Mapletree Industrial Trust

[heyjojos]

VB, still learning how to value REITS.  MIT FY22/23 announcement just came out 

In looking at the P/L, it indicated a $110.632M net loss from valuation.  Should we be concern on this?  Seems excessive??  What I'm not understanding is in the Portfolio Valuation (page 12 of Result Slides), it actually show an increase instead.  Anyone care to enlighten?

Also puzzled by the non-tax deductible is it also usual to see large fluctuations $46,441K (for FY22/23) vs -$107,018K (for FY21/22): see page 6 of Financial Result.  The footnotes says its valuation loss/gain property under development.  Do we normally expect such fluctations?

Appreciation any clarification. Thanks.

J

[weijian]

VB, still learning how to value REITS.  MIT FY22/23 announcement just came out 

In looking at the P/L, it indicated a $110.632M net loss from valuation.  Should we be concern on this?  Seems excessive??  What I'm not understanding is in the Portfolio Valuation (page 12 of Result Slides), it actually show an increase instead.  Anyone care to enlighten?

hi heyjojos,

Net loss from valuation I have observed that most REITs are recording a FVTPL loss due to increase in capitalization rates as a result of increase in risk-free rates in recent times. This is definitely not unique to MIT. But whether is this excessive or not, I do not have a good idea. If one has a good idea of cap rates, then one can compare what MIT is using vs relevant benchmarks. For example, HongKongLand is valued at <50% NAV. Could it be due to its super low cap rates? And whether should one be concern or not, I can only say this is part of the property cycle.

Portfolio Valuation Actually page12 seems to have given the answer and I will paste it as below:
The marginal increase in Singapore Portfolio valuation was largely attributed to the completion of 161, 163 and 165 Kallang Way
So while there is negative adjustment in FV for the properties, there is asset enhancement/development work that is been capitalized onto the balance sheet. From page24 of FY results, one can see that 124mil was added and then transferred to investment properties. And this money was largely funded through proceeds from the DRP (page33). As this AEI amount is larger than the FV loss, it may explain the the overall YoY increase in portfolio valuation.

As for the "Non-tax deductible/(chargeable) items" variations, unfortunately, I have not really look deep (nor invest in REITs), so I do not have a ready answer for that.

[heyjojos]

VB, still learning how to value REITS.  MIT FY22/23 announcement just came out 

In looking at the P/L, it indicated a $110.632M net loss from valuation.  Should we be concern on this?  Seems excessive??  What I'm not understanding is in the Portfolio Valuation (page 12 of Result Slides), it actually show an increase instead.  Anyone care to enlighten?

hi heyjojos,

Net loss from valuation I have observed that most REITs are recording a FVTPL loss due to increase in capitalization rates as a result of increase in risk-free rates in recent times. This is definitely not unique to MIT. But whether is this excessive or not, I do not have a good idea. If one has a good idea of cap rates, then one can compare what MIT is using vs relevant benchmarks. For example, HongKongLand is valued at <50% NAV. Could it be due to its super low cap rates? And whether should one be concern or not, I can only say this is part of the property cycle.

Portfolio Valuation Actually page12 seems to have given the answer and I will paste it as below:
The marginal increase in Singapore Portfolio valuation was largely attributed to the completion of 161, 163 and 165 Kallang Way
So while there is negative adjustment in FV for the properties, there is asset enhancement/development work that is been capitalized onto the balance sheet. From page24 of FY results, one can see that 124mil was added and then transferred to investment properties. And this money was largely funded through proceeds from the DRP (page33). As this AEI amount is larger than the FV loss, it may explain the the overall YoY increase in portfolio valuation.

As for the "Non-tax deductible/(chargeable) items" variations, unfortunately, I have not really look deep (nor invest in REITs), so I do not have a ready answer for that.

Thanks for your reply.  Have a better understanding now.  

[weijian]

It hasn't been fun to be a MIT shareholder and so I thought it would be interesting to take a peek at AR25/26 annual report. For FY25/26, NPI and DPU have dropped 5-6% YoY and largely attributed to (a) absence of income from Spore Portfolio divestments, (b) lease non-renewals + FX impact on North American Portfolio. Checking a little bit of history of its North American Portfolio:

Dropping occupancy rates Coming out of covid, the occupancy rate has dropped from ~93% (FY22/23) to the current 86.1% as end FY25/26 for its portfolio of 55 data centers in North America. The explanation from Mgt sounds worrying because it suggests that some of these data centers maybe becoming "technologically obsolete". The occupancy rate will continue to get worst because another 5.1% of leases (based on gross rental) will expire in the coming year and >90% of these are not renewing (~half of this non renewals is coming from a single tenant that is going to vacate an entire building). 

Origin of MIT's North America portfolio Interestingly, this portfolio of assets were largely only injected into MIT via 3 acquisitions during a period of 2 years (2020 - 2021). The sellers were 2 US REITs (Digital Realty and Sila Realty) + Sponsor themselves. Post acquisition, the North America portfolio would transform MIT into the biggest AP REIT holder of data centers and it would account for ~50% of its total assets. When I read back the ppt slides of the acquisitions, they were touting a longer WALE post acquisition as one of the reasons for buying. A few years down the road, the shorter leases in these acquisitions are starting to show via non-renewals. On hindsight, 2020-2021 was a good time for insiders (US REITs) to sell to foreigners with low interest rates + committed leases (at least for the next 3-4years).

MAPLETREE INDUSTRIAL TRUST ANNUAL REPORT 2025/2026

In FY25/26, MIT operated against a backdrop of portfolio transition and external headwinds, which affected our
financial performance. Net property income for FY25/26 decreased by 5.9% year-on-year to S$500.4 million. This was mainly due to the absence of income from the Singapore Portfolio Divestment, the non-renewal of leases in the North American Portfolio and the depreciation of USD against SGD.

----------------------------

The North American Portfolio recorded an average occupancy rate of 87.4% in FY25/26, compared with 89.3% in FY24/25. The decline was mainly due to a tenant’s downsizing of office space at the multi-tenanted property at 250 Williams Street NW, Atlanta, as well as the nonrenewal of leases at 400 Holger Way, San Jose and 2005 East Technology Circle, Tempe.

Looking ahead, Data Centres in North America account for approximately 5.1% of leases due for expiry in FY26/27 (by gross rental income), of which about 4.7% have been confirmed as non-renewals. Near-term leasing challenges are primarily associated with data centres leased to enterprise users, which are mostly located outside primary data centre markets and have configurations less optimised for data centre use. Notably, the majority of leases with enterprise users have already undergone their initial renewal cycles, which provided greater visibility and certainty over upcoming lease expirations.

AR25/26: https://links.sgx.com/FileOpen/20260622_...eID=893762

[weijian]

Shareholders have queried more about the underperformance of its North American data center portfolios and the reply reminds me of what I learnt in REITS 101 (in italics) - Industrial assets are generally more risky compared to generic use assets like commercial (office/retail) due to the risk of obsolesce. 

So one gets to enjoy the relatively high IRRs during its early life cycle and the pain comes when one is still holding the baby towards the end of its life cycle.

RESPONSES TO THE SUBSTANTIAL AND RELEVANT QUESTIONS FROM UNITHOLDERS FOR THE ANNUAL GENERAL MEETING ON 21 JULY 2026

Q1. Referring to the “Data Centres – Detailed Property Information” section (page 51 of the Annual Report 2025/2026), the valuation of 7337 Trade Street, San Diego declined significantly year-on-year from approximately US$151.0 million to
US$49.2 million as at 31 March 2026, despite maintaining the average occupancy rate of 100.0%. Could the Manager provide further insight into the key factors that contributed to this substantial reduction in valuation?

Ans: The year-on-year valuation decline of 7337 Trade Street, San Diego was primarily attributable to the valuer’s adoption of the sales comparison approach, which reflected property-specific risks arising from the non-renewal of lease. In addition, the property’s highest and best use was reassessed to an industrial land development site, which also contributed to the valuation decline. The lease for this property was originally due to expire in 2023 but was subsequently extended twice,
maintaining the average occupancy rate of 100.0% in FY25/26. The tenant confirmed that it would not renew the lease, which expired in May 2026.

Q2. The properties located at 2000 Kubach Road (Philadelphia), 5000 South Bowen Road (Arlington) and N15W24250 Riverwood Drive (Pewaukee) reported an average occupancy rate of 0% in FY25/26. Could the Manager elaborate on the key challenges that have prevented MIT from securing tenants for these data centres?

Ans These three properties became vacant as enterprise tenants did not renew their leases due to business consolidation. Securing replacement tenants has been challenging as these properties may not be efficiently configured for full data centre use. In June 2026, the Manager completed the divestment of 2000 Kubach Road, Philadelphia for US$14.5 million, which represented a 4.3% premium to its independent valuation. Since the expiration of its lease on 31 December 2024, leasing interest in this property has been limited. Repositioning or redevelopment of this property presents challenges, including limited power availability, layout constraint, as well as construction and execution risks. Following a thorough evaluation, the Manager concluded that divestment was the most prudent course of action to optimise capital allocation and redeploy capital into markets and assets with sustainable growth potential.

AGM Q&A: https://links.sgx.com/FileOpen/20260716_...eID=896631

[specuvestor]

Find it interesting that this lack of demand is happening in a data centre and capex infrastructure boom environment, which speaks a lot on the quality of assets which KDC REIT previously warned about

[swakoo]

Think the main issue faced by MIT in finding new tenants for it's DCs is that these are the older generation DCs which cannot meet the specs needed of newer generation (eg. AI-ready) DCs. And it is very difficult/costly to repurpose them.

So in spite of the mega trillion dollars projected for (AI-ready) DC capex in coming years, DCs are not generic nor homogenous and legacy ones may not be fit for purpose for AI workloads.

According to Claude:

Repurposing data centers is possible but rarely easy — it depends heavily on what you're converting from and to.

**What makes it hard:**

- **Power infrastructure** is usually the biggest constraint. A facility built for a certain power density (say, 5-10 kW per rack for traditional enterprise IT) can't easily support modern AI/GPU workloads, which often need 40-100+ kW per rack. Upgrading electrical distribution, transformers, and backup generation is expensive and sometimes physically impossible within the existing building envelope.

- **Cooling systems** are similarly load-specific. Air-cooled facilities designed for older equipment often can't handle the heat density of modern GPU clusters, which increasingly require liquid cooling. Retrofitting liquid cooling into a facility built for air cooling means new piping, pumps, and sometimes structural changes.

- **Floor loading and structural capacity** matter more than people expect — high-density racks and liquid cooling infrastructure are heavier, and older buildings may not be rated for it.

- **Network connectivity** — a facility built as a corporate colocation site may lack the fiber connectivity or carrier-neutral setup needed to become a cloud or hyperscale facility.

**What transfers well:**

- Physical security, access control, and basic building shell
- Fire suppression systems (though may need upgrading)
- General site attributes — land, substation proximity, water access — that made the location worthwhile in the first place
- Some racking and floor space, for lower-density workloads

**Common conversions in practice:**

- Legacy enterprise data centers → colocation facilities (moderate difficulty, mainly commercial/operational changes)
- Older data centers → lower-tier archival or backup facilities (relatively easy, since demands are lower)
- Traditional data centers → AI/GPU-optimized facilities (hardest conversion — often more economical to build new than retrofit)
- Industrial buildings/warehouses → data centers (increasingly common, but needs heavy infrastructure investment)

The general pattern: repurposing *down* in power/cooling density is much easier than repurposing *up*. This is part of why so much current data center capacity is being built new rather than retrofitted — the jump to AI-era power and cooling requirements is large enough that many older facilities become effectively obsolete for high-performance computing, even though they might live on for lower-intensity storage or general enterprise use.

[specuvestor]

Yes you need high powered DC for AI workload in training etc

Yet there's supposed to be flood of information exchanges or "distilled" data or streams to AI enabled laptops that doesn't require such power.

This is also the reason why some pundits worry about the US$1T capex on AI infrastructure when they are no longer shining swanky anymore in 10 years ie depreciating at $100b clip a year

It was not easy to spot the Amazon or Google in 2002

Think the main issue faced by MIT in finding new tenants for it's DCs is that these are the older generation DCs which cannot meet the specs needed of newer generation (eg. AI-ready) DCs. And it is very difficult/costly to repurpose them.

So in spite of the mega trillion dollars projected for (AI-ready) DC capex in coming years, DCs are not generic nor homogenous and legacy ones may not be fit for purpose for AI workloads.

According to Claude:

Repurposing data centers is possible but rarely easy — it depends heavily on what you're converting from and to.

**What makes it hard:**

- **Power infrastructure** is usually the biggest constraint. A facility built for a certain power density (say, 5-10 kW per rack for traditional enterprise IT) can't easily support modern AI/GPU workloads, which often need 40-100+ kW per rack. Upgrading electrical distribution, transformers, and backup generation is expensive and sometimes physically impossible within the existing building envelope.

- **Cooling systems** are similarly load-specific. Air-cooled facilities designed for older equipment often can't handle the heat density of modern GPU clusters, which increasingly require liquid cooling. Retrofitting liquid cooling into a facility built for air cooling means new piping, pumps, and sometimes structural changes.

- **Floor loading and structural capacity** matter more than people expect — high-density racks and liquid cooling infrastructure are heavier, and older buildings may not be rated for it.

- **Network connectivity** — a facility built as a corporate colocation site may lack the fiber connectivity or carrier-neutral setup needed to become a cloud or hyperscale facility.

**What transfers well:**

- Physical security, access control, and basic building shell
- Fire suppression systems (though may need upgrading)
- General site attributes — land, substation proximity, water access — that made the location worthwhile in the first place
- Some racking and floor space, for lower-density workloads

**Common conversions in practice:**

- Legacy enterprise data centers → colocation facilities (moderate difficulty, mainly commercial/operational changes)
- Older data centers → lower-tier archival or backup facilities (relatively easy, since demands are lower)
- Traditional data centers → AI/GPU-optimized facilities (hardest conversion — often more economical to build new than retrofit)
- Industrial buildings/warehouses → data centers (increasingly common, but needs heavy infrastructure investment)

The general pattern: repurposing *down* in power/cooling density is much easier than repurposing *up*. This is part of why so much current data center capacity is being built new rather than retrofitted — the jump to AI-era power and cooling requirements is large enough that many older facilities become effectively obsolete for high-performance computing, even though they might live on for lower-intensity storage or general enterprise use.

[swakoo]

Yes you need high powered DC for AI workload in training etc

Yet there's supposed to be flood of information exchanges or "distilled" data or streams to AI enabled laptops that doesn't require such power.

This is also the reason why some pundits worry about the US$1T capex on AI infrastructure when they are no longer shining swanky anymore in 10 years ie depreciating at $100b clip a year

It was not easy to spot the Amazon or Google in 2002

Yes, a lot of moving parts on how AI unfolds in coming years.

Recently come across this well written analysis of how AI should play out and which AI layers may be more important. Just the author's opinion but supported with sound analogies and makes an interesting read (in trying to spot potential winners).

The AI Economy: The Next Chapter

[weijian]

My wife bought a sack of durians last night and some of them were bad. Even though I am particular with how durians should taste, I decided to forgive her because she is after all....my wife. The problem of those bad durians obviously lies with the dishonest durian seller.

BUT if my CEO (of my vested holdings) tells me that lemons are inevitable when they bulk purchase oranges, it is simply not good enough. After all, there looks to be a decent amount of lemons than the occasional odd one.

MINUTES OF PROCEEDINGS OF 16th ANNUAL GENERAL MEETING

Proxy Mr. J G Thurlow had the following query: He asked whether an independent technical assessment was conducted prior to an acquisition (in terms of the existing power availability, upgrade and expansion potential, cooling capabilities, and the suitability of the data centre assets to support hyperscale customers and evolving artificial intelligence (“AI”) workloads) and which key assumptions underpinning the acquisition had proven to be incorrect.\

Response: Ms. Ler Lily explained that Management's assessment of acquisition opportunities considers the underlying attributes of the assets, including lease quality, tenant profile, portfolio risks and the sustainability of cash flows. While a portfolio acquisition may inevitably include certain assets that subsequently underperform or are less resilient than initially expected, acquisitions are evaluated on an overall portfolio basis, with the objective of ensuring that the portfolio as a whole generates sustainable cash flows and delivers accretive growth. As part of the acquisition process, independent technical due diligence was conducted by external consultants, alongside Management's commercial and financial assessments.

MoM: https://links.sgx.com/FileOpen/20260819_...eID=901755