Summary

  • Mercury paid US$30 million, or NZ$53 million, for 12.7% of Datagrid NZ and placed chief executive Stew Hamilton on its board.
  • The investment follows a 140MW power purchase option agreement. It creates a route towards a large electricity sale, not evidence that 140MW is being delivered or consumed today.
  • Official sources describe the planned campus as up to 240MW of IT capacity, 280MW of hyperscale capacity and a 360MW project. No public reconciliation explains the different scopes.
  • Resource consent and August site works are real execution milestones, while final investment decision, customer commitments, option commencement, substation energisation and metered demand remain the tests that matter.

Four megawatt numbers, four different claims

The most useful fact about Datagrid NZ’s planned campus is not a single capacity number. It is that four official numbers now surround the same project.

Environment Southland says the consented data centre may provide up to 240MW of IT capacity. Datagrid’s current website advertises 280MW of hyperscale capacity across three modules. Mercury’s July investment announcement calls it a 360MW project. A separate agreement between the two companies concerns a 140MW power purchase option.

Those numbers may describe different boundaries or different versions of the development. IT load can exclude cooling and other facility demand. A project envelope can exceed an initial build. A commercial option can cover only part of a campus. But the reviewed public sources do not publish the bridge between the figures.

That omission matters because megawatts are doing several jobs at once in the data-centre market. They describe engineering scale, advertise ambition, reserve grid attention, support fundraising and suggest future revenue. When one unit carries that many meanings, the market has to ask what each number controls.

The 140MW figure is especially easy to misread. It belongs to a contractual pathway, not to a meter at a working campus.

What Mercury actually bought

On 23 July, Mercury said it had invested US$30 million, or NZ$53 million, for a 12.7% minority stake in Datagrid Holding Group NZ Ltd. The investment came from existing capital facilities. Mercury chief executive Stew Hamilton also became a Datagrid NZ director.

The equity is small beside the multi-billion-dollar scale associated with the full development. Its strategic value lies elsewhere.

Mercury has moved closer to the point where Datagrid’s electricity demand will be defined. A board seat can improve visibility into financing, customer discussions, construction sequencing and the timing of power requirements. It can also give Mercury a voice in governance before a possible data-centre load becomes one of the largest new users on the South Island.

Mercury described the investment as staged, with governance and downside protection. The public announcement did not disclose the detailed shareholder rights, valuation method, dilution protections or exit terms. A minority position is not control. It is, however, more information and more influence than an arm’s-length electricity supplier normally receives.

Datagrid gets something different. The investment funded continued development and the start of horizontal works. It also puts a large renewable generator beside the project during conversations with lenders and prospective customers. A data-centre developer can show not merely an intention to find power later, but an existing commercial relationship with a utility that has capital and generation projects of its own.

This is why the transaction is more than a passive share purchase. It joins the formation of electricity demand to the planning of electricity supply.

The option is a path to a customer, not the customer itself

Datagrid and Mercury announced their 140MW Power Purchase Option Agreement in March. Datagrid described it as a 15-year arrangement covering 1.2TWh a year and said it provided long-term price certainty. Mercury’s later disclosures kept the more cautious label: a power purchase option that provides a pathway for renewable supply and enables possible new demand.

The distinction is substantive.

An option can be valuable before electricity flows. It can reserve a commercial route, reduce exposure to future prices or establish conditions under which a longer power purchase arrangement begins. That can help a developer model costs and speak to customers. It can help a generator assess whether new demand may support generation investment.

But an option is not the same as a commissioned substation, a live data hall or 140MW of metered consumption. The public material reviewed for this article does not disclose the exercise conditions, strike price, commencement date, volume ramp, credit support, curtailment allocation or termination rights. It also does not say that the option has been exercised.

The right conclusion is neither that the agreement is empty nor that the load is secured. It is a real contractual coordination device whose operating outcome remains conditional.

The scale illustrates why those conditions matter. Datagrid equates 1.2TWh a year with about 3% of national electricity demand. Mercury said three renewable projects that began generation in 2026 would together add about 1.1TWh of annual output. The quantities are not directly interchangeable: annual energy differs from instantaneous capacity, generation is delivered through a portfolio and a grid, and output varies by asset. The comparison nevertheless shows that the option is large enough to shape capital planning if it turns into sustained demand.

Why a generator wants to see demand earlier

New generation and new data centres face opposite versions of the same coordination problem.

A generator wants evidence of durable demand before committing capital to wind, geothermal, hydro or storage. A data-centre developer wants a credible power path before lenders and compute customers commit to the campus. Each side can wait for the other and make the project slower or more expensive.

Mercury’s structure narrows that gap. The equity gives it exposure and information. The board seat gives it governance access. The power option creates a commercial path to offtake. None of the three is sufficient alone, but together they let the companies plan against a shared set of contingencies.

Mercury states the logic plainly. Long-duration demand from users such as Datagrid can provide the revenue certainty that underpins new renewable developments. Its August results placed the newly approved NZ$506 million Puke Kapo Hau wind project in the context of South Island demand, including Datagrid.

That does not make Puke Kapo Hau a dedicated Datagrid power station. The wind project, the existing generation fleet, transmission constraints and other customers all form part of Mercury’s portfolio. It does show that the prospective data-centre load has moved from a developer’s presentation into a listed generator’s capital narrative.

For investors, that is progress in institutional commitment. It is not yet progress in operating consumption.

Earthworks prove activity, not usable compute

Datagrid announced on 18 August that horizontal works had begun at the 49-hectare Makarewa site. HEB Construction is building access works, upgrading roads, moving topsoil and constructing a six-metre bund. Datagrid expects foundation platforms for the future data-centre facility, cable landing station and power substation by year-end.

That is more than a rendering. Contractors are on site and physical capital is being placed.

It is also well short of an operating campus. The announcement concerns preparation for future structures. It does not establish that data halls have been vertically built, that the dedicated substation is energised, that compute equipment is installed or that a customer is paying for capacity.

The sequencing is noteworthy because Mercury’s July announcement said a final investment decision for the project was expected later in 2026. The August works release did not announce that a campus FID had been taken. Development expenditure and enabling works can proceed before the financing and construction commitment for a full campus is complete.

That is not inherently a warning sign. Early site works can protect a schedule and retire practical risks. The discipline is to label the stage correctly. A road is a road. A foundation platform is preparation. Neither is usable capacity.

Resource consent should be read the same way. Environment Southland records approvals concerning wastewater, air discharges from diesel generators, groundwater, wetland effects and the coastal data cable. Southland District Council issued land-use consent for the hyperscale facility and a grid exit point substation. These approvals remove regulatory obstacles. They do not supply capital, customers or electricity.

The capacity gap is a disclosure problem, not a puzzle to guess away

It would be easy to reconcile 240MW, 280MW and 360MW by inventing a neat hierarchy: IT load, initial facility size and ultimate grid envelope. That explanation may prove correct. It is not yet published in the source set.

Markets should not have to infer the definition of the largest number in a capital-intensive project.

Datagrid can improve the information surface by publishing a simple capacity bridge: consented IT capacity, total facility demand at each phase, contracted or optioned power, connection capacity, module timing and ultimate development envelope. The figures would not need to reveal customer identities or commercial prices. They would need to make like-for-like comparison possible.

Mercury has an interest in the same clarity. Its shareholders need to distinguish a strategic minority investment from contracted electricity revenue. Grid planners need to distinguish a possible future connection from a forecast load. Communities need to distinguish a development envelope from the site’s first operating phase.

Transparent definitions do not weaken a project. They make it easier to price.

A closer relationship moves risk; it does not erase it

The positive case for the transaction is credible. Mercury gains an early view of a rare long-duration demand opportunity. Datagrid gains capital and a sophisticated power counterpart. Each company can make decisions with better information about the other side.

The risk also becomes more connected.

If Datagrid’s customer demand arrives later or at a smaller scale, Mercury may have committed equity and management attention before electricity revenue appears. If generation, transmission or substation delivery is late, Datagrid may have a consented site and construction expenditure without the usable power needed for AI workloads. If the option terms allocate risk poorly, a delay on one side can become a financing problem on the other.

This is why the transaction should be understood as staged coordination, not vertical certainty. Its success will be visible through a chain of separate gates: campus FID, financing close, credible customer commitments, option commencement, substation construction, transmission readiness, generation delivery, commissioned data halls and finally metered load.

Mercury has bought a position beside that chain. It has not bought proof that the last gate is open.

Sources