CGSI Research estimates the AI-DC segment can contribute about RM0.9bil to YTL Power’s net earnings.
KUALA LUMPUR: YTL Power International Bhd
may gain from stronger than anticipated earnings from its artificial intelligence (AI) based data centres (AI-DC) on strong margins.
This is based on data from CoreWeave’s initial public offering disclosures as it is set for a listing on the US-Nasdaq stock exchange.
This provides a valuable reference for YTL Power amid limited available details, said CGS-International Research (CGSI Research).
It said after going through CoreWeave’s financials that its earnings expectations for YTL Power’s AI-DC appear reasonable.
“We estimate that a typical AI-DC operation generates an earnings before interest, taxes, depreciation, and amortisation (Ebitda) margin of circa 65% and Ebitda per megawatt (MW) of US$10mil.
“This is derived from the company’s fourth quarter 2024’s results, assuming an effective information technology load of circa 190MW during the quarter versus its 360MW active power as at end-2024,” it said.
This compares to its current Ebitda margin and Ebitda/MW forecasts for YTL Power’s AI-DC business of 65%, with capacity offtake scaling up from 20MW in mid-2025 to 100MW by mid-2027.
“At full ramp-up, we currently estimate the AI-DC segment can contribute to a circa. RM0.9bil to YTL Power’s net earnings,” CGSI Research said.
The research house maintained its “add” rating on YTL Power, with an unchanged standard operating procedure-based target price of RM4 per share (pre-warrants), but tactically see better risk-reward at lower levels, especially following the 17% spike in share price over the past three weeks.
It also notes of the potential overhang from the proposed one-for-five bonus issue of free warrants that’s expected to be completed by the second quarter.
“In our opinion, securing the remaining 80MW AI DC offtake and greater clarity on these projects are crucial for the stock to re-rate further,” it said.
Shares of YTL Power traded at RM3.44 at the time of writing.
