Breaking
GPU capacity is now billable by the second
Short training and inference jobs no longer pay for a full hour of idle GPU time.
Articles by
KGE Engineering
Platform Engineering
Technical notes from the team that runs our cloud, GPU and network platforms.
Postmortem: a load balancer failover that took longer than it should
A failover completed in 11 minutes instead of 90 seconds. Here is what we changed.
East Malaysia latency: what it actually costs
Hosting in Singapore and calling it regional does not serve Sabah and Sarawak well.
Opinion
Uptime SLA: what 99.95% actually means
It is about 22 minutes a month. Read the exclusions before you compare two numbers.
YTL and Nvidia Bring AI Data Centre Online in Kulai With Malaysia's ILMU Model
YTL Power and Nvidia have brought an AI data centre online at the 600MW YTL Green Data Centre Park in Kulai, Johor, running liquid-cooled NVL72 Grace Blackwell GPUs alongside Malaysia's own ILMU model.