Summary

  • Adani Enterprises and Jabil’s 13 October non-binding MoU proposes a 50:50 Indian company to make and assemble data-centre and AI equipment. It remains conditional on definitive agreements, approvals and other customary conditions.
  • Equal equity does not reveal who would fund each production line, choose the factory’s location, approve investment, own designs, commit orders or carry product and warranty risk. The release discloses none of those operating terms.

A precise split around an unbuilt structure

The most precise number in Adani Enterprises and Jabil’s announcement is also the one most likely to invite over-reading. The companies say their proposed Indian joint venture would be owned 50:50. They do not publish the governance, financing or commercial agreements that would turn that proportion into a working manufacturing business.

The 13 October memorandum is explicitly non-binding. It sets a framework for a venture that would manufacture and assemble AI and data-centre equipment, but says formation still depends on definitive agreements, applicable approvals and customary conditions. That is a step beyond the 15 June announcement of an intended strategic alliance. It is not yet evidence of a funded factory, a customer programme or a production start.

The distinction matters because 50:50 is an ownership description, not an operating chart. Public documents do not say how directors would be appointed, which decisions would require both partners, how a deadlock would be resolved, or what each side must contribute in cash, equipment, technology, personnel or customer access. A shareholder split may be the starting point for such terms; it cannot stand in for them.

A product list is not a production system

The proposed portfolio is broad. The announcement names AI racks, servers, storage and networking equipment, then adds power-distribution units, liquid-cooling systems and transformers. It also includes installation and commissioning. The list reaches from assembling a rack around purchased components to making power and cooling equipment with their own qualification, supplier, safety and warranty requirements. It does not say which products would be built first, which processes would sit in India, or which components would remain imported.

Nor is this a semiconductor-fabrication announcement. The disclosed scope concerns equipment around computing systems: racks, servers, storage, networks, power and cooling. That boundary matters when reading the companies’ “multi-gigawatt” language. The release describes intended rack-manufacturing capacity supporting computing systems with a combined power capacity of several gigawatts. It does not report installed compute, energised data-centre load, shipped racks or accepted customer capacity.

Manufacturing capacity has economic value only when a product configuration can be qualified, supplied, assembled at a repeatable yield, delivered on time and supported after installation. High-density systems raise the cost of getting those joins wrong. A rack may arrive on schedule while its power distribution, cooling loop, networking or site integration remains on a different schedule. The announcement does not assign responsibility for those interfaces.

The same care applies to market-size figures. The companies cite forecasts of 5–8 GW of Indian data-centre capacity by 2030, more than US$50 billion of planned hyperscaler spending in data-centre, cloud and AI ecosystems, and a global AI-compute infrastructure opportunity above US$3 trillion over seven years. Those figures describe a broad prospective market as presented in the release; they are not orders available to this venture, an addressable-revenue calculation or a committed share for either partner.

Two ventures, two different evidence trails

Adani already has a data-centre business whose public record gives a useful, separate reference point. Its FY2024–25 annual-report section on AdaniConneX reports 36.6 MW of operating capacity and a cumulative order book above 210 MW, alongside a target of a 1 GW platform by 2030. AdaniConneX is a distinct 50:50 venture with EdgeConneX. Its facilities, customers and operating figures cannot be assigned to the proposed Jabil partnership.

That comparison makes the evidence gap visible without implying that the new venture must follow the same path. Live megawatts and an order book say something different from an announced manufacturing capacity target. The new venture could eventually supply Adani-linked projects, outside data-centre operators, hyperscalers or all three. The October release names those market segments as intended customers but identifies no purchase order, qualification programme, delivery date or customer-specific product.

Jabil’s own scale is also not a financing commitment. In its 30 September FY2026 results release, the manufacturer described its results as preliminary and unaudited. It reported 21% revenue growth, a 40-basis-point expansion in core operating margin and US$1.532 billion of adjusted free cash flow; its supplemental data showed US$628 million of PP&E acquisition. These are company-wide measures. None is earmarked for the proposed venture, and the release provides no project budget or capital-contribution schedule. Jabil’s description of an asset-light model makes the question sharper: how would this particular platform allocate fixed plant, inventory and execution risk while preserving the economics of its wider manufacturing business?

Local demand policy is not a hardware order

India’s policy framework can support the wider data-centre ecosystem, but it should not be folded into the JV’s revenue case. The Finance Act 2026 provides a conditional exemption through the tax year ending 31 March 2047 for qualifying foreign-company income from procuring data-centre services from a specified data centre. The statutory definition and approval conditions concern data-centre services and the qualifying facility. The provision does not award an equipment-manufacturing contract, automatically exempt this proposed manufacturer, or require a cloud provider to buy its racks.

That separation is commercially important. Data-centre demand may attract new campuses, while a hardware venture still has to win a procurement slot against global suppliers, meet the buyer’s reliability and integration standards, secure components and price its products. Policy can alter the location economics of the service layer without deciding which manufacturer receives an order.

The conversion work is still ahead

The public case for the partnership is coherent: Jabil brings manufacturing and engineering experience; Adani brings infrastructure, energy and data-centre interests. Combining those capabilities could help a domestic platform address India’s and export markets’ equipment demand. But complementarity is a rationale for forming a venture, not evidence that the complements have been contracted, funded or connected to buyers.

The central unknown is not whether the two companies want a manufacturing platform. Their announcements say they do. It is how the platform would convert that ambition into controlled, financed and saleable output. The missing documents are therefore operational as much as corporate: final shareholder and governance terms; each party’s committed contributions; a named site and investment schedule; product-by-product capacity and ramp dates; customer qualification and purchase commitments; and allocation of design ownership, supply, installation, warranty and service liability.

Until those details appear, the 50:50 headline is best read as a proposed division of participation. It does not establish equal day-to-day authority, committed capital or a route from India’s data-centre forecasts to the venture’s factory orders. The next meaningful proof is a binding operating structure paired with funded capacity and identifiable demand.