA ENERGY, NOFAR ENERGY, SOLAER, PHINERGY, AMPA, and MESHEK ENERGY remain, at this stage, early optionality: no signed site, no customer, or no direct disclosure.

The conclusion is clear: mixing land, grid connections, cooling systems, memoranda of understanding, and active campuses under one label produces a distorted picture. The data-center value chain on the TASE is taking real shape, but the pie is not divided evenly. There are genuine platforms, there are single concrete projects, there are proven infrastructure plays, and there are quite a few stories that are still years away from creating shareholder value. That is the core filter now.