Presented by YourCFO
I need to be honest with you before we start, because otherwise this whole piece reads like a man with an axe to grind.
I do have an axe. Let me show it to you first.
I do not like Grab. I am not going to be coy about it. For my money it is one of the most frustrating companies to actually use in this part of the world, and frustrating precisely because there is often no other option. The app fights me. The map sends my driver in a slow circle around the block. Support, when something goes wrong, is a maze designed so you give up before a human ever appears. And all of it sits on top of a market position so dominant that none of it has to get better.
So this is not a neutral host reading you the news. I want that on the table up front.
Why give a company I cannot stand a whole week? Because this week Grab did something that fits the pattern I have been chasing for a month. The $13 billion funding number wearing the costume of a boom. Exit figures wearing the costume of liquidity. Seed rounds wearing the costume of a Series A. Real numbers, dressed up as something they are not.
This week the costume was not on a spreadsheet. It was on a company. In seven days, Grab put on a lab coat and announced to Southeast Asia, twice, that it is now an artificial intelligence company. I looked at the thing wearing the lab coat and I knew its face, because I argue with it in traffic every week.
I will be fair. Grab earns some credit here and it will get it. But the costume is coming off.
One. Two announcements in seven days
The facts first, clean, before I say a single sharp thing.
Announcement one. On 8 October in Kuching, SoftBank, Grab and PETROS, the Sarawak state energy company, signed a framework agreement to explore an AI and digital infrastructure platform in Sarawak. The scope, as described, could stretch across AI compute and services, advanced technologies, robotics, and local workforce and community development. SoftBank brings its AI infrastructure and computing expertise. PETROS brings energy. Grab, per the release, brings its Southeast Asian digital platform, its “applied AI capabilities” and its regional customer and partner networks.
Hold on to that last line. We are coming back to it, because it is the most honest sentence in the announcement.
Announcement two. Six days later, on 14 October in Bangkok, Grab stood up with OpenAI and launched GO Forward with AI: practical AI training for more than 3,000 drivers, riders and merchants in Thailand, with three months of free ChatGPT Plus for those who attend the workshops in person. It is the first step in a plan to reach more than 30,000 workers and small operators across the region within two years. Grab’s chief executive, Anthony Tan, framed the point well: access alone is not enough, the question is whether the training turns into earnings.
Now the fair hearing I promised.
The gig worker programme, at its best, is a genuinely good idea. A while back I got into a Grab car, mentioned what I do, and the driver spent the ride telling me what he had built on the side with AI tools. The people driving these motorbikes and cars and packing these bags are some of the most capable, hustling operators in the region. A driver who learns to use these tools to run a side business, market it and manage the money is a real thing. They are also the people most exposed if autonomy and automation arrive in their line of work. Teaching that class of worker to use the most powerful general tool of a generation is not nothing. I am not going to sneer at it.
And the Sarawak plan points at something the region genuinely needs. Sarawak has been thoughtful and aggressive about making its own moves. Real computing capacity, built here, with local energy behind it, is a serious ambition. No quarrel with the ambition, none with Sarawak wanting it, none with a local energy firm backing it. Build it.
So here is the question for the rest of this piece. Two announcements, both with the letters A and I stamped all over them, and one company at the centre of both, suddenly presenting itself as an engine of Southeast Asian intelligence. What does either of these actually require Grab to build? What does Grab have to deliver, and by when, or it has failed?
I went looking for the answer and found a lab coat with nobody inside it.
Two. The map in my pocket
Let me explain why I cannot take Grab seriously as an AI company, using the least glamorous technology in the world. A map.
A few years ago Grab decided to move off the outside mapping providers that powered its navigation and build its own, GrabMaps. The logic was reasonable on paper. Southeast Asia is genuinely hard to map. Alleys with no names. Addresses that are really a landmark and a vibe. Whole neighbourhoods the big global maps render as a grey blank. Grab said it had millions of drivers on these roads every day and could map them better than anyone. Good story.
Here is the part that is not on paper. It is on my phone, several times a week.
In my own daily use, since that switch, the experience has got worse. Type in an exact address and it cannot find it. The pin lands on the wrong side of the building, or the wrong side of the road. The driver is routed to a phantom entrance. I watch the little car on the screen drive past me, turn away and loop the block while the arrival time climbs, and the arrival time had no basis in reality to begin with.
I am not giving you a statistic. This is me, a regular user, telling you what the product does in my hand. The reporting on whether drivers are happier since the switch is thin, so I will not dress my annoyance up as data. But I will tell you what it did to my behaviour. I have cut my usage sharply and started using other providers for most of my travel, out of sheer frustration. It did not get better. It got worse, and then it stayed worse.
Three. The dominance pays for the mediocrity
Why would a company let its core product get worse and just leave it there? That is the real question, and it has nothing to do with AI.
A company that wins on the quality of its product cannot let the product rot. If you have three real competitors and your map sends people to the wrong door, you lose them by the end of the week. A company that wins on dominance, on the simple fact that there is nowhere else to go, can let the product sit at good enough forever. There is no punishment. The customer grumbles, screenshots the circling car, posts it, and opens the same app the next morning.
The map is just the part I can see. Try reaching a human when an order arrives wrong, or when the pizza box arrives tipped on its side with everything piled in one corner, which has happened to me more times than I can count. The support experience feels less like help and more like a series of locked doors, each one hoping you give up before you reach the next. That is not the accident of a young company still figuring things out. Grab is not young. It is a design choice a dominant company can afford, because the cost of your frustration is carried by you, not by them.
That is what sits underneath the costume. In my experience Grab is not a technology company that wins because its technology is excellent. It is a distribution position. It got big first. It bought its position with capital, got the drivers, got the merchants, got the habit. From inside that fortress, mediocre tech, a clumsy interface and poor customer experience are not a crisis. They are the resting state. Competing against that dominance costs too much, so the dominance pays for the mediocrity.
And notice who absorbs the resting state. Not only the irritated rider. The driver whose map sends him the wrong way and whose time is eaten by the loop around the block. The small merchant who lives inside an app she does not control and cannot meaningfully complain to. The people most dependent on Grab pay the quiet tax of a product allowed to stay just good enough.
So forgive me if I am slow to believe the fix is a rebrand as an AI company. The tools to make the core experience genuinely good have existed for years. In the age of AI, a product this clumsy is not a technical failure. It is a decision nobody made. The will to change costs money, and the dominance never required it.
When that same company walks on stage in a fresh lab coat and tells me it will be the engine of artificial intelligence for Southeast Asia, I have one honest reaction. You could not keep my driver from circling the block. You could not answer one customer inquiry. You could not show a written address in a major city. You built your own map, made it worse, and your position was so strong you never had to fix it. And now you will deliver frontier intelligence to 30,000 people?
The sharper version, and then I will move on. A company that is comfortable does not get better at the hard thing. It gets better at the announcement. When the product stops being how you win, the narrative becomes how you win. Right now AI is the richest narrative on the planet, the most expensive lab coat money can buy. Of course they put it on. Everyone comfortable is putting it on. The question is whether there is a scientist inside the coat.
Four. Follow the money to Tokyo
Look again at the two announcements, but this time look past Grab, at who is standing next to it.
In Sarawak, the partner bringing the compute is SoftBank. In Bangkok, the partner bringing the intelligence is OpenAI. And SoftBank is one of OpenAI’s largest backers on earth, committed to both the company and the physical infrastructure AI runs on. Line the two deals up and the picture reorganises itself.
This is not really a story about Grab becoming an intelligence company. It reads far more like SoftBank and the companies in its orbit wiring themselves into Southeast Asian AI, through the one asset in this region that is almost unbeatably valuable. Not Grab’s technology. Grab’s distribution. More than 50 million people transacting through that app every month: riders, eaters, drivers, merchants. That is the prize.
Go back to the honest line in the Sarawak release. Grab brings its regional customer and partner networks. That is the whole role. Grab is not bringing the chips or the models. SoftBank and OpenAI have those. Grab brings the front door, the daily habit of a region.
And that is worth pausing on, because it is the one genuinely scarce asset in the arrangement, and it is ours. Chips you can buy, even if supply gets tight. Models you can license. Engineers you can hire in any city on earth. Access to tens of millions of people who already let one app into the centre of their daily lives took years and billions of dollars of other people’s money to build. You cannot order it on a two-year framework. Everybody in the room knows it. That is why Grab is in the room. Not for its intelligence. For its reach.
Seen that way, the gig worker programme reads differently too. Free ChatGPT Plus for the most frequent, most loyal users of the region’s dominant app is, among other things, a beautifully efficient way to seed a US intelligence product into daily life in Southeast Asia, with a local champion as the delivery van. I am not calling it sinister. The upskilling is real. I am calling it distribution. In this arrangement, distribution is exactly what Grab is for.
So the question for founders and operators is not whether Grab is a good AI company. Wrong question. The question is who will own the intelligence layer being built on top of our daily transactions. Because it is being built, and SoftBank’s money is extremely real. Follow it to the end and it passes through Grab’s distribution on its way to Tokyo and San Francisco. The venue is Southeast Asia. The destination may be somewhere else.
If that sounds familiar, it should. It is the pattern I pulled the costume off a couple of weeks ago. The region as the stage, not the owner. This week it is wearing a lab coat.
One more thing, said carefully. This is the same SoftBank that, since late last year, has been among the shareholders pushing changes at GoTo that were widely read as clearing the way for a Grab deal. One investor, arranging the furniture so the regional champions consolidate and the intelligence on top of them is supplied from outside. I am not predicting that deal, and I am not feeding the rumour mill. Just keep an eye on the hand doing the arranging.
Five. What a signature is worth
Back to Sarawak, because I skipped the most important detail on purpose.
I called it a framework agreement. That is one of the most misunderstood objects in business, and once you understand it you will not read that kind of headline the same way again.
A framework agreement is an agreement to maybe, possibly, think about doing a thing later. This one carried no stated investment figure. And by its own terms it does not oblige the parties to enter definitive agreements or to implement the project at all. In plain language: three large organisations signed a piece of paper that commits them to nothing. They agreed to draw up a roadmap. They agreed to explore. They shook hands in front of cameras, and not one of them is on the hook to pour a bag of concrete or put a single ringgit into anything.
I am not mocking the deal. The ambition is real, and the people who want to build real capacity in this region are right to want it. Sarawak deserves credit for this and for a long list of other forward-looking moves. My problem is not the plan. It is the gap between what was signed and how it was reported. In plenty of places it read as though a data centre was already rising out of the ground. What actually happened was a signing ceremony and a promise to make a plan.
The lesson is universal. It applies to everyone, including me, including you. A memorandum, a framework, a letter of intent: these are the cheapest objects in business. They cost nothing to sign and they generate enormous headlines. The distance between the signature and a working thing is where credibility is won or lost, and most announcements never travel it.
So when you see two famous logos on a framework agreement, ask one question. What, in this document, is anyone actually obligated to deliver, and by when? If the answer is nothing, you are not looking at an investment. You are looking at a photograph.
Apply the same discipline to the other number. 30,000 workers trained over two years sounds big. Set it against the base. Grab’s drivers and workers across the region run into the millions. 30,000 over 24 months is a pilot and a press moment. A worthy pilot, and I am glad it exists. But it is not a labour policy, and it has not changed anybody’s income yet. Measure it in two years by what the 30,000 are actually earning, not by the size of the launch.
Six. The gift a comfortable giant hands you
I said I would not just diagnose and walk away. So let me turn this all the way around, because buried in my frustration is the most hopeful thing I can tell a founder this year.
A dominant company that has stopped improving its product is the single greatest gift an incumbent can hand you.
Think about what I just described. A company so far ahead it no longer has to be good. A map that misroutes. Support built to exhaust you. A product parked at good enough because the dominance pays the bill. Every one of those is a crack. Every customer screenshotting a circling car is someone who would leave in a heartbeat if something genuinely better showed up and was easy to switch to. Comfort is the vulnerability. The scale that makes Grab so hard to compete with is exactly what lets it ignore quality, and ignored quality is the opening.
This is not only about Grab. It applies to any dominant player, in any market. If you are building in the region right now, three things.
* Find the single workflow the giant does worst, and own it completely. Do not build a better everything app. You will lose. Find the one job the dominance lets it neglect and be ten times better at just that. When you are small, depth beats breadth.
* Build where switching is easy and quality is felt every day. A monopoly is safest where leaving is a hassle. It is weakest at the daily, visible, emotional touchpoints, where a customer feels the product succeed or fail in real time. Live there.
* Treat “good enough because we are the only option” as a target painted on someone’s back. Every time an incumbent says that to itself, it is telling you where to aim. Write it on your wall.
To make it concrete. Say you are looking at food delivery and you look at Grab and think: impossible, they have everyone. Wrong frame. Do not look at everyone. Look at the one merchant who hates the commission and feels invisible inside the giant. The one neighbourhood the big map renders badly. The one moment, the order that arrives wrong, when the customer feels abandoned. Win that merchant, that neighbourhood, that moment, completely. You are not trying to beat the fortress. You are trying to be undeniably better at the thing the fortress stopped caring about. That is a business. It starts small on purpose, and small and excellent beats large and comfortable more often than anyone in a pitch meeting wants to admit.
The bow
I told you at the top that I do not like Grab. That is still true. But I want to be precise about what I am against, because it is not a logo, and it is not a company I am jealous of.
I am against comfort pretending to be ambition. I am against a lab coat worn by an organisation that could not get my driver there on time or my dinner there in one piece. I am not anti-Grab. I am pro-product. I am for companies that win because the thing works when you touch it and the support shows up when it does not, not because there is nowhere else to go.
And here is the part I believe all the way down. The best thing for this region is not that our champions get comfortable and put on costumes. It is that somewhere out there right now, a founder is picking up the tool a giant left on the floor and building the version that actually works, for the customer that giant stopped respecting. That founder does not need a framework agreement or a ceremony. They need the incumbent to stay exactly this comfortable for about 18 more months.
If that is you, reply and tell me. I read every one, and the pushback is where the next episode comes from. And if you know a founder quietly building the better version while the big one announces the future, forward this to them. They need the part about the gift.
SEA of Startups is brought to you by YourCFO, an Indelible Ventures portfolio company. Most founders I meet run finance off a spreadsheet and hope. YourCFO is the finance team, minus the team. Code SOS gets you 10% off at yourcfo.tech
This piece accompanies this week’s episode of SEA of Startups. Real. Raw. Relatable. Listen on Spotify, Apple Podcasts, or YouTube, and subscribe here so the next one lands in your inbox.
