Posted on — Leave a comment

How Hostinger is helping solo founders in Southeast Asia go from idea to paying customers

You have an idea for a service or a small product. You have a free weekend and no developer. By Sunday night you also have a list: a domain, hosting, a site builder, a way to take payment, a mailing list, somewhere to keep customer data, and a plan for answering the same three questions every day.

None of it is hard on its own. Together, it is why many good ideas stay in a notes app.

Founders in Indonesia, Malaysia, the Philippines and Vietnam face this with fewer hours and smaller budgets than a funded team. The work between an idea and a paying customer is getting shorter, though, because more of it can now be handed to AI that does the task rather than only talking about it.

What “agentic” means in plain words

Most people have used AI that answers. You ask how to add a login page and it explains the steps. Agentic AI takes the steps. You ask for a login page, and it builds the page, connects the database and tells you what it did.

The difference is who does the work. An explanation still leaves you with a to-do list. An action makes the list shorter. For a founder who is not technical, or only partly so, that gap decides whether a project moves this week or next month.

Step 1: Test the offer before you build anything

Write your offer in three sentences: who it is for, what they get, and what they pay. Say it to ten people who fit that description. Then count how many ask a follow-up question, such as when they can start or how to pay.

Those questions are your signal. Polite compliments are not. If fewer than two of the ten ask, change the offer before you touch a website.

Step 2: Build the smallest version that can take a customer

A first version needs three things: a page that explains the offer, a way to collect a booking, order or sign-up, and a way to follow up with the people who respond. A logo redesign, a blog and a second product can wait until after your first five customers.

This is the step where tools matter most, because it is where solo founders lose the most time to setup.

Also read: The US$1M per person revolution: how AI is reshaping Southeast Asia’s startup landscape

Where Hostinger fits: build, run and grow from one account

Hostinger began as a web hosting company and now describes itself as an agentic platform for building, running and growing a business online. It reported more than 5 million users in June 2026, described as small businesses and solopreneurs across more than 150 countries, with Indonesia among the five markets it names as the core of that base.

Build with Hostinger AI Builder

You can start Hostinger AI Builder from a prompt, a PDF, a screenshot or a link to an old site. In agentic mode the AI leads while you describe what you need. In manual mode you move sections and edit elements yourself.

Data, user accounts and sign-in run on Hostinger’s own backend, so asking for customer accounts means the AI sets up the database and the sign-in. It works in an environment separate from your live project, and changes go public only when you publish them.

Know the limits too. Hostinger’s documentation notes that agentic mode cannot currently create some types of projects, including native iOS or Android apps and trading or banking applications. If your idea falls into one of those categories, plan a different route.

AI Builder is designed for founders working solo or without a technical team, and for getting a working first version live quickly. As your product grows more complex, or starts handling large volumes of customer data or payments, bringing in a developer to review and extend what the AI built is a sensible next step.

Builders using the Hostinger platform can choose between Manual and Agentic modes, which can be easily toggled

Run with Hostinger Agent and Agentic Mail

Hostinger Agent works through chat to help you manage your site, from plugins and domains to fixes and edits. Hostinger says it is included with every plan for site management, and that a paid upgrade adds ready-to-use skills for SEO, content, marketing and sales.

Agentic Mail connects a mailbox to AI agents so they can sort enquiries or send routine replies. It takes some tool setup, so most non-technical founders will start with Hostinger Agent.

Agentic chat is key to any AI-powered workflow for building websites and apps today, and Hostinger’s built-in agent chat supports exactly that

Sell and grow

To take orders, add Hostinger E-commerce, a separate subscription that lets you sell unlimited products. Hostinger Reach handles email marketing: describe the campaign or the customer segment and the AI drafts it.

Build your site with Hostinger

Step 3: Ask your first customers for something

After the first sale, ask that customer two questions: what nearly stopped them from buying, and who else should hear about this. Write the answers down. The first answer tells you what to fix on your page. The second gives you your next three conversations.

Do this for your first ten customers, not only the first. Patterns usually appear within a handful of answers, and they tend to be about trust, delivery time or unclear pricing rather than missing features.

Also Read: SEA’s venture capital shifts from mega-rounds to AI and SaaS

Step 4: Hand over the repeating work and keep the decisions

Make a list of what you did more than twice last week: answering the same enquiry, resizing images, drafting the same email. Those are the tasks to hand to an agent first. Pricing, which customers you take on and what you promise stay with you.

A simple weekly routine is enough to start. On Monday, read the week’s enquiries and decide which ones you answer yourself. Midweek, send one email to your list. On Friday, review what the AI changed and approve what goes live. While you are small, aim for about an hour of review a week.

Track three numbers for the first month: enquiries received, enquiries that became orders, and hours spent on admin. If the admin hours are not falling, the automation is not working yet, and that is worth knowing early.

The order matters more than the tool

Tools will change, and so will what each one can do. The order of work holds: test the offer, build the smallest version, get paid, then automate what repeats. If you want to try the build step this week, Hostinger AI Builder is a place to start.

Build your site with Hostinger

—

The e27 team produced this article sponsored by Hostinger

We can share your story at e27 too! Engage the Southeast Asian tech ecosystem by bringing your story to the world. You can reach out to us here to get started.

Featured image: Hostinger

The post How Hostinger is helping solo founders in Southeast Asia go from idea to paying customers appeared first on e27.

Posted on — Leave a comment

ShopBack’s Justin Lee on why it let Meta’s Muse in while Amazon shut the door

When Meta switched on Muse in Singapore last month, ShopBack was one of eight launch connectors the agent could plug into. In the US, Amazon had taken a different view. It blocked Muse within 12 days of launch, complaining in part that an agent shopping for a customer bypasses the personalisation built into its store.

ShopBack, a shopping and rewards platform, chose to sign up rather than shut the door. Justin Lee, its Chief Product Officer, says the web was built for people to read, not for agents to act on. e27 sent him ten questions on what that means for a business that earns its keep between the shopper and the merchant. His answers show a confident bet. They also leave a few things unsaid.

Also Read: Agentic commerce’s dirty secret: the data powering AI purchases is often wrong

The stakes are easy to state. A shopping and rewards platform, lives in the gap between buyer and seller. An agent that can compare prices on its own could route around that gap entirely. Lee’s wager is that the agent will keep a shopping and reward platform in the loop because the reward is part of the best deal.

Two weeks to write for a reader that never skims

Without a connector, Lee says, an agent shops like a person with a search engine, only faster. It combs through pages designed for human eyes and pieces an answer together from whatever it finds. That works, but it is slower, and the agent cannot always tell whether a price or a reward is still current.

The connector replaces that guesswork with a direct line into ShopBack’s live data: merchants, prices and the savings attached to each. “Building one meant writing for a new kind of reader, one that takes everything at face value and acts on it straight away,” Lee says. His team built it in two weeks.

He frames Muse as the next stop on a familiar road. ShopBack moved from website to app to browser extension, and then into the ChatGPT App Store, where it was among the first shopping and rewards apps. Asked whether ShopBack saw something Amazon missed, Lee declines to pick a fight. “Different companies will make different choices for their own business,” he says.

It is a diplomatic answer to a blunt question. Amazon owns the shop, so it can afford to keep agents out. A shopping and rewards platform owns no shelves. If AI shopping agents reshape retail the way their backers expect, being the tool the agent shops with is partly insurance against being bypassed.

Same link, same commission, new storefront

Affiliate commerce runs on attribution: tracked clicks, cookies, a visible referral path. Lee insists nothing changes when Muse does the clicking. A sale counts as ShopBack’s only if it goes through a ShopBack link, which is the link Muse gives the shopper. The merchant’s own tracking confirms the sale, and merchants pay only for confirmed sales, on their existing terms.

Justin Lee

What about commission on a sale the agent would have made anyway? Lee argues the agent weighs several merchants before answering. When a ShopBack reward tips the balance, “the reward is what won the sale”. It is the classic affiliate defence, the same logic behind Lazada’s US$100 million affiliate push. Whether it holds for an agent that might have picked that merchant regardless is harder to prove.

Also Read: Singapore is paying startups to secure the AI agents it is racing to deploy

On who owns the customer, Lee explains a traveller who books a flight through Muse buys from the airline, which issues the ticket and keeps the relationship. Muse is the channel and ShopBack the rewards layer.

To the questions on who (ShopBack or Meta) keeps the conversational data and whether money changes hands with the tech behemoth, Li says: “How Muse handles conversations is best answered by Meta. On our side, ShopBack never sees what the member types. When Muse decides a member’s request calls for ShopBack, it turns to the ShopBack connector, and the connector returns the relevant options and rewards.”

When it comes to ⁠payment, he claims that no fees changed hands between ShopBack and Meta. “We joined Muse’s Singapore launch at Meta’s invitation, because we are always looking for more places to reward our members, and personal AI agents are where more of them are starting to shop.”

Cheapest after cashback, or best for the merchant?

Meta’s own launch demo has ShopBack finding the cheapest return flight to Seoul after cashback. Can a user check that the answer really is the cheapest, and not the one from the merchant paying the highest commission? Lee’s reply is that the agent sees each deal’s price and savings and weighs them “on the facts”. Anyone who wants to check can open the ShopBack app and find the same deal there.

That is a consistency check, not an independent one. The app and the connector draw on the same data, so finding the same answer in both tells a shopper ShopBack is consistent, not that it is cheapest across the whole market.

The pay-to-rank question gets the most interesting answer. Merchants can already boost the rewards they offer through ShopBack, and Lee says that boost reaches the shopper as a bigger saving. “We give the agent the facts, and the agent decides,” he says. It echoes the promise that Muse doesn’t care about your ad budget.

The line is sensible, but it is thinner than it sounds. A boost is merchant money spent to win the agent’s pick, and it works best when the agent weighs savings heavily. For brands already worried that if an AI agent cannot find you, you do not exist, boosted cashback may become the cheapest way to be found.

When the agent books the wrong flight

Meta delayed Muse’s US launch over security concerns, and The Information reported that the agent sent emails without approval during testing. So who pays if Muse books the wrong non-refundable fare using ShopBack’s data? Lee’s answer puts the purchase with the merchant, under the merchant’s terms. ShopBack handles only the rewards side, through its help centre.

That is a neat division of labour, and probably the legally correct one. It is also the kind of handover that makes liability for autonomous agents so murky. Lee concedes the point in broader terms. “I believe trust will decide whether people shop with agents at all,” he says, adding that the industry must stay clear about who does what. In a market where Singapore firms lead APAC in deploying AI agents and in rolling them back, that clarity is not optional.

Plumbing or destination? Lee says both

Much of cashback’s pull is emotional: the user sees the reward, feels clever and comes back. If the agent does the shopping, Lee says, the satisfaction simply moves into the answer, with Muse showing the savings on each hotel option. Members still collect their rewards in ShopBack, and spend them online or in store with ShopBack Pay, a product backed by the MPI licence it secured from MAS.

The bigger worry is becoming back-end plumbing for someone else’s interface, especially as ShopBack builds its own AI tools. Co-founder Henry Chan has described searching 15 retailers for a single flight query. Lee rejects the choice. The connector runs on the same services as the app, so improving one improves the other. The app, with tools like its Travel Planner, remains where members cash out and stumble on deals they did not ask for.

Also Read: AI agents need someone standing behind them, says AmanTrue CEO Evin McMullen

Muse also has rivals: OpenAI’s Operator, Google’s Gemini Spark and Perplexity’s Comet. ShopBack built its connector to work with more than one agent. What stops it from becoming a swappable plug-in? Lee points to members’ own accounts and balances, merchant partnerships and more than a decade of understanding how people shop.

That moat is real but untested. Agents are built to find the best answer, as Lee keeps saying. The day a rival offers a fatter reward, the agent will report that fact without loyalty. ShopBack has bet that it can keep winning on the facts. In agentic commerce, that may be the only bet available.

The post ShopBack’s Justin Lee on why it let Meta’s Muse in while Amazon shut the door appeared first on e27.

Posted on — Leave a comment

Copper is choking AI. Staera secures US$4.6M to swap it for light

Every new AI chip arrives with a bigger compute number. Fewer arrive with an honest admission that those chips increasingly sit idle, waiting for data to trickle in over copper wires that were never designed for this kind of workload.

Singapore’s Staera Semiconductor is betting its future on that gap. The deeptech startup, formerly known as New Silicon Corporation (NSC), emerged from stealth today with US$4.6 million in pre-seed funding to commercialise Lightwire.

Lightwire is an optical interconnect that generates light inside the silicon chip itself rather than bolting a light source on next to it.

Also Read: Nexstrom lands US$12M to bring 2D semiconductors to 12-inch wafers

Openspace Capital led the round, with participation from SGInnovate, Black Kite Capital, 28 Holdings, In Group and Decimus Technologies. Staera is a spin-off from the Low Energy Electronic Systems (LEES) research group at the Singapore-MIT Alliance for Research and Technology (SMART), MIT’s first research centre outside the US.

Copper is running out of road

The problem Staera wants to solve is not new, but AI has made it acute. Inside an AI system, data has to travel between processors, memory stacks, boards and racks, over distances ranging from a millimetre to a metre. According to the company, copper’s electrical performance degrades roughly a thousandfold across that range. Push more data further over copper and you pay for it in power and heat.

That is a painful trade at a time when electricity is becoming the binding constraint on AI. The International Energy Agency expects global data centre electricity use to more than double, from about 415 terawatt-hours in 2024 to around 945 terawatt-hours by 2030. In Southeast Asia, where operators are already running short of power for AI servers, every watt saved on moving data is a watt that can go to compute, or not be drawn from the grid at all.

Even Nvidia, which said in 2025 it would stay with copper for as long as it could, now plans to use both copper and optics in its AI systems. Analysts tracking AI’s emerging bottleneck economy increasingly point to interconnects, packaging and memory bandwidth, rather than raw compute, as the place where the next squeeze, and the next wave of spending, will land.

Building the lamp into the wafer

Most optical links today are assembled from separate parts: a laser or LED made elsewhere, attached next to the chip and coupled to waveguides. Every join adds cost, power draw and another thing that can fail.

Staera’s pitch is that it fabricates the whole optical link on a single commercial silicon wafer using standard fab steps. That means gallium nitride (GaN) LEDs as the light source, silicon nitride waveguides to route the light, and the CMOS circuits that drive it, all on one piece of silicon. Lightwire sits inside the bridge or interposer chip that already connects processors and memory in today’s AI hardware, so, in theory, it slots into existing packaging flows.

“Copper performance plummets over the millimeter-to-meter distances that matter most in AI accelerators, and bolt-on optics add power and complexity,” said Dr Andrew Kim, CEO of Staera. “We build the light into the chip, the entire optical link inside a chip, so the connection gets smaller, cooler and faster with each generation.”

Also Read: NASA-linked, MIT-trained founders’ nSWX raises US$2M for AI chip packaging

The headline numbers are striking. Staera says Lightwire is designed to move data at under 0.2 picojoules per bit, against about 5 pJ/bit for co-packaged optics, the current state of the art. At memory scale, it says moving 24 terabytes a second would take under 50 watts with Lightwire, versus nearly a kilowatt with co-packaged optics. It also claims a reach beyond 50 mm, compared with 2 mm for the UCIe-A die-to-die links used in today’s chiplet designs.

The fine print hides in “designed to”

Note the verb. Almost every efficiency figure Staera has released is a design target, not a measured result from a shipping product. Commercial release is pencilled in for 2030, four years and, almost certainly, several funding rounds away.

There is also an engineering trade-off worth understanding. Each Lightwire link runs at a modest 4 to 8 Gbps, far slower per lane than the high-speed electrical and laser-based links found in data centres today. The bet is on density rather than speed: more than 600 links per millimetre of chip edge, all running in parallel. That approach plays to the strengths of LEDs, which are cheap and frugal with power, but cannot match lasers on per-channel bandwidth.

Staera does have manufacturing pedigree on its side. It says it has been building GaN LEDs into silicon chips in commercial foundries since 2016, with yields now at 98 to 99 per cent. That longevity cuts both ways, though. A decade of R&D culminating in a US$4.6 million pre-seed cheque is a modest sum for a company that wants to rewire AI hardware, and a reminder of how capital-hungry chip startups are.

Investors seem comfortable with the long horizon. “We backed Staera because of the strength of its technology, the depth of its team and its potential to define a new category in semiconductor connectivity,” said Hian Goh, co-founder of Openspace Capital.

The startup also won the ECTC Start-up Competition at the IEEE Electronic Components and Technology Conference in May, and placed runner-up at Lam Research’s 2026 Lam Capital Venture Competition the same month.

A crowded race to dethrone copper

Staera’s claim to be the first and only optical interconnect built fully into the chip will be tested by a well-funded field. In the US, Ayar Labs and Lightmatter have raised hundreds of millions of dollars for silicon photonics, while Marvell agreed to acquire photonic interconnect startup Celestial AI in late 2025.

California’s Avicena is the closest conceptual cousin, using microLED arrays for short-reach chip links. Broadcom and Nvidia, meanwhile, are pushing co-packaged optics into networking switches.

Closer to home, Staera joins a small but growing cohort of Southeast Asian semiconductor startups moving beyond assembly, including photonics player LightSpeed Photonics. Its real differentiator is monolithic integration: if the light source is grown on the same wafer as the circuits, it sidesteps the assembly and alignment headaches that dog bolt-on optics. Whether that holds at volume is the question customers will ask first.

Singapore wants a bigger slice of the AI hardware stack

Staera’s emergence fits neatly into Singapore’s push to climb the chip value chain. A*STAR and the Economic Development Board recently unveiled SG Semiconductor as a national identity for the sector, A*STAR has teamed up with Synopsys on AI chip packaging, and Applied Materials has expanded its Singapore footprint with a US$500 million campus. Interconnects and advanced packaging, the unglamorous plumbing between chips, are exactly where the city-state hopes to matter.

Also Read: Amplicity raises US$1M to put Singapore’s sleeping data centre batteries to work

The fresh capital will go towards advancing the technology, expanding the team into the US, and deepening work with customers and supply chain partners. The US push is telling: the buyers who will decide Lightwire’s fate, hyperscalers and AI chip designers, mostly sit in California.

For now, Staera has a compelling physics argument, a manufacturing track record and a four-year runway to prove its numbers survive contact with real hardware. If they do, the most important part of tomorrow’s AI chips may be the light nobody sees.

The post Copper is choking AI. Staera secures US$4.6M to swap it for light appeared first on e27.

Posted on — Leave a comment

Four hundred business cards. Zero customers

The show is over. The founder sits on the hotel bed and counts the cards.

Four hundred, give or take. A dozen promises to call next quarter. Several requests for samples. Everyone said the product was good. Not one conversation with someone who would actually use it.

Who comes to the booth

Importers. Distributors. Consultants. Accountants. Some buy. Others advise on regulation, tax and market entry. None of them uses the product.

The distributor knows what moves off the shelf, what gets returned, where the margins are. The consultant knows the rules. The accountant knows the tax treatment. None of them knows what happens after the product leaves the store: who opens it, at what hour, for what purpose. That was never their job.

So you hear the same sentence at 10 shows in a row. Great. Send samples.

Los Angeles, 1959

Honda opened its first overseas subsidiary in Los Angeles in June 1959, in a storefront on Pico Boulevard with eight employees. By the end of the year it had sold just 170 motorcycles. Americans wanted big motorcycles. Harleys. Triumphs.

Then the 50cc Super Cub started to move. Americans didn’t ride it the way the Japanese did. Some hauled it to campsites and fishing holes. Honda went where they were: direct mail to sporting-goods, boating and fishing shops, ads in general-interest magazines like Life, shelf space where no motorcycle had been sold before.

Monthly sales passed 1,000 in May 1961. By 1962, Honda was selling 40,000 a year in the US.

Honda’s answer wasn’t a bigger motorcycle. It was a different customer.

Nobody at a trade show tells you their customers load this into a pickup. You see it only by being there.

Also Read: Start with the problem, not the token: What Wall Street’s tokenisation push means for Southeast Asian founders

Samsung’s year abroad

In 1990, Lee Kun-hee launched Samsung’s regional-specialist programme. Not an expat posting. One employee, one country, a year or more. Learn the language. Walk the streets. Learn how ordinary people live.

Insiders resisted: the cost, the good people pulled off the line. Lee pushed it through. “We cannot do anything without globalised manpower,” he reportedly said.

Today each specialist costs more than 100 million won (US$68,000) a year on top of salary. Harvard Business Review later called the programme a vital part of Samsung’s globalisation.

Honda learned by selling there. Samsung paid to learn how people lived there. Neither mistook a visit for knowing the market.

The three-day expansion

For many startups, “global” means three days a year. A booth bought with a government subsidy. A conference someone else organised. Then the report home: productive trip. A few dozen meetings. Hundreds of cards.

Trade shows are easy to count. Meetings held. Cards collected. Samples requested. Subsidy spent. But what did the founder learn about the customer? There is no column for that.

The trip fills the CRM. The customer stays a stranger.

The customer never came to the booth. The founder met the people whose job is to explain a country’s market, and they relay a life they don’t live. If your overseas strategy came out of their mouths, it isn’t a strategy. It’s hearsay. The company’s next three years may depend on a second-hand account of the market.

Also Read: The frontier AI money is getting picky. Southeast Asia’s AI founders should price that in now

Why nobody goes

A week on a foreign street is harder to count. It is also uncomfortable. No shared language. Strangers who can’t imagine why they’d need your product. The hotel and the booth are the safest ground a founder has abroad.

But the discomfort is where the questions start. Questions that never come up in a convention hall. You might learn why no one needs your product. Or why they need it for something you never imagined. Either answer is worth more than 400 cards.

Skip it, and everything downstream rests on second-hand knowledge. A country manager hired by someone who has never met a customer. Local pricing set from a distributor’s spreadsheet. Local marketing built on a consultant’s slide. We call it going global.

Four hundred cards, and not one customer’s day understood. That founder didn’t enter a market. He took a business trip, and burned through more of a shrinking runway.

One question

In the country you want to enter, can you walk through one customer’s day, morning to night? What’s their name?

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

Join us on WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post Four hundred business cards. Zero customers appeared first on e27.

Posted on — Leave a comment

Airwallex’s US$1.5B Singapore pledge dwarfs the island’s entire fintech funding haul

Airwallex wants Singapore to know it is staying put. The cross-border payments firm said on Wednesday it will invest more than SGD2 billion (about US$1.5 billion) in the city-state over the next five years and grow its local headcount from over 750 to more than 1,000.

The move cements Singapore as one of its two global co-headquarters, alongside San Francisco.

The pledge coincided with the official opening of its expanded office at Guoco Tower in Tanjong Pagar, which the company describes as Singapore’s tallest building. A new floor on level 35 is now occupied and another is due in early 2027, taking Airwallex’s local footprint to more than 80,000 sq ft across four floors.

Also Read: Growth-minded Singapore SMEs turn to fintech amid cost pressures: Airwallex survey

On paper, it is a hefty vote of confidence at an awkward moment. The US1.5 billion figure is roughly three times what every Singapore fintech raised combined in H1 2026. That half-year haul, US$499 million across 53 deals, was the city-state’s weakest first-half fintech showing in close to a decade. Nearly two-thirds of it came from a single cheque: Airwallex’s own Series H.

What US$1.5 billion actually buys

Airwallex did not break the figure down. The release says the money will support workforce expansion, product and technology capabilities, and the strengthening of “global and regional operations”. That is a broad bucket. Over five years, it could easily absorb salaries, office leases and compliance costs that a fast-growing company would incur anyway.

The hiring target deserves the same scrutiny. Adding roughly 250 roles over five years works out to about 50 a year. By comparison, Airwallex’s global headcount went from more than 2,300 when it announced its US$320 million Series H in June to more than 2,700 today. That is some 400 people in about five months. Singapore is growing, but it is not where most of the new desks are going.

What the company stresses instead is the kind of jobs. Airwallex says it will build “specialised, high-value capabilities” in product development, engineering and AI.

The Singapore Economic Development Board (EDB), which worked with the company on the expansion, struck the same note. “We welcome Airwallex’s continued investment in Singapore, which will significantly increase its footprint and anchor more product and AI capabilities here,” said Herman Loh, Senior Vice President and Co-Head, Digital Industry Singapore at EDB.

That is the more meaningful signal. Singapore’s pitch to global firms has long rested on being where regional decisions get made. Anchoring engineering and AI work, rather than just sales and compliance teams, is what keeps that pitch credible as other financial centres court the same companies.

Ninety per cent, of what?

The growth numbers are eye-catching. Airwallex Singapore reported 90 per cent year-on-year growth in both revenue and total transaction volume in H1 2026. It did not disclose absolute figures, so it is hard to tell how much of that is a modest base compounding quickly.

Group-level numbers offer some perspective. In June, Airwallex said it had reached US$1.3 billion in annualised revenue in March, up 74 per cent, and US$287 billion in annualised transaction volume, up more than 120 per cent. The periods are not like-for-like, but Singapore appears to be outpacing the group on revenue while trailing it on volume.

Also Read: The hidden margin killer: Why CFOs are rethinking cross-border payment infrastructure

The company credits demand from local businesses that want their finances in one place. “Singapore SMEs increasingly value integrated financial solutions that bring their financial operations together on one platform,” said Lionel Tan, Head of SME and Growth at Airwallex. He added that interest in AI tools for managing finances is also rising.

The customers it names are telling. Far East Flora, Stryv, Circles, Moom Health, ZENXIN Organic and Chronos Agency span floristry, food, telecoms, health and marketing. That mix suggests Airwallex is reaching well beyond venture-backed startups into traditional SMEs taking their first steps overseas.

A crowded corridor out of Singapore

Airwallex is hardly alone in treating Singapore as a launchpad. Homegrown unicorn Nium has been pushing into stablecoin rails, most recently by acquiring Cypher. Circle’s US$400 million deal for Tazapay showed that stablecoin giants want a slice of regional flows too. Aspire courts the same SMEs with business accounts and spend tools, while Stripe, Wise and Adyen fight for merchants.

Airwallex’s answer is regulatory depth. It holds more than 85 licences worldwide and runs its own settlement integrations. It has also shown it will buy its way into markets rather than wait, as with its acquisition of Paynuri in South Korea.

From Singapore, the firm is expanding into Malaysia, Indonesia and Vietnam. In these markets, payment habits, foreign exchange rules and licensing regimes differ sharply, and cross-border payments remain fragmented enough that licences matter more than marketing.

A co-headquarters is a promise, not a guarantee

“We’re not just expanding our presence in Singapore, we’re investing in the talent, technology and capabilities that will power our next phase of global growth,” said Arnold Chan, General Manager, Asia-Pacific.

Also Read: Before the licence: What Singapore’s regulated institutions ask a fintech partner first

The open question is how much decision-making weight “co-headquarters” really carries. Dual-HQ structures tend to tilt over time. Airwallex earmarked its Series G to power US expansion, so San Francisco will exert a strong pull.

Still, in a year when fintech capital in Singapore has thinned, a company committing multi-year spending, four floors in a landmark tower and AI engineering roles is a signal EDB will gladly amplify. The real test comes in 2031: whether the 1,000 jobs materialise, and how many of them build products rather than sell them.

The post Airwallex’s US$1.5B Singapore pledge dwarfs the island’s entire fintech funding haul appeared first on e27.

Posted on — Leave a comment

Ecosystem Roundup: Nvidia board probes Jensen Huang over affair with Megaspeed exec

Nvidia’s board has opened an independent investigation into Jensen Huang after the CEO admitted in a staff memo to an affair with Alice Huang, an executive at Singapore-based Megaspeed. A memo circulated among US policymakers earlier this year claimed China’s Ministry of State Security had recruited her to spy on him. Beijing’s embassy in Washington has dismissed the allegations, and no one has been charged.

The board said it has no reason to believe national security or company secrets were compromised, and that it retains confidence in Huang. Independent directors will run the probe with outside legal counsel.

The Singapore angle is what the region should watch. Megaspeed, a cloud company that grew out of a Chinese gaming business, became Nvidia’s largest buyer in Southeast Asia, purchasing roughly US$2 billion worth of chips, according to The New York Times. That reporting raised questions over whether the chips reached Chinese users in breach of US export rules, claims Nvidia says it investigated and debunked.

The timing could hardly be worse. Nvidia reports earnings on 17 November, and three House committee chairs want Huang at a closed-door briefing a day later. For Southeast Asia’s data centre and AI infrastructure players, who rely on steady access to Nvidia hardware, any tightening of export scrutiny would land close to home.

Read the full article here.


REGIONAL

Thailand files criminal complaint against Bitget, moves to block it: The SEC alleges operator BG Limited served Thai users without a licence from 2022 to 2025, and wants access cut from December 5. The case clouds Bitget’s bid to buy licensed exchange GMO-Z.com.

A quarter of Filipino workers use AI agents, but only 15% of firms do: Gobi Partners puts worker adoption above the 16% global average, yet AI startups took under 1% of 2026 deal value. IBPAP has cut its 2028 headcount target from 2.5M to a best-case 2.14M.

Equinix opens US$290M SG6 data centre in Singapore for AI workloads: The first phase adds about 1,550 cabinets and scales to 20MW of IT load. SG6 is Equinix’s first site approved under Singapore’s Data Centre Call for Applicationframework, which rations new capacity by efficiency and economic value.

Chocolate Finance takes its cash management app to Hong Kong: The Singapore fintech, which manages over US$1.25 billion for more than 130,000 users, offers a 3.8% annual return on the first US$12,800 deposited and will cover any shortfall until June 2027.

U Mobile opens 5G-A sandbox to Cradle startups, minus the cheques: Malaysia’s second 5G operator brings test kit, engineers and enterprise introductions, but no guaranteed funding or contracts. The real test is whether pilots turn into paid deals as Singtel writes bigger cheques.

Shopee says Singapore affiliate sales grew nearly 50% in a year: Its local affiliate network now exceeds 300,000, and creator content drives one in 10 platform visits. The figures are unaudited, and Shopee disclosed no merchandise value ahead of its 11.11 campaign.

INTERVIEWS & FEATURES

Gobi pitches one-person AI firms to fill Philippines’ Series A gap: Philippine Series A value has fallen 98% this year. Gobi argues AI-native solo founders can validate ideas on US$5,000 to US$20,000, though 86% of the country’s 8,640 one-person corporations remain traditional service businesses.

Yale’s Sonnenfeld puts Jensen Huang’s survival odds at 50-50: Nvidia has no obvious successor and no dual-class shares shielding its founder. Its role as AI’s chief financier, with stakes in OpenAI, Anthropic and xAI, means a leadership crisis could ripple across the industry.

Data centre backlash and Musk’s losses top tech’s midterm hangover: TechCrunch’s Equity team notes voters approved most data centre restrictionson local ballots, while Musk’s half-billion-dollar Republican spend bought little. OpenAI has already dropped NDA demands and rebranded sites as compute campuses.

INTERNATIONAL

South Korea fines Coupang US$220M over food delivery dominance: The Korea Fair Trade Commission says Coupang required merchants to match termsoffered on rival platforms and penalised those who refused, curbing their pricing freedom and hampering rivals such as leader Baedal Minjok.

India’s NPCI launches AI and blockchain technology subsidiary: NPCI Tech Solutions, led by former CTO Vishal Kanvaty, is the UPI operator’s fourth wholly owned unit. It will build digital public infrastructure services such as eKYC Setu; budget and revenue targets remain undisclosed.

Tesla to end Cybertruck production in 2027 to build more Cybercabs: Musk once forecast 250,000 annual sales; Tesla has sold only tens of thousands in three years. It is the third model Tesla has axed, after the Model S and X, as it pivots to robotaxis and robots.

SNL’s new Elon Musk impression lands after a rough week for Musk: Jeremy Culhane plays Musk as desperate to be liked after his half-billion-dollar midterm bet flopped and the Cybertruck’s demise. SNL has also skewered Anthropic’s Dario Amodei and data centre backer Kevin O’Leary this season.

CYBERSECURITY

Fake Wi-Fi and a bogus security workshop drained traders at TOKEN2049: ‘Evil Twin’ hotspots and a side event that pushed remote-control malware let attackers empty hot wallets in Singapore, including over US$4M from trader Frogman. The writer urges eSIMs, VPNs and cold storage.

Intrusion at SoftBank’s IDC Frontier hit 495 cloud customers: SoftBank Corp. says attackers took no data from customers’ virtual servers in the October 7 incident, which hit corporate and local government users. It still lifted its FY2026 revenue forecast to US$49.9B.

Binance CFO Kaiser Ng exits a week after US$6.3B North Korean hack: The FBI blamed North Korea’s TraderTraitor group for the breach. Once accounts unfroze, users pulled a net US$3.6 billion within hours. Justin Yek takes over, and Binance says it will absorb all losses.

SEMICONDUCTOR

Nvidia, SK Telecom bring Korean startup Rebellions into sovereign AI: Rebellions’ R200 chips will plug into Nvidia’s NVLink Fusion and MGX racks, handling decode while Nvidia GPUs run prefill. Seoul backs the project, but the systems will only be ready for deployment in 2028.

Analog Devices closes US$1.35B buy of edge-AI chipmaker Alif: The deal brings Alif’s low-power AI microcontrollers and its Singapore and Bengaluru engineering teams into ADI. ADI has not said how those teams will be integrated or whether US$200 million in contingent payments were triggered.

US memo claims Megaspeed exec was recruited to spy on Nvidia’s CEO: Separately, The New York Times found New Matrix, a Chinese firm cleared to buy Nvidia chips, listed an office that was an empty shell, raising fears it fronts a blacklisted private equity firm.

AI

OpenAI reportedly raises US$11B at a US$1.25 trillion valuation: The round could reach about US$20 billion, far below September’s US$30B goal. Potential buyers of Anthropic’s IPO shares are also eyeing a lower valuation as midterm results rattle AI stocks.

Ramp data shows enterprise AI spending stalling as token use climbs: Token spend grew just 1.5% across August and September, against 42% in July, as price cuts and tighter budgets bite. Ramp’s economist says the hypergrowth era may be over just as Anthropic prepares to list.

UBTech, Tianyi Telecom aim to push home robots past 1M units: The Shenzhen robotics maker signed the strategic pact with Tianyi Telecom Terminals on November 9, also setting 100,000-unit milestones for highly biomimetic and wheeled humanoid robots. The targets cover the wider industry, not just their own shipments.

THOUGHT LEADERSHIP

SEA healthtech funding fell 85% in two years. AI hasn’t rescued it: Funding dropped from US$397M in 2023 to US$59.7M in 2025. The author argues founders must prove paying customers, measurable value and repeatable deployments, because unpaid hospital pilots are not product-market fit.

Why SEA healthtech struggles for funding despite the AI boom: Regional healthtech funding fell from US$397 million in 2023 to US$59.7 million in 2025. The author argues founders must prove paying customers, repeatable deployments and measurable value, not just impressive AI demos.

The Indonesian tax traps foreign startups underestimate: New PTs lost access to the 0.5% turnover tax in April 2026. The author flags permanent establishment risk from sales-closing local hires, 20% withholding without treaty paperwork, and low transfer pricing thresholds.

One input window is replacing dashboards, but code must check AI: Lobby founder Serge Berezhnoy runs operations through an agent on a Linux server. After a model pitched a cat-meme creator as a beauty influencer, his team moved validation into deterministic code.

AI became my extra team member, but it still needs managing: The Wellness Insider founder Melissa Fann uses AI for first drafts, campaign plans and sponsorship pitches. Her verdict: vague instructions produce vague results, and responsibility for errors still sits with her.

Founders stuck on ‘almost’ are deferring a loss, not stalling: Priyanka Raina argues indecision shields founders from closing off possible futures. Rather than listing obstacles, she has clients picture the decided state vividly until the pull of that future outweighs the fear.

The post Ecosystem Roundup: Nvidia board probes Jensen Huang over affair with Megaspeed exec appeared first on e27.

Posted on — Leave a comment

GCash’s record quarter shows Mynt is now a lender that happens to run a wallet

Three weeks after a stock market debut that ended flat, Mynt has handed investors the number they were told to wait for. The parent of GCash, the Philippines’s dominant e-wallet, said its third-quarter revenue hit ₱24.0 billion (about US$400 million), up 16.6 per cent year-on-year and the highest in its history.

The timing matters. When Mynt listed on 20 October after raising ₱60.9 billion (roughly US$1 billion) in the largest IPO in Philippine history, local brokers advised would-be buyers to hold off until the Q3 numbers landed. Their worry was simple: the hype was loud, and Mynt’s second quarter had slipped from the first. The stock, priced at ₱6.60, a hefty 34 per cent below its ₱10 ceiling, closed its first trading week at ₱6.27.

Also Read: The Philippines doesn’t need more fintech apps; needs rails

On paper, the quarter answers that worry. Nine-month revenue rose 12.2 per cent to ₱67.1 billion (about US$1.1 billion), quicker than the 9.8 per cent growth posted at the half-year mark. Net income for the period climbed 17.4 per cent to ₱16.5 billion (about US$275 million), up from 7.3 per cent growth in the first half. Net margin stood at a healthy 24.5 per cent.

The loan book is outrunning the user base

The more revealing numbers sit further down the release. Monthly active users rose 9.7 per cent to 42.1 million, which is respectable but hardly explosive for an app that is already the country’s default wallet. Daily transactions averaged 64.4 million over nine months, up 16 per cent, so existing users are doing more rather than new ones flooding in.

Payments gross transaction value, the total money flowing through the app, grew 23.1 per cent to ₱15.1 trillion (about US$252 billion). The standout, though, is credit. GCash’s loan portfolio swelled 43.3 per cent to ₱79.3 billion (about US$1.3 billion), growing more than four times faster than its user base. Active borrowers reached 7.8 million, roughly one in every five monthly users.

That arithmetic tells you what Mynt is turning into. Payments win the users; lending makes the money. It is the same quiet pivot Grab’s own third-quarter numbers laid bare last week, and part of a wider regional shift in which fintech, having solved distribution, now has to scale on the balance sheet.

The non-performing loan (NPL) ratio, the share of loans where borrowers have stopped paying, held at 4.8 per cent, flat on the previous quarter. Mynt calls that sustained asset quality. A more cautious reading is that it is stable, not low. Nearly one in every 20 pesos lent has gone bad, a sign that GCash is lending to customers traditional banks often skip. That is both the social mission and the risk. The average loan works out to roughly ₱10,200 (about US$170): small-ticket, high-frequency credit that can sour quickly when household budgets tighten.

What the press release does not say

Two caveats deserve attention. First, the release claims record quarterly net income but discloses profit only on a nine-month basis. Investors therefore cannot cleanly set Q3 profit against Q2, the very comparison analysts flagged before listing.

Second, the cost line is flattered. Mynt says operating expenses grew in line with revenue, but most one-off IPO costs will be booked in the fourth quarter, alongside what it calls seasonally higher costs. Q4, in other words, is where the margin story gets its real test.

Also Read: MoneyHero’s Q2 exposes the rising cost of fintech growth in Southeast Asia

The macro backdrop is not friendly either. The peso hit a record low in October and the Philippine Stock Exchange index sank to an 11-month low, with foreign investors pulling about US$623 million out of local shares this year.

Chief Financial Officer Evelyn Ng nonetheless struck a confident note, arguing that GCash has shown “there is good business in doing good” and that sizeable room remains to grow in both payments and lending.

A crowded fight for Filipino borrowers

GCash is not lending into an empty field. Maya, the PLDT-backed fintech under Voyager Innovations, runs its own wallet and digital bank. Specialist lenders are pushing in too: Salmon is betting US$140 million on secured lending, while Grab’s US$1.49 billion Atome deal brings a regional buy-now-pay-later player with Philippine ambitions under a superapp roof.

GCash’s edge is distribution: 42 million monthly users and a stream of transaction data no bank can match, built on regulatory foundations that, as one e27 contributor argued, are what let Philippine fintech scale this fast. Its weakness is concentration. Unlike Grab or Sea, every one of Mynt’s pesos is earned in a single market, exposed to a single currency and a single central bank.

April’s lock-up is the next exam

For the Philippine Stock Exchange, which has struggled to win tech listings in a year when Malaysia carried Southeast Asia’s IPO market, these numbers are welcome relief. More than 140,000 GCash users, many of them first-time investors, bought into the IPO through the app’s own trading platform. Their money now rides on the same loan book the company is racing to grow.

Also Read: Southeast Asia’s fintech apps don’t have a literacy problem, they have a fear problem

The quarter proves Mynt can grow profit faster than revenue even as user growth cools. It does not yet prove the bad-loan line can stay flat while the loan book expands by more than 40 per cent a year. The fourth-quarter results, and the expiry of the 180-day lock-up on pre-IPO shareholders in April, will show whether the market believes it can.

The post GCash’s record quarter shows Mynt is now a lender that happens to run a wallet appeared first on e27.

Posted on — Leave a comment

AquaBloom raises seed funding to turn Indonesia’s seaweed into climate armour for farmers

Along Indonesia’s coasts, smallholder farmers tie seaweed to lines strung across shallow water. A few hundred kilometres inland, other smallholders watch El Niño bake their rice paddies.

Jakarta-based AquaBloom wants the first group’s harvest to rescue the second group’s yields, and it now has fresh capital to test whether that idea scales.

The agri-biotech startup has raised an undisclosed seed round led by Spiral Ventures, with participation from Potato Impact Partners (PIP), Archipelago VC and several angels.

The money will go into expanding manufacturing capacity, R&D and field trials, and distribution across Indonesia.

Also Read: Thirty climate tech startups in Southeast Asia that would rather build than pledge

Founded in 2024 by CEO Michelle Arsjad, AquaBloom makes seaweed-based biostimulants for crops and feed additives for livestock. It previously raised from Altrui Investment Management, among others.

Indonesia ships the seaweed, and the value with it

Indonesia is the world’s second-largest seaweed producer after China, yet most of the crop leaves as a low-priced commodity. According to UNCTAD, close to two-thirds of the country’s seaweed exports are raw, which caps how much value stays at home.

Between January and October 2025, Indonesia exported US$264.6 million worth of seaweed, mostly dried seaweed and carrageenan, with China the dominant buyer at US$183.6 million.

Jakarta knows this. In October, Luhut Binsar Pandjaitan urged faster seaweed downstreaming so the commodity is turned into value-added products rather than shipped out raw. AquaBloom is pitching itself as one small answer to that policy itch.

Biostimulants are not fertilisers. Rather than feeding the plant directly, they help it absorb nutrients more efficiently and cope with stresses such as heat and drought. The promise for farmers is better yields with less synthetic fertiliser, which matters when input costs swing and subsidised supply is unreliable.

A pitch built on yield, and on El Niño

The target customer is the smallholder who has been squeezed from both sides: rising costs and an increasingly erratic climate. Other startups have spotted the same pressure point, from low-emission rice farming to precision agri-drones.

“Smallholder farmers need and deserve tools designed for the realities they face, especially as the climate makes farming less predictable,” Arsjad says. “The seaweed we use is grown by smallholder farmers along Indonesia’s coasts. I find it beautiful that smallholder farmers at sea can help smallholder farmers on land grow stronger crops.”

Also Read: Life Lab Resources grabs US$1M to turn food waste into aquaculture feed

The company says its products have been validated on almost 30 crops across more than 40 field trials in Indonesia, with yield improvements of 20 to 30 per cent. It points to rice plots hit by El Niño this September, where treated paddies showed taller shoots and more tillering than adjacent control plots.

Those are striking numbers although AquaBloom has not share the details such as who ran each trial, across how many seasons, or how results varied by crop. Yield gains in a drought year can look very different in a normal one.

Why the investors signed the cheque

Spiral Ventures’s Irawan Ganda, Partner and Head of Southeast Asia, points to the regulatory groundwork: three products already registered for sale in Indonesia and third-party trials backing the yield claims. He also highlights a team whose product people came out of Bayer, Syngenta and Corteva. Spiral plans to open doors to Japanese and regional agriculture groups.

PIP’s Investment Lead Tessa Charupatanapongse cites the product’s range across almost 30 crops and the distribution networks already reaching smallholders.

The most telling detail comes from Archipelago VC’s Principal Olivier Lentze. His firm first met Arsjad in 2025, before AquaBloom had earned a single rupiah. Over the following 18 months, he says, the team landed larger-volume clients, drove farmer adoption and increased production while holding prices accessible. This means AquaBloom’s commercial track record is roughly a year and a half old.

Imported incumbents, and the hard part of selling to smallholders

AquaBloom’s main argument against rivals is provenance. Established seaweed biostimulant brands sold in Indonesia are largely imported. Globally, Canada’s Acadian Plant Health and South Africa’s Kelpak have built businesses on cold-water kelp, while India’s Sea6 Energy works with tropical seaweed. The agrochemical majors also run biologicals divisions. A local product made from local seaweed, tuned to local crops, could undercut imports on price, but the incumbents have deeper pockets and longer trial histories.

The tougher fight may be distribution. Indonesia’s first generation of agritech learned that reaching smallholders is expensive and extending credit to them can be fatal. Farmers also have to be persuaded to pay for an input whose benefits are less visible than fertiliser’s, in a sector already battling a youth exodus from farming.

Also Read: Why Indonesia’s agritech winners will be phygital, not purely digital

Some in the sector argue that the farmer need not always be the one paying, with offtakers, retailers or corporates footing part of the bill. AquaBloom’s plan to grow through retail partners, and Spiral’s push towards regional agriculture groups, hint at that route. A strategic buyer may also be the likeliest exit, since Southeast Asian agritech is arguably better built for acquisitions than IPOs.

Over the next 12 to 18 months, AquaBloom plans to scale production and widen retail distribution. Investors in Southeast Asian climate tech have heard plenty of impact stories. The one that matters here will be told by a shallot grower deciding whether to buy a second bottle.

The post AquaBloom raises seed funding to turn Indonesia’s seaweed into climate armour for farmers appeared first on e27.

Posted on — Leave a comment

SoftBank and Grab put US$1.85B down on a 5GW AI bet in Malaysia

Grab started life in Malaysia as a taxi-booking app. Fourteen years on, it is heading back to its birthplace with a far heavier load: AI data centres.

SoftBank Group and Grab Holdings have signed joint development agreements with Petroleum Sarawak (PETROS), the Sarawak state oil and gas company, to build AI infrastructure across Sarawak and Peninsular Malaysia. The initial seed capital is US$1.85 billion: US$1 billion from SoftBank, US$750 million from Grab and US$100 million from PETROS.

Also Read: Malaysia’s sovereign AI bet: Local context becomes the next startup moat

The long-term target is five gigawatts or more of AI compute, which is where the maths gets interesting. Industry estimates routinely put the all-in cost of a single gigawatt of AI capacity, chips included, in the tens of billions of US dollars. The seed cheque, in other words, is a down payment on a project whose final bill nobody has published.

From a Kuching handshake to binding cheques

A month ago, this was a statement of intent. In October, the three parties signed a framework agreement in Kuching to explore an AI platform in Sarawak, with no capital attached. The new agreements, the companies say, carry binding development and funding obligations.

Binding, with caveats. The joint venture entities do not yet exist, and each party’s investment terms still hinge on definitive agreements, due diligence, regulatory approvals, and securing land, affordable power, connectivity and contracted customers. Money will be released in stages, tied to development and customer milestones. The risk disclosures run longer than the announcement.

The two projects are built differently. In Sarawak, SoftBank is the majority partner, putting up US$800 million of the US$1 billion initial investment, with Grab and PETROS adding US$100 million each. Grab holds an option to raise its stake, diluting SoftBank, which would still keep control.

On the peninsula, the roles flip. Grab is the founding shareholder with US$650 million of the US$850 million seed, SoftBank adds US$200 million, and Grab leads development. Further growth is meant to come from third-party equity and non-recourse project financing, debt repaid from the project’s own cash flows rather than its sponsors’ balance sheets. The companies have not said where on the peninsula the facilities will sit, a notable silence given how Johor has absorbed much of the region’s data centre spillover, and the scrutiny that came with it.

A superapp takes on a power-hungry side hustle

For Grab, US$750 million is a hefty commitment for a business whose bread and butter is rides, meals and, increasingly, loans. Its third-quarter results showed net profit of just US$18 million against a record Adjusted EBITDA of US$237 million, while gross cash liquidity slipped to US$7.1 billion as buybacks drew on the pile.

Grab is also on the hook for its US$1.49 billion deal for 60 per cent of Atome Financial. Add compute to the list and its capital now chases driver incentives, a fast-growing loan book, share repurchases and data centres. Investors will want to know which gets priority when cash tightens.

Also Read: Southeast Asia’s US$1T digital dream slips five years, and AI is the new reason to keep believing

Grab’s case is that it needs the compute itself, to speed up localised AI products for its users. Co-founder and CEO Anthony Tan will oversee both projects while continuing to run Grab, with dedicated teams handling day-to-day operations. Grab’s board has approved the arrangement. Still, running a Nasdaq-listed superapp and two infrastructure ventures is a heavy load for one executive, and the overlap invites related-party questions: Grab will be shareholder, likely customer and, through its CEO, steward.

SoftBank, one of Grab’s longest-standing backers, has sweetened the deal by agreeing in principle to a five-year lock-up on Grab shares held directly by a SoftBank entity, subject to customary exceptions. Lock-ups rarely feature in infrastructure announcements. This one reads as a signal that the partnership is as much about cementing a shareholder relationship as about megawatts.

Malaysia’s compute queue keeps getting longer

The consortium joins a crowded line. Just last week, Google committed US$15 billion to Malaysian digital infrastructure across 2027 and 2028. Microsoft has pledged US$2.2 billion, AWS about US$6.2 billion through 2037, and Oracle US$6.5 billion for a public cloud region. Regionally, AI infrastructure dealmaking hit a record this year as Singapore, Indonesia and Thailand chase the same capacity.

What sets this consortium apart is its business model. Hyperscalers build for their own clouds; SoftBank and Grab plan to build capacity and then court hyperscalers as tenants, alongside governments, multinationals and small businesses. That makes demand the central risk. Without signed customers, five gigawatts is a forecast, not a facility.

The demand case leans on the e-Conomy SEA 2026 report by Google, Temasek and Bain & Company, which puts per capita AI usage in the region’s six largest economies at about a tenth of US levels, with usage projected to grow up to 60-fold by 2030. Projections of that magnitude have a habit of being revised.

Sarawak’s gas, and the numbers nobody gave

Sarawak’s pull is energy. PETROS brings power supply and local supply chain links, and the project is pitched as serving both the state’s Post COVID-19 Development Strategy 2030 and Putrajaya’s AI Nation 2030 plan. The companies promise facilities that follow Malaysia’s renewable energy guidelines and use designs that cut water use and reliance on local grids.

Also Read: SEA’s AI boom has a water problem it cannot offset away

They did not say how much power or water the sites will draw, or how much of that power will come from gas. Google, by contrast, pledged last week to pay its own electricity bill and fund lower household tariffs near its campuses. The communities living beside Malaysia’s next server halls will want similar numbers.

The promised high-value jobs in engineering, design and technical fields, plus talk of advanced manufacturing, also arrive without figures. Malaysia wants 300,000 AI jobs by 2030, yet data centres are famously lean to run once built.

For now, Malaysia gets another blockbuster headline and Grab gets a seat at the infrastructure table. Whether this becomes five gigawatts or a well-funded pilot will be decided by the milestones buried in the fine print.

The post SoftBank and Grab put US$1.85B down on a 5GW AI bet in Malaysia appeared first on e27.

Posted on — Leave a comment

What happens to the subscription economy when AI agents review every payment?

Subscription businesses have spent the past two decades making revenue more predictable. Instead of selling a product once and persuading the customer to buy again, they persuade the customer to keep paying every month or every year. That creates an important advantage: once the subscription has started, continuing to pay usually requires no decision at all. Stopping does.

Research published in the American Economic Review shows how valuable that inertia can be. Across 10 large subscription services, monthly retention fell four times faster when customers were forced to make an active renewal decision. The researchers estimated that consumer inertia increased seller revenue by 87 to 120 per cent on average, depending on the model used.

AI agents could change this advantage by giving continuous attention to every recurring payment, checking whether the service is still being used, whether the price has risen and whether a better alternative exists. Recurring billing starts to become recurring justification.

The subscriptions most exposed are those we stop thinking about

Some subscriptions pay for products customers use constantly and consciously. A Spotify user who listens every day does not need an agent to explain why they are paying. Others, including gyms, dating apps, creative software and some media products, are useful intermittently, so the customer’s need can fade before the payment stops.

A third group is more exposed. PDF tools, résumé services, travel apps, photo utilities or mod-download services may solve a problem that lasts minutes, days or weeks while billing continues for months.

RevenueCat’s 2026 analysis of more than 115,000 subscription apps and over US$16 billion of revenue shows how different these businesses already are. Productivity apps had a median first annual renewal rate of only 23 per cent, while business applications were substantially stickier. AI apps also retained monthly subscribers worse after 12 months, at 6.1 per cent compared with 9.5 per cent for non-AI apps. The key distinction is whether the customer’s need itself genuinely recurs.

Agents make that easier to see by combining the recurring charge with renewal terms, actual use and available alternatives. Instead of requiring the customer to assemble the information, the agent can reduce it to one decision: keep, downgrade, switch, renegotiate or cancel.

Also Read: AI agents broke into 100 firms for pocket change. SEA’s online shops fit the target profile

Losing customers can make the remaining customers more expensive

Subscription businesses often benefit from customers who continue paying while using very little of what they bought.

Gyms provide an easy example. Research on US health-club members found that people on monthly contracts attended only 4.3 times a month on average. They were effectively paying more than US$17 per visit when a US$10 pay-per-visit option was available. If an agent repeatedly compares the membership fee with actual attendance, those low-use members should be more likely to question whether the subscription still makes sense.

The members who remain are more likely to be the people who actually use the gym, so revenue can fall much faster than visits. The gym still needs enough equipment, floor space and staff to serve them, worsening the economics from both directions.

The same effect can appear in software. A company might buy 1,000 licences even though only 600 employees use the product seriously. If an agent cuts the contract to 650 seats, the supplier loses 35 per cent of its paid seats but may lose very little actual usage. For software running on the customer’s computer, that is mainly a revenue problem. For cloud services and AI tools, where use creates real compute costs, it can become a margin problem as well.

Price increases create another value test

A company facing that squeeze may raise prices, but AI agents make that harder too. A customer might ignore a service increasing from US$10 to US$13 a month. An agent can immediately compare the higher price with recent usage and competing products.

That creates a feedback loop. Low-use customers leave, average usage among the remaining customers rises, margins come under pressure, and the supplier raises prices. Those increases trigger another round of comparison, cancellation and switching. A company could therefore maintain headline revenue while the economics underneath it deteriorate.

Also Read: Banks want AI agents to move money, as long as someone is accountable

Some products may stop making sense as subscriptions

This eventually raises a bigger question: should some products be subscriptions at all?

Software companies have spent years moving from packaged products toward subscriptions because recurring revenue is more attractive than repeatedly asking customers to buy a new version. Agents could push some products back in the opposite direction. If a desktop utility costs almost nothing to provide after installation and customers only need it occasionally, charging US$60 for a permanent licence may become more attractive than repeatedly asking an agent to approve another US$10 monthly payment.

Cloud products face a different problem because heavy usage has a real cost. They may move toward credits, metered consumption or subscriptions combined with usage charges. NVIDIA’s GeForce NOW already limits Performance and Ultimate customers to 100 hours of premium playtime each month, with additional hours available separately. The company says the limit helps maintain quality, speed and shorter queue times.

Gaming illustrates the difference clearly. A US$70 game running mainly on the customer’s console creates relatively little continuing computing cost for the publisher. A cloud gaming service must keep paying for processors, bandwidth and infrastructure every time the customer plays. If low-use subscribers disappear while power users remain, outright purchases, expansions, sequels and usage-based pricing become relatively more attractive.

Businesses face the same mechanism at a much larger scale

The same logic applies to corporate spending, where hundreds or thousands of subscriptions can accumulate. xAI gave its Grok procurement agent access to contracts, spending and usage information. It found 43 paid software seats with no activity for 90 days and identified more than US$100,000 of direct savings across the spending it reviewed.

This turns occasional procurement reviews into a continuous process. An agent can check unused seats, overlapping products, alternative suppliers, renewal prices and whether the company is still buying what employees actually use. The relationship between employee numbers and software revenue consequently becomes less secure.

Gartner calls a broader version of this “agentic arbitrage” and estimates that up to US$234 billion of enterprise application spending could be exposed by 2030, around 20 per cent of projected enterprise application software-as-a-service spending. That does not mean US$234 billion disappears. It means a very large pool of recurring expenditure becomes easier to question, optimise and renegotiate.

Also Read: AI doesn’t need crypto, AI agents do

The subscription economy is larger than it looks

US consumers report spending roughly US$1,300 to US$1,900 a year on subscriptions and recurring services, depending on the survey and definition used. US software-as-a-service revenue alone was estimated at more than US$160 billion in 2025. The affected spending extends beyond obvious subscriptions. An agent may not cancel electricity or car insurance, but it can compare tariffs, shop an insurance renewal or find a cheaper mobile plan.

That does not mean subscriptions disappear. Services providing strong continuing value should survive repeated scrutiny, while bundles may become more valuable because they give customers several reasons to keep paying.

What becomes harder to defend is recurring revenue supported by inattention, underuse and the effort required to reconsider a purchase. The subscription economy was built around recurring billing. AI agents may turn it into an economy of recurring justification.

Read a more detailed version of this article, with full sources, here.

Read “Abundance Economics: The Economic Mechanisms of the AI Age” here, or find more on the research framework behind it here.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

Join us on WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post What happens to the subscription economy when AI agents review every payment? appeared first on e27.