Aolani Commits to 500 MW AI Data Center Capacity in Malaysia by H2 2027
The global AI infrastructure build-out faces increasing power and funding constraints, even as new data center capacity commitments emerge in Asia.
The story
The rapid expansion of AI infrastructure in the United States is encountering significant headwinds, with a new Morgan Stanley analysis predicting a 32-gigawatt power shortfall for US data centers by 2028, a gap that could disproportionately affect smaller enterprises. This power deficit is exacerbated by existing bottlenecks, as nearly half of the US data centers planned for 2026 are already facing delays or cancellations due to equipment shortages and insufficient grid capacity.
The Financial Times reported on October 18 that America's AI build-out is struggling with funding pressures, construction setbacks, and local opposition, with a regulatory intelligence firm counting at least 47 major projects delayed or blocked in 2026. Underscoring the economic implications, BlackRock CEO Larry Fink warned on October 17 that a nationwide moratorium on data center construction in the US would lead to a recession. These challenges highlight a critical shift where the deployment of AI compute is increasingly constrained by foundational infrastructure, rather than just chip supply.
Silicon
| Chip | Maker | What | For Whom |
|---|---|---|---|
| Maixstation | Sipeed | A Chinese rival to NVIDIA's DGX Spark, featuring 128GB of LPDDR5X memory. | Local AI enthusiasts and developers |
The build-out
AI Data Center Capacity Expansion
Who: Aolani
Scale: Up to 500 MW
Where: Multiple sites in Malaysia, with capacity targeted to come online starting in the second half of 2027
On-site Power Island Units for AI Data Centers
Who: Venture Global with GE Vernova and Baker Hughes
Scale: Up to 16 new 720 MW combined-cycle gas power island units, valued at approximately $17.2 billion
Where: Company-owned sites in the US, with the first power-island purchase order expected by the end of 2026
Supply & policy signals
Morgan Stanley forecasts a 32-gigawatt power deficit for US data centers by 2028.
Implication: This impending shortage poses a risk for smaller enterprises and could shift hyperscaler demand to Asia.
Bernstein's October analysis indicates DRAM and NAND flash contract prices rose nearly 20% in Q3 2026, with 2026 HBM production sold out to AI data centers.
Implication: This will likely lead to increased prices for consumer PCs and GPUs, maintaining tight supply for non-AI sectors.
The Financial Times reports that at least 47 major US AI projects were delayed, withdrawn, or blocked in 2026, and Bernstein analysts estimate only about 35% of the proposed US data center pipeline is credible.
Implication: This suggests a significant slowdown in compute capacity expansion, challenging investor confidence and developers' ability to deliver.
What we'll be watching
- Venture Global's expected first power-island purchase order for its AI utility by the end of 2026.
- The rollout of Aolani's 500 MW AI data center capacity in Malaysia, commencing in the second half of 2027.
- Further developments regarding the funding and construction challenges impacting US data center projects.
- Updates on global memory chip prices and the supply of High-Bandwidth Memory (HBM) as 2026 concludes.
Reporting + analyst voices: grounded via Google Search at publish time.