Singtel launches AI cloud service; to process data locally for sensitive sectors, government

SENSITIVE sectors, such as banking and healthcare, as well as the public sector, will soon be able to tap telco giant Singtel’s artificial intelligence (AI) cloud offering.

The company launched its offering, called RE:AI, on Thursday (Oct 10) at its George Street office, along with five memoranda of understanding with academia, industry players and AI startups.

Chief executive of Singtel’s Digital InfraCo unit Bill Chang said at a media briefing that unlike other cloud players that may bring AI hardware to Singapore, the company has the advantage of being a homegrown player that is focused on the local market.

He noted that companies with sensitive use cases, such as those in the banking and healthcare sectors, need to process their data locally.

“Would they want to put this (data) with a non-Singaporean player who’s got GPUs here, or with a Singaporean player who’s got GPUs here, that’s one key question,” he said.

GPUs, or graphics processing units, are used to both train and run AI models. As part of RE:AI, Singtel is offering access to Nvidia H100 GPU chips, which are one of the fastest chips on the market now.

As Nvidia updates its chips, Chang said that the company is committed to bringing the latest chips, such as the upcoming Nvidia GB200 GPU chips, to market as well.

“If you think about the challenge of these cloud players, they cannot just think about Singapore, they have to think about their global data centres,” he said, adding that as a national AI factory of sorts, Singtel will be able to move faster to adopt new technologies.

Furthermore, Chang noted that aside from having sufficient GPU capacity, the telco can also ensure fast and secure connections for its customers as it has access to network infrastructure, such as the 5G core and quantum safe networking.

“Autonomous vehicles, robotics, all sorts of use cases at the endpoint can be delivered with assured latency,” he said, adding that the company has learned to optimise its solutions after working with companies such as Hyundai and Micron in previously announced trials.

He added that companies in the manufacturing sector have not been able to adopt AI more readily due to the prohibitive cost of the server hardware involved, as well as the complexity behind powering, cooling and running these servers with software.

Instead of selling companies raw GPU compute power, Chang said that it is more useful for the company to sell AI as a service to clients so that companies do not need to incur the manpower costs of developing and troubleshooting AI software.

One of the memoranda of understanding the company signed is with the Advanced Remanufacturing and Technology Centre, Singapore Institute of Manufacturing Technology, Singapore Manufacturing Federation (SMF), and Singapore Institute of Technology (SIT) to create an AI Centre of Excellence for Advanced Manufacturing.

Singtel expects 150 companies from SMF’s network to be involved in trials with students and academia from SIT over the next three years. They will develop smart AI models and applications for the sector, the company said.

The company is also partnering with AI startups such as AlphaSense, H2O.AI, Hippocratic AI and Hive AI to launch software, applications, tools and models for customers.

For instance, Hippocratic AI is a healthcare AI assistant that can play the role of a nurse who can make proactive calls and answer questions that patients may have about their condition.

Hippocratic AI co-founder and chief executive Munjal Singh said that the company’s solution can only work if the company has access to GPUs that are close by.

While the company could access GPUs on the US East Coast, data would take 400 milliseconds to be sent to and fro. Instead, Singtel said that it can bring AI processing to customers in less than 10 milliseconds.

“It has to be somewhat snappy so it has to be local, (data) can’t be going across the ocean and back,” he said.

Furthermore, Singh said that local data sovereignty laws prevent the company from sending local healthcare groups’ patient data out to other jurisdictions for processing.

He added Singtel has introduced the company to different healthcare groups in Singapore and given it a chance to demonstrate its solutions to them.

While Chang did not disclose how much the company has spent on capital expenditure to build the service or the clients that have expressed interest, he said that he has observed interest from the public sector, AI research companies, financial institutions and advanced manufacturing players.

“We have done our sums, based on the projection of the huge growth opportunity … we’ve committed ourselves to this bet,” he said.

He added that the company is starting its AI cloud offering locally while aiming to move it into the region.

These include collaborations with Thailand’s AIS, Indonesia’s Telkom Group and Malaysia’s Maxis to bring Singtel’s AI Cloud offerings to enterprises in the region.

Chang said that these partners are “keenly watching and working with us to see how we can bring those capabilities into their countries as sovereign AI factories”.

At the announcement, Deputy Prime Minister and Minister for Trade and Industry Gan Kim Yong said that many companies and organisations may not be capable or large enough to develop and deploy their own AI solutions.

“We must therefore also develop platforms on which companies can train their AI models as well as run their AI workloads,” he said.

“To this end, Singtel’s AI cloud service is a turnkey platform that provides access to compute infrastructure in a development workspace,” Gan said, adding that the launch supports the National AI Strategy 2.0, which was unveiled in December 2023.

The announcement comes after a string of data centre collaborations between Singtel and other parties.

In August, Singtel’s regional data centre arm, Nxera, signed a memorandum of understanding with Hitachi to develop data centres across Japan and potentially the wider Asia-Pacific region.

In July, Singtel also announced that Nxera and Malaysia telco Telekom Malaysia broke ground for a data centre campus in Iskandar Puteri in Johor. The campus will house a 65 megawatt (MW) data centre, which can be scaled up to 200 MW with market demand.