An easy question about Grab's announcement on Monday is whether the driverless cars work. A harder one, and the only one that reaches the share price, is what 50 vehicles are a share of. Grab said on 14 September that it will take its Punggol autonomous fleet from 10 cars to 50 over the next six months, after more than 110,000 kilometres driven autonomously and more than 12,000 unique riders. Singapore's roads carried 634,000 point-to-point trips a day in July.
Fifty cars against 634,000 trips a day
Singapore's Land Transport Authority counts the sector monthly and publishes it. July 2026 averaged 43,000 street-hail trips and 591,000 ride-hail trips a day, 634,000 together.
Set the finished fleet against that. Fifty vehicles at 20 paid trips each a day gives 50 × 20 = 1,000 trips, and 1,000 ÷ 634,000 = 0.16%. The 20 is ours, not Grab's, and deliberately generous: the only rider figure Grab discloses, worked out below, is running at about six new riders per vehicle per day. Nothing in the table above changes the order of the answer.
Note which way the 0.16% errs. Grab publishes no Singapore trip count, so the island's sector is the only divisor in print, and Grab is not the whole island: measured against Grab's own book the share would be larger. Call 0.16% a floor.
The 110,000 kilometres belong to a free shuttle
Grab's two operating figures moved between dated statements, which turns them into a rate. On 29 July, in figures The Business Times carries from that day's release, Grab put the fleet at about 90,000 kilometres and more than 9,000 unique riders; on 14 September, more than 110,000 and more than 12,000. Forty-seven days separate them. Roughly 20,000 kilometres ÷ 47 days ÷ 10 vehicles works out to 43 kilometres per vehicle per day, and about 3,000 additional riders gives 6.4 first-time riders per vehicle per day. Both rates are ours and both are approximate, because Grab's endpoints are "more than" figures.
Neither is revenue. The Punggol shuttle runs free, and Grab told CNA it stays free while the on-demand service is trialled. Fares begin when the service opens to the general public in the fourth quarter of 2026, a date Grab published on 29 July and repeated on its 4 August call. Monday's statement moved the fleet target, not the revenue date.
Grab added 50 taxis in March alone
The regulator counts Grab's other Singapore fleet every month, and that comparison settles the scale. GrabCab, the street-hail subsidiary of Grab Rentals, took its operator licence on 9 April 2025 and first appears in the LTA fleet table that June with 20 taxis. By July 2026 it had 557. Across the six months from January to July 2026 it went from 345 to 557, adding 212 cars, while the plan announced on Monday adds 40 over the next six. Divide the two and Grab's human fleet grew 5.3 times faster.
March 2026 on its own was worth 50 taxis, 370 to 420. In one month Grab added as many cars as the whole autonomous fleet it is aiming for, drivers included.
The taxi side is also not optional. LTA gave GrabCab a three-year grace period to reach a minimum fleet of 800 taxis, falling due on 9 April 2028. Grab holds 557 and owes 243 more human-driven cabs to keep a ten-year licence, against 40 driverless ones it has chosen to add.
What "thousands by 2030" would actually take
A pilot is meant to be small, and the fair objection is that 50 is a staging post. Grab has put a figure on the destination: thousands of vehicles across Southeast Asia by 2030. Work back from the island's 634,000 trips at the same 20 a day and the threshold is legible: 1% of Singapore's point-to-point trips needs 317 vehicles, 5% needs 1,585, 10% needs 3,170. Thousands is the right order for a fleet that registers in group results, and thousands is what Grab claims, four years out and across eight countries.
Anthony Tan made the case himself in August, more bluntly than anyone outside the company. More than half of all Southeast Asian transactions are two-wheel rides below a dollar, he said, which makes electric vehicles uneconomical for commercial rollout, and Singapore is only about 10% of all four-wheel transactions regionally. Grab's most autonomy-friendly market is a tenth of the business the technology could serve.
Two disclosures would overturn the reading. If Grab publishes a Singapore trip count that puts Punggol above 1% of it, the 0.16% is wrong. And if LTA's monthly table shows GrabCab stalling near 557 while Grab reaches the 50, the two programmes are converging and the gap here is a snapshot rather than a trend.
Grab's 14 September statement is reported by CNA (Daphne Yow) and The Business Times (Renald Yeo), which between them carry the move from 10 to 50, the ten WeRide GXR vehicles, the kilometres and riders, and the 2030 ambition. The 9,000 riders and the fourth-quarter fare date are Grab's own, from its second-quarter call, where Anthony Tan also gave the two-wheel mix and Singapore's share of four-wheel transactions; the 90,000 kilometres logged by late July reach us through The Business Times rather than from the July release. Grab's listing is at GRAB. Point-to-point trip counts, the taxi fleet table and GrabCab's monthly vehicle count are LTA published statistics through July 2026; the 800-taxi minimum and 9 April 2025 licence date are from LTA's award release of 2 April 2025. Every per-vehicle rate, share and fleet threshold here is ours, built on the stated 20-trips-a-day assumption.
