Inside YTL and JLand's Gigawatt Data Center Plan

On August 19, 2026, YTL Power International and JLand Group announced a strategic partnership to build a new gigawatt-scale data center campus at Sedenak Tech Park West (STeP West) in Johor. If you have been watching how AI infrastructure is eating Southeast Asia's power grid, this is another significant piece of that puzzle — and it lands about 30 minutes from the Singapore border.

What Was Actually Announced

The deal is structured as a joint development between two subsidiaries: JLG Technopark Sdn Bhd (owned by JLand Group, the real estate arm of Johor Corporation) and SIPP Power Sdn Bhd (a 70%-owned subsidiary of YTL Power International).

Under the arrangement, SIPP Power will acquire approximately 58.6 hectares (~145 acres) within STeP West, with an option to acquire up to a further 162 hectares (~400 acres) for future expansion. If both options are fully exercised, the campus footprint reaches roughly 220 hectares (~545 acres).

The site sits inside Ibrahim Technopolis (IBTEC), a 2,951-hectare development spearheaded by the Johor government to support the Johor-Singapore Special Economic Zone. JLand Group is the master developer of IBTEC, and it is targeting RM40 billion in investments for STeP West alone.

The Numbers That Matter

The headline figure is capacity. The new Sedenak campus will complement YTL's existing Green Data Centre Park in Kulai, which has a planned capacity of up to 1.2 GW. Analysts covering the announcement put the potential numbers like this:

For scale: as of March 2026, YTL Power had around 298 MW of data center capacity contracted at Kulai, with plans to lift the park's capacity from 600 MW toward 1,000 MW. Management has guided that it expects to secure roughly an additional 200 MW of contracts annually. There is also chatter about a possible IPO of the data center business by 2027, with analysts floating valuations north of RM80 billion — comparisons were drawn to the DayOne DC IPO valuation of USD20 billion.

None of these are small numbers, and none of them are guaranteed. But the direction is clear: one company is assembling more than two gigawatts of planned digital infrastructure in a single Malaysian state.

Why Johor Keeps Winning These Deals

Johor's rise as a data center hub is not an accident. Three factors stack in its favor:

Power and land. Singapore effectively paused new data center construction for years because of its tiny footprint and power constraints. Johor has land, an expanding energy pipeline, and a government actively courting digital infrastructure investment. Water and electricity readiness was specifically called out for the Sedenak site.

Proximity. The IBTEC cluster sits within the Johor-Singapore Special Economic Zone. For workloads that need to be near Singapore — finance, low-latency regional services — but cannot afford Singapore power prices or real estate, Johor is the obvious answer. Fiber routes across the causeway make the latency story credible.

State backing. JLand Group manages assets worth over RM25 billion on behalf of the Johor state government. When the master developer of your tech park is state-owned, land deals move faster and the political risk drops.

What a Gigawatt Actually Means

Gigawatt gets thrown around so often in AI infrastructure news that it is easy to lose sight of what it means physically. A quick back-of-envelope calculation shows why these campuses take years to build:

# Rough sizing: how much GPU fits in a 1.2 GW campus?
TOTAL_POWER_KW = 1_200_000   # 1.2 GW campus
PUE = 1.25                   # power usage effectiveness target
IT_POWER_KW = TOTAL_POWER_KW / PUE

GPU_TDP_W = 1_200            # per accelerator module
OVERHEAD = 0.75              # networking, storage, idle headroom

usable_watts = IT_POWER_KW * 1000 * OVERHEAD
gpus = usable_watts / GPU_TDP_W

print(f"IT load:        {IT_POWER_KW:,.0f} kW")
print(f"Approx GPUs:    {gpus:,.0f}")

Run it and you get an IT load of 960,000 kW and room for roughly 600,000 accelerator modules at 1,200 W each. Even if you assume denser future hardware at double the efficiency, you are still provisioning for hundreds of thousands of devices — each needing network connectivity, cooling, and maintenance access. That is the physical reality behind "gigawatt-scale."

And the demand side is real. Training runs for frontier models consume clusters of tens of thousands of GPUs, and inference is becoming the bigger consumer as agentic workloads multiply. We covered recently how Cerebras is pushing ultrafast inference modes for GPT-5.6 — every token served out of those systems draws power somewhere, and increasingly that "somewhere" is Johor.

The Developer Angle

If you are building in Malaysia or the region, this matters beyond the business pages:

Co-location becomes viable locally. YTL's model is data center co-location — you bring the hardware or rent capacity, they provide power, cooling, and connectivity. A 2+ GW supply pool means less waiting list and potentially better pricing than flying your workloads to Singapore or Hong Kong.

Latency-sensitive architecture changes. When serious compute sits 300 km from Kuala Lumpur instead of 3,000 km away, decisions change. Self-hosted inference, edge caches, and hybrid architectures where sensitive data stays in-country all get easier to justify. Compliance-wise, keeping data on Malaysian soil also simplifies conversations about PDPA obligations.

The open-source agent ecosystem benefits too. Frameworks like DeepSeek's Harness for open-source agents assume cheap, abundant inference. Regional compute supply is one of the inputs that determines whether that assumption holds here.

What to Watch

Three things will tell us whether this deal delivers:

1. Contracted capacity conversion. Planned gigawatts mean nothing until tenants sign. Watch whether YTL keeps hitting its ~200 MW/year contracting pace.
2. Power execution. A 1.3–1.5 GW campus needs serious generation and grid upgrades. YTL owns generation assets, which is exactly why it is well positioned — but execution risk remains.
3. The IPO. If the data center unit lists by 2027 anywhere near analyst estimates, expect a wave of copycat projects across Johor and possibly other Malaysian states.

Southeast Asia's AI buildout is turning into a land-and-power game, and this partnership is one of the clearest signals yet that Malaysia intends to play it at scale.