Grab Leverages AI to Combat Rising Fuel Costs, Drive Growth

Grab, Southeast Asia's leading ride-hailing and delivery organisation, is banking on artificial intelligence-led products and services to foster growth and overcome challenges such as affordability and increasing fuel costs in the wake of the war in Iran. CEO Anthony Tan stated the Singapore-based company believes strongly in its AI strategy, which is already yielding positive results.

Credit: UNSPLASH

Tan confirmed the fuel cost situation is a real concern for everyone, emphasising how companies like Grab must become more conscious of customers’ wallets. The CEO highlighted Grab’s scale, with an LSEG-estimated market value of USD 14.5 billion, as a crucial differentiator, providing "tremendous data" to aid growth.

This data allows Grab to make services more affordable, which in turn encourages more people to use them, according to Tan. He described this as the best way to drive growth, by building AI-led strategies that are unique.

Earlier this year, Grab expanded beyond Southeast Asia for the first time by acquiring Delivery Hero's Foodpanda delivery business in Taiwan. Despite this expansion, the company’s fiscal 2026 revenue and adjusted EBITDA forecasts fell short of Wall Street expectations.

This signals slower momentum in its core ride-hailing and delivery businesses, as consumers grappled with economic uncertainty even before the war in Iran. Grab's share price has decreased by nearly 30% this year, despite the Nasdaq-listed company announcing its first-ever full-year net profit 14 years after its founding.

Among the 13 products Grab recently unveiled was a "group ride" feature, designed to save customers up to 40% on fares. This feature uses AI to precisely calculate fare splits among groups of travellers.

The company did not disclose the investment value for these 13 AI products. The "group ride" product will soon see a wider rollout in Indonesia, the region’s largest economy and the biggest of Grab's eight operational markets.

Tan expressed satisfaction with Grab’s presence in Indonesia, affirming the company's commitment to "keep doubling down" its efforts there.

---

• Grab is leveraging artificial intelligence to counter rising fuel costs and drive growth.

• CEO Anthony Tan affirmed the company's strong belief in its AI-led product strategy, citing positive results.

• Grab recently unveiled 13 new products, including an AI-powered "group ride" feature, which can save customers up to 40% on fares.

Source: REUTERS

Meta Unveils Muse Spark AI Model, Superintelligence Team's First Release

Meta Platforms has unveiled Muse Spark, its initial artificial intelligence (AI) model developed by a newly assembled "superintelligence" team. Shares of the company extended gains, trading up nearly 7% following the announcement. This new model represents Meta's efforts to compete with rivals in the AI sector.

Credit: META

The company assembled its superintelligence team last year, hiring Chief Executive Officer Alex Wang from Scale AI under a $14.3 billion deal. Some engineers were offered pay packages of hundreds of millions of dollars to staff the team, aiming to advance Meta’s position in AI development.

Credit: META

"Superintelligence" refers to AI machines that could potentially surpass human thought capabilities. Muse Spark is the first in a new series of models internally known as Avocado, developed by this specialised team.

Muse Spark will initially be available on the Meta AI application and website. The company stated that the model will replace existing Llama models that power chatbots on WhatsApp, Instagram, Facebook, and Meta’s smart glasses in the coming weeks.

Meta did not disclose Muse Spark’s size, a metric typically used to compare an AI system’s computing power with competitors. Unlike previous open releases of its Llama models, Meta shared only a "private preview" of Muse Spark with unnamed partners.

According to a company blog post, "This initial model is small and fast by design, yet capable enough to reason through complex questions in science, math, and health. It is a powerful foundation, and the next generation is already in development."

Independent evaluations of Muse Spark’s performance showed it catching up with top models from market leaders Google, OpenAI, and Anthropic in some areas, such as language and visual understanding. However, the model lagged in coding and abstract reasoning, tying for fourth place on a broad index of AI tests compiled by Artificial Analysis.

Meta Chief Executive Officer Mark Zuckerberg tempered expectations in January. He stated the team’s first models "will be good but, more importantly, will show the rapid trajectory that we’re on." Zuckerberg added, "I expect us to steadily push the frontier over the course of the year as we continue to release new models."

Wang, who leads the superintelligence team, acknowledged that "there are certainly rough edges we will polish over time in model behavior." He confirmed that larger versions of Muse Spark are under development, with plans to release some openly.

Meta also provided insights into its monetisation strategy, teasing shopping features embedded within its Meta AI chatbot. These features aim to direct users to purchasable products.

The company is betting that applying AI to everyday personal tasks will increase engagement among its more than 3.5 billion users across its social media platforms. This strategy could provide Meta an advantage over rivals with smaller user bases.

Muse Spark can assist users with tasks such as estimating calories in a meal from a photo or superimposing an image of a mug on a shelf. An additional Contemplating Mode runs multiple agents simultaneously to boost reasoning power.

This mode would allow Muse Spark to take on the extended thinking modes of Google’s Gemini Deep Think and OpenAI’s GPT Pro. Users could leverage this for tasks like efficiently planning a family vacation, with one agent drafting an itinerary while another searches for kid-friendly activities.

---

• Meta Platforms launched Muse Spark, the first AI model from its superintelligence team.

• The model will initially be available on the Meta AI app and website, later replacing existing Llama models across Meta's platforms.

• Independent evaluations show Muse Spark performs well in language and visual understanding but lags in coding and abstract reasoning.

Source: REUTERS

China Pioneers Offshore Wind-Powered Underwater Data Centres

Shanghai has launched an underwater data centre (UDC) directly linked to an offshore wind farm, a global first. This pioneering facility aims to meet soaring demand for artificial intelligence (AI) computing power.

Credit: UNSPLASH

The UDC is located 10 kilometres off Shanghai’s eastern coast, 10 metres beneath the water. Shanghai HiCloud Technology, a subsidiary of Highlander, built the facility within Shanghai’s Lingang Special Area.

This project, which represents a total investment of USD 232.4 million, has a planned capacity of 24 megawatts. It serves as a key pilot to explore the technical and commercial viability of subsea data centres connected to onshore cloud and telecoms infrastructure.

The cluster will power multiple applications, including AI scenarios, embodied intelligence, and autonomous driving. This initiative is part of China's broader efforts to expand its computing power supply by looking to the seas and skies.

Shanghai's coastal project follows an earlier UDC initiative by Highlander off the southern island province of Hainan. The Hainan facility was installed in a 1,300-ton underwater data cabin, equivalent to the weight of 1,000 passenger cars.

The first phase of the Hainan UDC project finished construction in 2023. The first phase of the Hainan UDC project, which finished construction in 2023, was projected by Deputy General Manager Li Jiawen of HiCloud to save 26,000 tons of water and 3.4 million kilowatt-hours of electricity annually.

Li added that this first phase of the Hainan facility was also projected to reduce carbon emissions by about 2,720 tons compared with traditional land data centres of the same size. Analysts noted HiCloud had demonstrated "comparative advantages of submarine data centres" through its commercial operation in Hainan.

The combination of subsea data centres and offshore wind power could help Shanghai solve computing bottlenecks. The city faces limited land resources, while AI workloads demand high-efficiency, low-latency local computing power.

However, analysts also pointed out difficulties in building subsea facilities, as the technology is still in its early stages. Challenges include a lack of standards, operational and maintenance difficulties, and economic viability.

Globally, there has been wider interest in submerged computing hubs. In the US, Microsoft explored similar possibilities through its Project Natick, though it stated in 2024 that the project was no longer active due to concerns over its operational feasibility.

---

• Shanghai has launched the world’s first underwater data centre directly linked to an offshore wind farm.

• The facility, built by Shanghai HiCloud Technology, aims to support artificial intelligence and other high-demand computing applications.

• The project, located 10 kilometres off Shanghai, has a total investment of USD 232.4 million and a planned capacity of 24 megawatts.

Source: SCMP