Plans to build New Zealand’s first large-scale AI data center in Makarewa, a farming district near Invercargill in Southland, have drawn local opposition over concerns about electricity consumption, groundwater use, and reliance on diesel backup generators.
The dispute reflects a broader question facing New Zealand: where should an island nation with limited suitable land for large-scale digital infrastructure build its next generation of AI data centers?
Samsung Heavy Industries believes it has an unconventional answer. The South Korean shipbuilder plans to commercialize a floating AI data center platform by 2028, using purpose-built barges cooled by seawater instead of conventional land-based facilities.
Samsung has not announced any plans for New Zealand. Even so, the concept addresses several pressures already emerging locally, including limited development sites, growing demand for electricity, and increasing community scrutiny of projects that place greater demands on land and natural resources.
Rethinking where AI data centers get built
Samsung’s proposal represents more than a different way to house servers. It reflects a broader shift in how the industry is responding to AI’s rapidly growing computing requirements.
Instead of constructing another conventional campus on valuable industrial land, the company is designing floating platforms that can operate close to shore while drawing power from existing electricity grids. Seawater would provide the primary cooling source, reducing reliance on conventional cooling systems.
The approach is intended to ease three major constraints facing AI infrastructure: limited land availability, lengthy permitting processes, and the rising cost of cooling increasingly power-hungry AI hardware. Samsung Heavy Industries CEO Choi Sung-an has described floating data centers as a major opportunity for the shipbuilding and offshore industries, with the company already holding discussions with prospective customers ahead of its planned launch.
Why land constraints matter in New Zealand
New Zealand is actively positioning itself as a destination for AI and hyperscale data center investment, with Invest New Zealand seeking billions in new investment by promoting the country’s renewable electricity, cool climate and political stability.
Securing those investments, however, involves more than attracting capital. Large AI facilities require access to high-capacity electricity transmission, fiber connectivity, transport infrastructure and environmental approvals, while also gaining community support. For a country with relatively limited developable land near that infrastructure, those competing demands are becoming increasingly difficult to balance.
Samsung’s floating concept does not eliminate every challenge. Still, it offers a different approach by shifting part of the infrastructure footprint offshore rather than competing for increasingly valuable land-based sites.
More about data centers
Southland shows that those pressures are already emerging
The debate surrounding Datagrid’s proposed AI campus in Makarewa illustrates the trade-offs New Zealand is beginning to confront.
Supporters say the NZ$3.5 billion development would strengthen the country’s AI and cloud computing capabilities while creating an estimated 1,200 construction jobs and around 50 permanent positions. Once operational, the facility is expected to consume up to 280 megawatts of electricity, making it New Zealand’s second-largest electricity user after the Tiwai Point aluminum smelter.
Critics, meanwhile, have questioned the project’s projected electricity demand, groundwater consent allowing withdrawals of up to 604,800 liters a day, approval for as many as 84 diesel backup generators, and what they describe as limited public consultation before key approvals were granted. The project has become one of New Zealand’s clearest examples of how AI infrastructure can quickly become as much a community issue as a technology investment.
Cooling is becoming a defining AI infrastructure challenge
Cooling is rapidly becoming one of the biggest engineering and operational challenges facing AI data centers as increasingly powerful processors generate substantially more heat than traditional enterprise servers.
Samsung is betting that direct seawater cooling can reduce both operating costs and pressure on freshwater resources in suitable coastal locations.
New Zealand already holds one natural advantage. Southland’s average annual temperature of about 10°C reduces the amount of active cooling required, which helps Datagrid with water consumption despite its approved groundwater allocation. Even so, water use has remained one of the project’s most closely watched issues, highlighting how environmental considerations are becoming central to decisions about future AI infrastructure.
Where the next chapter in AI infrastructure points to
Samsung’s proposal is ultimately less about floating platforms than about the direction of AI infrastructure itself. As demand for computing capacity accelerates, developers are moving beyond conventional approaches in search of faster, more efficient, and less contentious ways to build.
For New Zealand, the next competitive advantage may not simply be having space for AI infrastructure, but being ready for how that infrastructure is changing.
Read the full article here
