S&P Puts a US$20bn Price on Malaysia's Next Two Gigawatts of Data Centres
The ratings agency says domestic banks cannot carry the build-out alone, and Johor's efficiency rules are already deciding which projects get built.
S&P Global Ratings estimates that Malaysia's data centres will need more than US$20bn of funding over the next three years. That figure covers only powered shells and equipment (the building, power and cooling, networking and storage) for about two gigawatts of new capacity, The Star reported on 20 July. Chips come on top, and S&P puts them at between one and four times the cost of the powered shell.
The headline number is large, but the more important point is where it collides with the banking system. S&P expects the funding need to exceed domestic banks' sector concentration limits, which it assumes add up to about US$30bn. It also says that a fresh US$30bn of headroom might still not cover the projects plus the silicon. Until now, data centre projects have mostly been funded with bank loans and equity. S&P's view is that this will not be enough for long.
From boom to reset
S&P frames the market as moving from rapid expansion to slower, more sustainable growth; its credit analyst Spencer Ng described the sector as being in a reset mode, TechNode Global reported. Alongside the funding gap, S&P expects data centres to account for around 31% of Malaysia's electricity demand by 2035, up from 7% now, and warns that delays in power or water infrastructure could bottleneck growth.
Johor is where the regulation is tightest. According to The Edge's report on the S&P note, the state has halted approvals for Tier 1 and Tier 2 data centres that are significantly less water and power efficient. In other words, Johor is filtering by design quality and not only by megawatts.
S&P is still bullish on the long run. It expects Malaysia to nearly triple capacity by 2030. It notes that Malaysian power costs are now slightly above the Southeast Asian average, but it does not expect demand to move to neighbouring countries on cost alone, given proximity to Singapore, connectivity and growing utility infrastructure.
Why a software team should care about project finance
It is tempting to treat this as a story for bankers. It is not, for three reasons.
First, the expensive part is the chips, not the shells. If the shell alone needs US$20bn and accelerators add one to four times that, GPU capacity is where financing is thinnest. Malaysian builders hoping for cheap, local, in-country inference capacity should not assume the build-out produces it automatically. Much of the new capacity is being built for large tenants with long contracts. Local GPU capacity priced for smaller teams is a different product, and its financing will be the hardest of all.
Second, alternative capital comes with alternative terms. Private credit, infrastructure funds and capital markets price risk differently from a domestic bank with a relationship to protect. Higher financing costs flow into colocation and cloud pricing eventually, alongside the tariff increases operators are already absorbing.
Third, Johor's efficiency rules make it more likely the capacity that does get built is modern, efficient and AI-capable. That is good for the grid and for water. It also means fewer cheap, older halls, which is where small hosting providers and regional SaaS companies have traditionally found affordable racks.
What is unclear
The US$30bn concentration figure is S&P's assumption, not a published regulatory limit. How Bank Negara and the banks treat data centre exposure from here is the real variable. It is also unclear how much of the two gigawatts is already financed and how much is still on slides. A pipeline measured in announcements always looks bigger than one measured in signed power agreements.
What I will be watching: data centre sukuk and bond issuance, any private credit funds set up specifically for Malaysian builds, and whether tighter money slows the Johor pipeline faster than the regulators do.
Sources