AI Infrastructure on the TASE Splits Into Orders, Pilots, Land and Power
Rimon, Buff, Solair and Alon Tavor show four different levels of exposure to AI and data centers. The reader value is not another company list, but a ranking by paying customer, commitment, financing and remaining conditions.
Two reporting days created a small AI infrastructure map on the TASE: Rimon with an order of about $17 million for cooling systems at a U.S. data center, Buff with a small AI-data pilot for a major technology company, Solair with a land lease in Chile for a renewable-powered data-center project, and the Alon Tavor layer through Generation Capital, Rapac, Rapac Energy and Mivtach Shamir around planning for a gas-fired power plant of up to 900MW. The common theme is infrastructure demand around AI and data centers. The differences matter more than the common theme. A signed order, a pilot, land optionality and power planning are completely different in commitment, cash path, capital need and time to revenue.
Four Proof Levels
The right way to read this wave is by proof level, not by the magic word AI. The first question is who already has a paying customer. The second is who still needs a permit, financing, customer or pilot expansion. The third is how long it may take before the filing appears in revenue or cash flow.
| Company | Exposure layer | Current proof level | What is still missing |
|---|---|---|---|
| Rimon | Data-center cooling and air handling | Signed order of about $17 million | Margin, working capital and revenue recognition |
| Buff | Data for the AI industry | About $300 thousand pilot plus a small additional deal | Expansion into a meaningful and recurring contract |
| Solair | Land and renewable energy for a Chile data center | Land lease and development option | Customer, permits, financing and construction |
| Alon Tavor | Available power for energy-intensive infrastructure | Authorization to prepare a national infrastructure plan | Approvals, licensing, financing and investment decision |
The chart is not a valuation. It is a way to read evidence quality: a signed order carries more weight than a pilot, and a pilot is different from land optionality or power-plant planning.
Rimon Is Closest to Revenue
Rimon has the most concrete filing: an order of about $17 million for Fan Wall Unit systems, cooling and air handling for a Pennsylvania data center. The product sits in a critical part of the data-center construction chain. Without cooling and air flow, high-power computing equipment cannot operate reliably.
Even here, caution is needed. Deliveries are expected only from June 2027 through April 2028, and profitability is not disclosed. An order can increase backlog, but the stronger thesis requires execution, payment terms, margin and follow-on orders. Rimon's advantage is that the filing starts with a customer and a product, not with an idea.
Buff Is Still Testing the Market
Buff is in a different place. A roughly $300 thousand pilot with one of the five largest global technology companies is interesting mainly because of the customer's identity. It suggests the company's data may have relevance beyond gaming. But commercially, this is still a small pilot, with obligations to be completed by the end of July and no commitment to expand.
So the question is not whether Buff's headline is interesting. It is. The question is whether the company can turn data into a recurring, priced and saleable product. If the pilot expands into a meaningful agreement, it can become a turning point. If not, it remains a good marketing signal rather than a proven business.
Solair Has Land, Not a Data Center
Solair reported a lease agreement in Chile for a renewable-powered Data Center project. The exposure is interesting because it connects land, green power and the possibility of an energy-intensive facility. But it is also the earliest-stage operating-company exposure in this map. A lease is not a customer, not a permit, not financing and not construction.
The filing matters as a strategic option, especially if Solair can connect the land to a customer, licensing and financing. Until then, it should be read carefully. It opens a window into the sector, but does not prove commercial demand for the specific project.
Alon Tavor Is the Power Layer Behind the Story
A gas-fired power plant is not a data center and not an AI product. Still, if power demand from data centers, industry and cooling keeps rising, available generation becomes part of the infrastructure that enables the whole story. That is why Alon Tavor belongs in the map, but not in the same layer as Rimon or Buff.
Authorization to prepare a national infrastructure plan for a power plant of up to 900MW improves planning probability. It does not create cash flow. For the public companies exposed to the project, this is an optional asset inside a broader electricity thesis. The monitoring point is whether planning advances into approvals, licensing, financing and an investment decision.
The read: the story is not "who is exposed to AI." That question is too broad and often misleading. The story is who already has business proof, who has a pilot, who has a land option and who has a more distant infrastructure route. Rimon is in the order layer. Buff is in the pilot layer. Solair is in the land and option layer. Alon Tavor is in the power and planning layer.
That is the distinction readers should take from the current wave. AI and data centers are not one investment category. They are a value chain with very different certainty levels. Ignoring the difference between an order, a pilot, land and power risks assigning the same weight to events with very little economic similarity.
Disclosure: Deep TASE analyses are general informational, research, and commentary content only. They do not constitute investment advice, investment marketing, a recommendation, or an offer to buy, sell, or hold any security, and are not tailored to any reader's personal circumstances.
The author, site owner, or related parties may hold, buy, sell, or otherwise trade securities or financial instruments related to the companies discussed, before or after publication, without prior notice and without any obligation to update the analysis. Publication of an analysis should not be read as a statement that any position does or does not exist.
The analysis may contain errors, omissions, or information that changes after publication. Readers should review official filings and primary sources before making decisions.