Let's deep dive into exactly how this portfolio is structured, and why its biggest segment is now facing an unexpected uphill battle.
Section 1: Types of Data Centres in MIT's Portfolio
MIT holds 55 data centres across North America (spanning major tech hubs like Northern Virginia, California, and North Carolina), alongside data centre assets in Singapore and Japan (such as its newly fitted Osaka Data Centre).
To serve a diverse client base, MIT categorizes its data centre assets into three primary physical and operational structures:
- Hyperscale Data Centres: These are massive, high-capacity facilities designed to support large-scale computer applications. MIT structures these as fully fitted facilities.
- Target Customers: Large cloud service providers and massive technology giants (colloquially known as "Hyperscalers" like AWS, Microsoft, or Google) who require immense scalability and computing infrastructure.
- Powered Shell Data Centres: These are buildings that are physically constructed with robust utility connections, fiber optic infrastructure, and power capacity routed to the property line, but the internal fit-outs (such as cooling units and server racks) are left blank.
- Target Customers: Established data centre operators and large enterprise tenants who prefer to install their own custom technology stacks, hardware, and operational designs while letting the REIT handle the physical real estate.
- Colocation & Mixed-Use Data Centres: Multi-tenanted facilities where space, power, and cooling infrastructure are shared among smaller users.
- Target Customers: Corporate enterprise clients, healthcare providers, financial institutions, and telecommunications firms that need reliable data hosting but do not operate at the scale of a tech giant.
Section 2: Why MIT's North America Data Centres Are Becoming Its Achilles Heel
If you read the news, artificial intelligence is driving an insatiable global appetite for server space. Yet despite this undeniable tailwind, MIT’s North American portfolio has paradoxically turned into its primary operational drag, causing its overall Distribution per Unit (DPU) to slide by 6.3% year-on-year to 12.71 cents for the latest fiscal year.
The issues come down to specific property dynamics and a structural mismatch between older assets and new AI requirements:
1. The Power Bottleneck and Smaller, Legacy Footprints
AI training workloads require massive, highly advanced data centres with specialized cooling systems and unprecedented power allocations. Many of MIT’s North American facilities are smaller, legacy enterprise spaces or powered shells. When major leases expire, backfilling or converting these properties is incredibly difficult because upgrading power networks involves massive construction risks and years of grid approval lead times. For instance, following the departure of a major tenant (Vanguard) at its Philadelphia facility, MIT faced minimal leasing interest due to these exact power and execution limitations, ultimately choosing to divest the asset.
2. Confirmed Tenant Non-Renewals and Downsizing
The portfolio is wrestling with major tenancy transitions. Secular shifts have led to corporate downsizing and lease terminations. Most recently, a major tenant downsized its footprint at 250 Williams Street in Atlanta, and the trust faces a succession of upcoming non-renewals across secondary clusters. Because finding new, niche tenants takes months of downtime, these vacancies hit the bottom line immediately.
Key Portfolio Metrics to Note:
- Occupancy Rate: Driven by these lease expiries and space downsizing, the average occupancy rate for the North American portfolio has fallen to 86.1%. This stands in stark contrast to its highly resilient Singapore portfolio (93.4% occupancy) and its Japan data centres (100% occupancy).
- Revenue Contribution: The North American portfolio represents a massive chunk of MIT's business, making up approximately 46.5% of its total geographic AUM. Because it commands nearly half of the trust's structural asset value, the lower occupancy and weaker rental reversions in the US exert outsized downward pressure on total rental performance.
Section 3: Pure MIT Bad Luck?
Sadly, MIT finds itself in a classic real estate conundrum: being in the right sector, but caught with some of the wrong property types for the current technological shift. While the boom in AI is very real, it is highly selective—favouring newer, massive, power-optimized hyperscale facilities over older legacy layouts.
Nevertheless, all is not lost. The good news here is that the management of MIT isn't sitting on its hands. They are actively recycling capital—selling underperforming US sites at premiums to valuation and pivoting towards modern data centre growth in Japan and Europe. Combined with a robust, highly stable industrial portfolio back home in Singapore, MIT has the financial buffer to navigate this bumpy patch.
Parting Thoughts
Personally, my thoughts are that for long-term income investors, the key will be watching how fast MIT can plug the vacancy leaks in North America and shift capital into high-power digital real estate.
Of course, the key risk here is that by the time MIT fixed its North America data centre portfolios, the AI boom would have already gone to bust. I am currently still holding on to my MIT units and observing the execution by MIT.
Ok folks, that's all from me today. Have a great week ahead!
