Grab wanted Taiwan to be its first market outside Southeast Asia. Taiwan’s competition regulator has told it, in effect, that it is not really an outsider.

Taiwan’s Fair Trade Commission (TFTC) on Wednesday blocked Germany-based Delivery Hero’s planned sale of foodpanda’s Taiwan business to Grab, scuttling a US$600 million all-cash deal agreed on March 23. The commission said it had discussed possible remedies with the parties, but the commitments on offer were not enough to offset the damage to competition.

Also Read: Grab’s US$1.49B Atome deal signals a deeper race for SEA’s credit economy

Delivery Hero said the decision does not affect its right to appeal. Grab said it was disappointed, arguing the ruling stops it from bringing new services, better earning opportunities and improved service standards to Taiwanese consumers, couriers and merchants. It added, a little pointedly, that its core business across Southeast Asia “remains robust”.

This is the second time in under two years that Taipei has stopped foodpanda Taiwan from changing hands. In December 2024, the regulator blocked Uber’s roughly US$950 million bid for the same business because the combined entity would have controlled more than 90 per cent of the market. Grab’s bid was meant to be the antidote: an entrant with no Taiwanese operations, buying foodpanda to keep the island’s two-horse race alive. When the deal was announced, e27 argued it could save Taiwan from a delivery monopoly. The TFTC has concluded that the two horses would share a stable owner.

Uber in the room

That owner is Uber. The US company holds about 13.1 per cent of Grab, a legacy of the 2018 deal in which Grab swallowed Uber’s Southeast Asian business, and an Uber executive previously sat on Grab’s board. Grab’s dual-class share structure caps Uber’s voting power at 3.9 per cent, but the commission was unmoved.

Its reasoning runs on four tracks:

1. Uber would pocket a slice of Grab’s profits, which blunts Uber Eats’s incentive to undercut a Grab-owned foodpanda with cheaper fees or promotions.
2. The shareholding changes Grab’s own incentives, voting caps notwithstanding: it is hard to wage a price war on a company that owns a chunk of you.
3. The two companies have worked closely together in other markets, which the TFTC said makes coordination easier to predict and sustain.
4. Taiwan’s delivery market is difficult to break into, because a newcomer needs deep pockets and enough restaurants, couriers and customers all at once, leaving little room for anyone to punish a comfortable duopoly.

The third point is the sharpest. The commission singled out the 2018 Grab-Uber transaction, saying both parties knew or should have known it broke competition law. That is not a speculative charge. Singapore’s competition watchdog fined the two companies a combined S$13 million (about US$9.5 million) over that merger. Taipei has effectively entered that history into evidence, and it said Uber would need to cut its Grab stake further, or exit altogether, to ease the risks.

The Delivery Hero twist

There is also a complication the four points do not spell out. In July, Uber agreed to buy Delivery Hero itself in a deal valuing the German group at about US$14.8 billion, on top of the roughly 25 per cent stake it already held. Taiwan was carved out of that transaction, but the TFTC had already asked Uber whether owning Delivery Hero could still give it influence over the foodpanda sale.

Also Read: Superbank under Grab: what the takeover means for Indonesia’s crowded digital banking scene

Strip it down and the picture is awkward. The seller is about to become Uber’s subsidiary. The buyer counts Uber as a major shareholder. The incumbent rival is Uber Eats. A regulator that had already said no to Uber once was never going to be relaxed about a deal in which Uber sat, in some form, on every side of the table.

Local opposition did not help. A lawmaker and delivery workers urged the commission to reject the deal outright, calling it a “de facto monopoly”. The National Delivery Union cited a government-funded Institute for Information Industry study that ranked Grab last among five potential buyers. Critics also flagged Grab’s use of Huawei mapping services, a concern outside the TFTC’s competition remit, but hardly a tailwind in Taipei.

What foodpanda Taiwan is worth, and to whom

The prize is substantial. Foodpanda has operated in Taiwan for about 12 years, with 3-4 million users and 100,000 merchants, and generated US$1.8 billion in gross merchandise value last year. The deal would have handed Grab just over 50 per cent of the market and a presence in 21 cities.

For Grab, the setback is strategic rather than financial. The company has been on a buying spree, from the US$425 million Stash acquisition to its move on Atome, and Taiwan was supposed to prove that the playbook honed across its eight Southeast Asian markets could travel. That proof must now wait. Investors who have questioned the quality of Grab’s profits may not mourn a US$600 million cheque that stays in the bank.

For Delivery Hero, the options are narrowing. An appeal is one route. A third buyer is another, but it would need to be a well-funded operator with no Uber ties. Meituan’s Keeta, which overtook foodpanda in Hong Kong within 29 months of launching, is the obvious candidate, although a Chinese-owned platform would face political headwinds of its own in Taiwan. Asked whether foodpanda might simply leave the island, a TFTC official called the question hypothetical.

The Southeast Asian mirror

Southeast Asian founders and investors should watch closely, because the Taiwan ruling previews questions closer to home. If Uber’s Delivery Hero deal closes, foodpanda’s operations in markets such as Singapore, Malaysia and the Philippines would fall under Uber’s roof. That would put Uber back in direct competition with the company it sold its regional business to in 2018, while remaining one of that company’s biggest shareholders. Regulators in the region will have to decide whether that counts as rivalry or a family dinner.

Also Read: GoTo’s first profit signals a fintech-driven future

And with Grab-GoTo merger talk resurfacing every few months, Taipei has shown that Asian competition authorities are now willing to look past market-share tables to the cap table.

Grab chose Taiwan to prove it could grow beyond Southeast Asia. Its biggest obstacle turned out to be the shareholder it inherited from its most famous deal.

The post Taiwan kills Grab’s US$600M foodpanda deal, and Uber’s shadow is the reason appeared first on e27.

I knew within the first two minutes of glancing through the deck.

A strategy workshop for our senior stakeholders had just wrapped. The session itself was good: the partner facilitating it followed our brief, the debate was sharp, and an AI tool captured the notes. When the deck arrived, I opened it and felt that familiar dip. Clean layouts. Confident headings. Icons beside every bullet. And almost nothing that sounded like the room we had just sat in.

I can’t prove how it was made, and I won’t pretend to. But the output looked like what happens when notes go in one end of a tool and a deck comes out the other, and nobody with judgement reads it as the audience would.

What the AI got right

Capture was mostly there based on what could be heard visibly. Almost every comment was transcribed, and the notes were more complete than anything a human note-taker would have produced. If you still think the risk with AI is that it can’t capture pronunciation properly, you are a year behind, especially if you are using a fit for purpose transcription tool.

The deck is where it fell apart, and the failure came in four forms.

• Contextual failure. Several inputs we had spent real time on never reached a slide though they were transcribed. The context was missing or misunderstood, which means the human-in-the-loop was clearing missing when reviewing the outputs.
• Misconstrued inputs. This one worries me most. One point had been softened into a generic statement, and another had its emphasis flipped. A missing idea is perhaps due to a missed captured but the classification of info under the wrong category is a human oversight failing to validate AI’s outputs.
• No narrative. The deck was a tidy inventory of topics. There was no argument, no sequence of “therefore,” and no answer to “so what?” A workshop deck exists to move a group from where they started to what they now believe. This one just reported that a conversation had occurred.
• Decorative design. The visuals were there to look polished, not to say anything. A chart with no point. A framework graphic that framed nothing. Design should carry positioning and tell the story, and here it sat on top of the content like wallpaper.

Also Read: Hong Kong builds a test track for robots that keep failing outside the lab

This is not a vendor story

It would be easy to write this as “my partner used AI badly.” But I use these tools constantly, and I have shipped drafts that weren’t ready, so the point is not that I’m exempt.

Research backs up what many of us have experienced. A Stanford and BetterUp Labs study, published in Harvard Business Review, gave this phenomenon a name: “workslop”, AI-generated output that looks like finished work but pushes the effort of fixing it onto the person who receives it.

That is exactly what happened here. The partner saved time, and my team inherited the cost. We spent hours checking the deck against our own notes, rewriting the narrative, and working out which statements were ours and which were the tool’s best guess of our intent.

What I do now

Two things changed in how I work, and in how I brief partners.

• First, I run a traceability check before anything AI-assisted leaves my hands. Every input from the session must map to a slide, or I must be able to say why it was dropped. It is tedious, and it catches what a skim never will.
• Second, I ask one question of every slide: what should the audience believe after seeing this? If I can’t answer in a sentence, the slide is decoration, and it goes or gets rebuilt.

Also Read: Employers are buying AI tools. Their managers are being left behind

With partners, the briefing conversation has shifted too. I now ask who the named reviewer is on any deliverable, and whether outputs will be checked against source inputs before they reach us. It is a slightly awkward question. It is far less awkward than the conversation after a bad deck.

The decision that stays human

Here is what I keep coming back to. AI can capture, summarise and format. It cannot decide what mattered most, or what the room really meant, or what story will land with this particular group of stakeholders on this particular day. That is judgement, and judgement needs a human owner.

The tool did not ship that deck. A person did. “AI slop” is rarely a tool failure. It is a sign-off failure: someone decided the output was good enough without reading it as the audience would.

So use the tools. Let them take the notes, and even draft the first version. But before you send anything, read it as the person who wasn’t in the room, and ask whether it would still make sense to them. If you wouldn’t put your name on it, don’t send it.

—

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 AI took perfect notes. Then it ruined the deck appeared first on e27.

Bitcoin rose 1.51 per cent to US$84,055.22 in 24 hours, and Ethereum rose 0.97 per cent to US$2,471.00. The total crypto market cap rose 1.36 per cent. A rebound in technology and AI stock sentiment lifted risk appetite across both assets. Bitcoin showed a 79 per cent correlation with the S&P 500, and its 24-hour correlation with the S&P 500 ETF SPY was 0.79. That shared direction matters. My view is that this rally is real but narrow and derivative-heavy. It supports a cautiously bullish stance, not a conviction breakout call.

The initial driver came from technology and AI sentiment. SpaceX revenue soared 161 per cent, and Super Micro Computer offered positive guidance. Those earnings signals revived broader risk appetite, and that optimism flowed into crypto. Bitcoin’s tight link to the S&P 500 confirms that its short-term direction remains tied to traditional market sentiment, especially around growth and technology narratives.

The next test comes from upcoming earnings at major technology firms such as Meta and Tesla. Those reports could sustain the current mood or reverse it. If you have been reading what I wrote recently, I said to treat technology earnings as a near-term macro input, not background noise.

Bitcoin also received a mechanical push from a short squeeze. The asset saw US$84.26 million in total liquidations over 24 hours, and shorts made up 92 per cent of that amount. Rising prices forced leveraged short sellers to buy back positions, and that forced buying created a feedback loop that accelerated the upward climb.

This matters because derivative market mechanics amplified the rally, not just organic spot demand. That distinction shapes my view. A squeeze can drive a fast climb, but it can also increase volatility and leave the market vulnerable once forced buying runs out.

Watch for sustained high funding rates or a sharp contraction in open interest, either of which could signal the squeeze is running out of fuel. This is more tactical rather than structural. A squeeze can pull price toward resistance, and it can also fade once the forced buying stops. I would give more weight to spot demand than to liquidation data when judging whether the rally can last.

Also Read: Bitcoin at US$82,829: Why a 95% correlation with the S&P 500 should worry long term believers

On price structure, Bitcoin is testing the upper boundary of a well-defined US$80,000 to US$87,000 range. The seven-day simple moving average sits near US$83,100. If Bitcoin holds above that level and breaks US$87,000, it could target US$88,000. A rejection could bring a retest of the US$80,000 support. The near-term triggers are clear.

A daily close above US$84,200 would confirm near-term bullish momentum, while a close below US$82,000 would signal bearish momentum. The key macro trigger is the Federal Reserve’s rate decision on October 28. Bitcoin is consolidating and awaiting a macro catalyst for its next decisive move. I would want a decisive close above US$87,000 after the Fed announcement before treating this as a breakout rather than a range trade.

That plan keeps risk defined because the range gives both sides a clear line in the sand. My humble advice is not to chase the climb into US$87,000 before the Fed meeting, and I would not treat a single close above US$84,200 as proof of a new trend. The central bank decision deserves more weight than one session of price action.

Ethereum has its own supply-side catalyst. On October 20 to 21, Tom Lee’s Bitmine staked an additional 159,728 ETH worth approximately US$390 million. That brought its total staked holdings to about 5.23 million ETH. This action locks a significant portion of supply, reduces potential sell pressure, and signals strong institutional belief in Ethereum as a yield-generating asset. Large long-term holders are actively choosing to earn yield rather than sell.

That provides underlying demand support. The watch point is continued growth in the staking queue, which would indicate sustained demand to lock supply. It gives Ethereum a firmer floor than pure sentiment alone would provide, though a supply lock does not guarantee an immediate breakout.

Ethereum also benefited from the broader market lift. The total crypto market cap rose 1.36 per cent, led by Bitcoin’s 1.51 per cent gain. A rebound in global technology and AI sentiment supported the move, with strong earnings forecasts from companies such as SpaceX helping the mood. Ethereum saw US$47.99 million in liquidations over 24 hours, with shorts comprising the majority.

That suggests a squeeze added momentum. Ethereum did not rise alone. It benefited from a risk-on shift in correlated assets. The key watch point is how US equity markets react to upcoming technology earnings from Meta and Tesla, as those reactions can provide continued correlation cues.

Also Read: Institutional buyers face a massive test at the US$81,000 Bitcoin support level

Ethereum now faces immediate resistance at the psychological US$2,500 level. A sustained break above that level, especially on a daily close, could open the path to the 30-day simple moving average at US$2,579. The scheduled activation of Ethereum’s Hoodi testnet upgrade on October 26 could renew developer and investor focus.

Failure to hold US$2,470 risks a pullback toward the recent low near US$2,380. A daily close above US$2,500 would confirm bullish momentum, while a drop below US$2,380 would signal renewed weakness. My view is that Ethereum needs spot buying volume to accompany a break above US$2,500. Without that volume, a rejection would suggest the rally came primarily from derivatives positioning and short covering.

My overall point of view is cautiously bullish on both assets, with a preference for patience over chasing. Bitcoin’s improved risk sentiment and a derivatives squeeze provided a clear catalyst, but the price stayed within a multi-week range and showed little breakout conviction. Ethereum’s substantial supply lock-up and favourable macro backdrop provide a solid floor, but it must break above US$2,500 to sustain momentum.

The key watch is whether Bitcoin can decisively close above US$87,000 after the Fed’s policy announcement and whether Ethereum can break US$2,500 with rising spot volume. If those conditions fail, both markets risk reverting to range midpoints or retesting support. The data supports a constructive stance, but not complacency. Macro sentiment and spot demand matter more than a short squeeze, and I’d track those most closely.

My stance leans toward selective exposure rather than aggressive accumulation. The data point to a constructive backdrop, but also to fragile momentum. A trader can respect the range, wait for confirmation, and let the Fed decision and spot volume decide the next leg.

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 92% of Bitcoin liquidations were shorts: What happens when the squeeze ends? appeared first on e27.

Until this summer, the morning numbers at LDC Import-Export Co., Ltd. were built by hand.

LDC is a Lao distribution and retail group. It started in Vientiane in 2011, distributing local and imported goods, and its retail arm now runs 39 convenience stores across seven provinces, including Jiffy shops on PTT fuel forecourts and its own Sky Mart chain. The ERP behind those stores keeps the business in 52 databases that the company could only read through manual spreadsheet exports. Every morning the business-analysis team spent hours assembling the Excel report packs from those exports. The cost of empty shelves across the chain had never been put in money, and a question like “which items sold at zero cost last week” had no one to ask.

I’m the CEO and co-founder of QUEST EDTECH, and LDC is one of its clients. I took the work on because LDC was losing money it couldn’t see, and its analysts spent the first hours of every morning assembling numbers instead of using them.

Since May, my company has rebuilt that layer with AI coding agents working under my direction. A chain now restores the ERP at 4:45am each morning, rebuilds a warehouse and emails the packs by 7:45am with nobody awake, and a check at one in the afternoon ties 160,000 cells back to the source system. The analysts got their one to two hours a morning back, several hundred hours a year.

The first replay of the year found about US$570,000 of sales lost to empty shelves, more than half of it while the item sat in LDC’s own distribution centre, plus a duplicate-sales event the ERP had hidden. The AI running cost LDC carries for all of it is about US$100 a month. Five department files that staff used to compile by hand now rebuild themselves every night.

Before this year, the business-analysis team assembled the morning packs by hand, with Power Query pulls, SUMIFS across every branch and copy-paste from the ERP. When I proposed live dashboards instead, the answer was that the analysts need yesterday’s number in the format they know, and that stays.

So the format was fixed before a line of code existed and the daily reports are the original manuscript. Agents rebuilt the packs as codes, by mid-July they had loaded the ERP estate into a warehouse of about 60 million rows, the nightly chain has run unattended since 24 August via automation.

QUEST EDTECH realised that three things had to be fixed before the output was usable, and none of them were codes. The first fix was definitions, because the “online sales” column in one pack had been carrying wholesale accounts and ran several times too high. The definition that fixed it came from the analysts’ own requirements document, because the ERP codes credit returns differently from cash ones and the second code had never been filtered. Both fixes came from asking the people who own the numbers, and that’s where I’d start next time.

The second fix was time, after a date computed on a server that runs on UTC produced a pack with the wrong day’s data. The rule now is to compute a Bangkok day on a box set to Bangkok time. The third fix was a rule for the agents themselves, using the semantic layer derived from the database as the fact, then write the code. Agents read the database’s own schema before they name a column.

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

When the AI was confidently wrong about LDC’s numbers

Once nobody was assembling the packs by hand, nobody was looking at each number on its way to LDC either. In August every target-achievement figure in one pack read exactly 100 percent, which should’ve been the first clue. LDC’s target workbook is a rolling file, and when a month closes, its target is overwritten with the actual result.

The agents couldn’t know that, so they loaded actuals as targets and the pack was checking itself. What caught it was a test that rebuilt the targets from the sales data and found the two were identical. That same month I gave the agents the rule I’d give any team verify the fact, then write the code.

What AI informs but can’t decide

Before I could put the cost of LDC’s empty shelves in money, someone had to decide what an empty shelf is. Four defensible definitions of out-of-stock on the same data gave 0.6, 10.1, 67.2 or 77.3 percent. The agent could compute all four in minutes, but it couldn’t tell me which one the business should manage. I settled it with the rule the buyers already work to, from the supply-chain team’s own workbook. The US$570,000 of lost sales is counted on that rule’s zero-stock test.

What reaches whom is a human call too, and the AI “chief strategy officer” I set up to write weekly briefs for LDC’s CEO showed why. Its first version flagged four branches as the week’s news, and all four were already red in the pack the analysts read every morning. I rebuilt it as a decision sheet with an owner, a date and the amount at stake on each line, and who gets it is my call.

Also Read: Tokens are the wrong yardstick for AI agents, says JFrog’s Sunny Rao

What stopped, and what it has been worth

The analysts stopped building the packs, and five department spreadsheets that staff uploaded by hand were retired once a system source replaced them. I stopped waking up for failures too. In September I told the agents that when something breaks between 4 am and 9 am I need them to fix it, not notify me, since silicon doesn’t need sleep, and a DevOps agent now runs four checks a day and re-runs known-safe steps itself.

The scale of what agents build is the part I’d have doubted a year ago: about 240 dashboard cards in one three-day stretch in September, at roughly 4.1 minutes of my time per card, and 766 saved questions and 198 dashboards on the BI site against the client’s requirement of 76 reports.

The honest way to count the money is in what the system has found, because realised savings are LDC’s to report. With AI, LDC has identified more than half was lost while the distribution centre had the item in stock. Annualised, that’s roughly US$770,000, which is why the next phase is a replenishment fix rather than another report.

Who does what and the outcome

On QUEST EDTECH’s side the agents work in lanes, with Claude Code building and operating. The bigger change is on LDC’s side, where the analysts got Claude Desktop skills that turn a question in Thai into SQL. As of October, the BA lead added her own card to the finance dashboard, written through that skill with AI in Thais and Laos, and the build was changed so her cards survive every rebuild. The people who used to assemble the numbers manually now ask questions to AI and it just builds reports, dashboard and identifies with simple languages instead of power query or even a line of SQL.

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 How a 39-store retail chain in Laos moved from Excel to AI agents appeared first on e27.

Southeast Asia has a talent for making climate pledges sound like wedding vows: grand, emotional and frequently broken. Net-zero deadlines stretch to 2050 and beyond, yet coal still dominates several national grids, petrol motorbikes still choke the cities and mangroves still vanish faster than ministers can name them. Into that gap has stepped a scrappy generation of founders who would rather build than promise.

We combed e27’s startup database for climate tech companies founded between 2015 and 2026. The result is 30 startups worth watching. Some are flush: VFlowTech raised US$20.5 million for its vanadium flow batteries, and MediSun Energy closed an US$8.75 million seed round at a US$44 million valuation. Others survive on grants, angels and sheer stubbornness. Between them, they are turning chicken manure into coal substitutes, used cooking oil into biodiesel and idle backup batteries into revenue.

Also Read: Passing 1.5°C global warming: Performative ESG is no longer acceptable

None of them will save the planet alone. Collectively, though, they are a better barometer of the region’s climate ambition than any COP communiqué.

Energy, solar and storage

The region’s biggest emitter is its power sector, so the most capital has flowed here, led by VFlowTech’s US$20.5 million round.

1. VFlowTech

Builds vanadium redox flow batteries for long-duration energy storage.

• Founded: 2018
• Founders: Avishek Kumar, Arjun Bhattarai
• Headquarters: Singapore
• Funding: US$3M pre-Series A (2021, led by Wavemaker); US$10M Series A (2023, led by Real Tech Holdings); US$20.5M (2025, led by Granite Asia)

2. GetSolar (formerly Solar AI Technologies)

Runs a rent-to-own rooftop solar programme for homes and SMEs.

• Founded: 2020
• Founders: Bolong Chew, Gérald Chablowski, Luke Ong
• Headquarters: Singapore
• Funding: S$450K pre-seed (2020, ENGIE); US$1.5M seed (2023, led by Earth Venture Capital)

3. Verta Bioenergy

Turns farm and livestock waste into biomass pellets that replace coal in industrial boilers.

Also Read: NEU Battery Materials bags US$7.6M to mine lithium from the world’s least-loved battery

• Founded: Not disclosed (incubated under EDB’s Corporate Venture Launchpad)
• Founders: Jeremy Tan (co-founder and CEO)
• Headquarters: Singapore; plant in the Philippines
• Funding: S$1.1M (US$845K) seed (2025; ENGIE Factory, Wright Partners, AlphaGen VC)

4. Amplicity Energy

Software that turns idle backup batteries at industrial sites into grid revenue.

• Founded: Not disclosed
• Founders: Not disclosed
• Headquarters: Singapore
• Funding: US$1M seed (2026, ENGIE-backed)

5. SolX Technologies

Provides digital energy services that promise electricity savings.

• Founded: 2022
• Founders: Sergius Santos, Matt Levin Tan, Bradley Cua
• Headquarters: Makati, Philippines
• Funding: US$780K seed (2024; UntroD Capital Asia, Leave A Nest, Real Tech Holdings, Foxmont)

6. Arysun Energy

Sells affordable solar starter kits to middle-class households.

Also Read: EVs to make up over half of Singapore vehicle sales in 2026, BMI says

• Founded: Not disclosed
• Founders: Ng Aik Phong (listed on e27 profile)
• Headquarters: Singapore; selling in Indonesia
• Funding: US$575K pre-seed (2025; First Move, Wavemaker Partners)

7. TailorTech

Power management system that cuts fuel use at off-grid generators.

• Founded: Not disclosed
• Founders: Brothers Ts. Dr Indera Shaiful Baharudin and Ts. Indera Shafiq
• Headquarters: Malaysia
• Funding: US$110K pre-seed (2024, Antler)

8. Altent Renewables

Converts wet waste into clean fuels.

• Founded: 2022
• Founders: Tan Wee Han (founder and CEO)
• Headquarters: Singapore
• Funding: Not disclosed

E-mobility and clean logistics

Two-wheelers and delivery fleets are where the region’s transport transition will be won or lost, and charging infrastructure remains the bottleneck.

9. Blitz Electric Mobility

Operates an EV last-mile delivery fleet with AI routing and driver financing.

Also Read: VinFast’s ride-hailing arm Green SM takes on Jakarta’s ojek economy with e-scooters

• Founded: 2019 (operating since 2022)
• Founders: Saivya Chauhan
• Headquarters: Jakarta, Indonesia
• Funding: US$2.6M seed, including extension (2024; Third Derivative, Iterative, ADB among backers); undisclosed pre-Series A led by Vynn Capital (2025)

10. ChargeSini

Builds and operates EV charging infrastructure.

• Founded: 2022
• Founders: James Goh and partners from the automotive and tech sectors
• Headquarters: Malaysia
• Funding: US$1.25M seed (2023, via pitchIN)

11. Aerion Motors

Building a premium e-motorcycle brand for Indonesia, with production planned for Q4 2026.

• Founded: Not disclosed
• Founders: Adrian Cheng (listed on e27 profile)
• Headquarters: Singapore
• Funding: US$150K angel (2025)

Carbon accounting, carbon markets and climate finance

Disclosure rules are tightening across ASEAN exchanges, which has turned emissions data from a CSR nicety into a compliance product.

12. Unravel Carbon

AI carbon accounting software for enterprise supply chains.

• Founded: 2021
• Founders: Grace Sai, Marc Allen
• Headquarters: Singapore
• Funding: US$7.38M seed (2022, led by Surge), taking total to US$8.75M; undisclosed round from Japan Energy Capital and TIS (2025)

13. CarbonEthics

Develops blue carbon, peatland and forest restoration projects.

• Founded: 2019
• Founders: Bimo Soewadji (Agung Bimo Listyanu), Jessica Novia
• Headquarters: Jakarta, Indonesia
• Funding: US$220K pre-seed (2023; Spiral Ventures, Ecoxyztem); US$2.1M seed (2024, led by Intudo Ventures)

Also Read: EVs gain ground in the Philippines as fuel costs squeeze drivers

14. Refy

Asset-based financing for small and mid-scale green projects.

• Founded: 2023
• Founders: Ed Chin
• Headquarters: Singapore
• Funding: US$525K pre-seed (2024, Wavemaker Impact)

15. CREX

Turns hotels’ sustainability data into AI-ready signals that drive bookings.

• Founded: Not disclosed (first round in 2022)
• Founders: Sam Hon (listed on e27 profile)
• Headquarters: Singapore
• Funding: US$750K angel (2022); US$300K seed (2024)

16. Carbonwize

Audit-grade emissions reporting linked to green financing.

• Founded: 2022
• Founders: Natalie Lerthatasilp (listed on e27 profile)
• Headquarters: Bangkok, Thailand
• Funding: US$265K pre-seed (2023); about US$1.73M raised in total, per PitchBook

17. Carbon GPT

AI-driven greenhouse gas accounting across all scopes.

• Founded: Not disclosed (on e27 since 2024)
• Founders: Matthew Wong (listed on e27 profile)
• Headquarters: Malaysia
• Funding: US$20K Cradle Fund grant (2023); US$120K angel (2024)

Circular economy and waste

Waste is the least glamorous corner of climate tech and, arguably, the one with the fastest payback: methane from landfills and open burning is an immediate problem.

18. Rekosistem

End-to-end waste collection, sorting and material recovery.

• Founded: 2018 (formally launched 2021)
• Founders: Ernest Layman, Joshua Valentino
• Headquarters: Jakarta, Indonesia
• Funding: US$5M (2023, led by Skystar Capital); US$7M Series A (2025; led by Saratoga Investama Sedaya and K3 Ventures)

19. Remind

Collects and processes e-waste and recovers metals.

• Founded: 2020
• Founders: Muhammad Dzikri Ahira Soefihara
• Headquarters: Tangerang, Indonesia
• Funding: US$1.3M (2025; co-led by Bali Investment Club and BEENEXT, with Spiral Ventures)

20. GEPP Sa-Ard

Digital waste-tracking and trading platform for businesses.

• Founded: 2017
• Founders: Dome Boonyanurak, Mayuree Aroonwaranon
• Headquarters: Bangkok, Thailand
• Funding: US$300K pre-seed (2020); backers include Delta Electronics Thailand, depa and AIS The StartUp

21. Arus Oil

Collects used cooking oil and supplies it to biodiesel refineries.

• Founded: 2017
• Founders: Syazwan Majid
• Headquarters: Subang Jaya, Malaysia
• Funding: Pre-seed, angel, seed and debt rounds totalling about US$192K (2021–2023); 1337 Ventures Accelerator Fund (2025)

Food, agriculture and biomaterials

Agriculture and land use drive a large share of the region’s emissions, and these startups attack it from the farm, the feed trough and the lab.

Also Read: Sustainability reporting is not the goal. Better business decisions are

22. Agros

Solar-powered irrigation and soil products for horticulture farmers.

• Founded: 2019
• Founders: Max Nelen
• Headquarters: Singapore; operations in Indonesia
• Funding: US$2.7M pre-Series A (2023); US$4.25M Series A (2025; co-led by Gaia Impact, Schneider Electric Energy Access Asia, Wavemaker Impact)

23. Terra Oleo

Uses microbes to make palm oil and cocoa butter alternatives from agro-industrial waste.

• Founded: 2023
• Founders: Shen Ming Lee, Boon Uranukul, Min Hao Wong
• Headquarters: Singapore
• Funding: US$3.1M (2025; ADB Ventures, Better Bite Ventures, Elev8.vc, The Radical Fund)

24. AgriG8

Credit platform that links rice-farmer loans to sustainable practices.

• Founded: 2021
• Founders: Aries Dwiputera, David Chen
• Headquarters: Singapore
• Funding: US$900K seed (2024)

25. Virdalis

Duckweed-based protein for animal feed, as an alternative to soy.

• Founded: Not disclosed
• Founders: JM Aujero (listed on e27 profile)
• Headquarters: Singapore
• Funding: US$700K pre-seed (2026)

26. Cleantech and Beyond

Advanced functional materials for sustainability uses.

Also Read: N&E Innovations nets US$1.7M to turn cashew husk waste into fresh produce protection

• Founded: Not disclosed
• Founders: Not disclosed
• Headquarters: Thailand
• Funding: US$100K pre-seed (2024); US$50K seed (2025)

Water, buildings and efficiency

Cooling, desalination and industrial water treatment are energy hogs in the tropics, which makes efficiency a climate play in its own right.

27. MediSun Energy

Brine management and osmotic energy technology for desalination and heavy industry.

• Founded: Not disclosed
• Founders: Joseph Chua, Dusun Kim
• Headquarters: Singapore
• Funding: US$8.75M seed (2024, led by Vynn Capital, at a US$44M valuation); further seed funding of about US$3M reported (2026)

28. Aprisium

Real-time contaminant detection, including PFAS, for water and industrial fluids.

• Founded: 2021
• Founders: Raghav Narayan, Dr Lars-Henrik Skjolding
• Headquarters: Singapore
• Funding: US$800K seed (2023, Cocoon Capital); US$7M Series A (2024; led by Deep-Tech Venture Partners and Cocoon Capital)

29. BIZSU

Passive add-on that cuts air-conditioning energy use.

• Founded: 2019
• Founders: Brice Degeyter
• Headquarters: Singapore
• Funding: About US$250K angel (2021); US$100K angel (2025)

30. intlife

IoT and AI energy efficiency for homes and offices.

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

• Founded: Not disclosed
• Founders: Kok Huang Er
• Headquarters: Malaysia
• Funding: US$25K Cradle Fund grant (2023)

What the list tells us

Read the headquarters lines and a lopsided map emerges. Seventeen of the 30 are based in Singapore, against five in Malaysia, four in Indonesia, three in Thailand and one in the Philippines. Vietnam, the region’s fastest-rising factory floor, does not appear at all. Singapore’s grants, incubators and corporate venture arms clearly work, but the city-state is not where most of the region’s emissions come from.

The cheques are also small. About half of these companies have yet to disclose a single round larger than US$1 million. A battery-swapping network or a desalination plant cannot be built on angel money. Until growth-stage investors back climate hardware with the conviction they once reserved for super apps, many of these founders will stay stuck proving the same pilot over and over.

The post Thirty climate tech startups in Southeast Asia that would rather build than pledge appeared first on e27.

A year ago, some of the most nervous entrepreneurs in the world were in Hong Kong, Taipei and Shanghai. Trade tensions, a property slump in mainland China and a stalled listings market had put founders there among the most cautious respondents in HSBC’s annual survey of business owners. Twelve months on, they are the most cheerful people in the room.

The reason has little to do with apps and almost everything to do with the machines the apps run on.

Also Read: Gen Z is bringing its own AI to work. Singapore’s SMEs need rules, not bans

HSBC’s Global Entrepreneurial Wealth Report 2026, which polled more than 3,000 founders across 17 markets, finds that 54 per cent of Hong Kong entrepreneurs now feel “very positive” about their business prospects, up 20 percentage points year-on-year. Optimism about personal wealth rose 22 points to 53 per cent. In Taiwan, the share of “very positive” founders climbed 15 points to 43 per cent.

Follow the shipments and the mood swing makes sense. Mainland China (18 per cent), Taiwan (14 per cent) and Hong Kong (12 per cent) together accounted for nearly half of all global AI-related exports by value in the first quarter of 2026. The report says Taiwan’s AI hardware exports are now equivalent to almost half the island’s GDP, a figure that says as much about concentration risk as it does about success.

Hong Kong, meanwhile, logged its busiest first half for IPOs in five years, with AI listings leading the queue. It is the same report that told us AI will create jobs, provided you read the fine print.

For Southeast Asia, the uncomfortable takeaway is that the region barely features in this story. Its startup ecosystem spent a decade perfecting super-apps, ride-hailing and digital marketplaces: software businesses that monetise users rather than ship silicon. The money now flowing through the AI supply chain is largely going elsewhere.

Also Read: Gen Z doesn’t need more AI courses, it needs the skills AI can’t replicate

There are efforts to change that. Thailand has approved a semiconductor strategy that targets roughly US$80 billion in cumulative investment by 2050, while Singapore has given its chip sector a national identity in SG Semiconductor. But national roadmaps run on decades, and founders run on 18-month runways.

The generation that treats AI as plumbing

The second half of the report is about who is building companies, not where. Gen Z founders lead every age group on AI adoption: 65 per cent say they have fully embedded AI into their operations, compared with 57 per cent of Gen X and older founders. In mainland China, the Gen Z figure reaches 91 per cent.

Desmond Kuang, Chief Investment Officer for Asia at HSBC Private Bank, describes the gap as a matter of temperament. Young founders, he says, are “more comfortable experimenting, iterating and tolerating model risk”, while older peers “often require clear ROI, stronger governance”.

In practice, that means AI agents handling scheduling, research, customer support triage and code refactoring from day one. Some young founders in financial services say they have redesigned entry-level hiring altogether, handing routine junior analyst work to software and hiring humans only for senior roles.

That looks efficient on a spreadsheet. It also raises an awkward question for an ecosystem that already struggles to grow mid-level talent: if nobody hires juniors, where do tomorrow’s seniors come from? Singapore’s job market is already treating agentic AI as the hottest skill on the CV, and the people being priced out are rarely the ones writing the job ads.

Worried, but only a little

The report is keen to show that Gen Z founders are not reckless. They are twice as likely as older founders to worry that AI will erode critical thinking in their teams. Female founders show a similar gap over male peers, at 8 per cent versus 4 per cent.

The framing hides the base, though. “Twice as concerned” means 10 per cent versus 5 per cent. Roughly nine in ten of the most AI-fluent founders surveyed did not flag cognitive decline as a concern. That is less a generation guarding against automation bias than a thoughtful minority within one.

Also Read: The AI agent boom is exposing Southeast Asia’s startup codebase problem

Data privacy is a different matter. It is Gen Z’s top adoption concern, cited by 36 per cent, while every other age group puts implementation cost first. That matches what engineering leaders across the region describe: teams are happy to let AI agents draft and test, but not touch production. The threat surface is widening too, with cheap AI-driven attacks now within reach of the region’s small online merchants.

The barbell, and the gold

The most revealing data may sit outside the business. Thirty-eight per cent of Gen Z founders name inflation as their biggest personal wealth worry. Their response is what the report calls a barbell strategy: high-risk bets on their companies paired with defensive personal holdings. Sixty-seven per cent of Gen Z business owners plan to buy more physical gold over the next year, and 60 per cent intend to increase direct investments in private companies, compared with 44 per cent of Gen X.

That last number deserves attention from Southeast Asian VCs and family offices. A cohort of young, AI-native founders who want to write their own cheques into private companies is a potential co-investor, not just deal flow.

What the region should actually take from this

The report’s implicit prescription, plugging into North Asia’s hardware engine, is easier written than done. Partnering with Taiwanese foundries or Shenzhen robotics makers takes purchasing power, technical credibility and, increasingly, geopolitical nerve, as US-China export controls redraw who can sell what to whom.

A more useful reading is narrower. Southeast Asian software startups that want to matter in this cycle should build for the hardware that is arriving, such as edge devices, robotics, EVs and factory automation, rather than another chatbot wrapper. Taiwan, for its part, is trying to move in the opposite direction, betting on Gen Z founders to outgrow its chip-supplier image. The two ecosystems have more to trade than either tends to admit.

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

The Gen Z lesson, meanwhile, is not simply “adopt more AI”. It is to adopt it with the privacy discipline young founders say they want, and with the human judgement that only a minority of them are actively protecting.

North Asia’s founders are optimistic because they sell the picks and shovels. Southeast Asia’s will need a sharper plan than renting them.

The post Hong Kong and Taiwan founders are cheering the AI boom. Southeast Asia isn’t in the room appeared first on e27.

HSBC’s latest survey of wealthy entrepreneurs contains an awkward finding for the men who still write most of the cheques. Their own confidence is slipping, while the confidence of the women they rarely fund is rising.

The Global Entrepreneurial Wealth Report 2026 polled more than 3,000 business owners across 17 markets. Over the past year, the share of male entrepreneurs who call themselves “very positive” about their business prospects fell nine percentage points to 47 per cent. Among women, it rose three points to 54 per cent.

The gap widens on money. Forty-nine per cent of female business owners expect their personal wealth to grow substantially, up a point. Male expectations dropped 10 points to 42 per cent.

Also Read: HSBC says AI will create jobs. Southeast Asia should read the fine print first

Before anyone puts that on a pitch-deck slide, there is a caveat. Every respondent holds at least US$2 million in investable assets. These are established and often large businesses, not a pre-seed founder in Manila juggling payroll.

The report also offers no Southeast Asia-specific gender split, so any regional reading is an extrapolation. What the numbers can show is how women who have already made it think about the next wave of technology. That is where it gets interesting.

Self-belief, with receipts

Asked why they feel good about their prospects, women cite self-belief more often than men do (31 per cent versus 26 per cent). Thirty-nine per cent say they have what it takes to ride out economic uncertainty, against 33 per cent of men. The report points to long-run economic research suggesting women earn disproportionately more running their own firms than in salaried jobs, which may explain why the conviction holds.

That conviction contrasts sharply with the capital that reaches them. Female-only founding teams secured just 2.3 per cent of global VC funding in 2024. In Southeast Asia, only 16.5 per cent of venture investment went to female entrepreneurs as recently as 2020. Confidence, evidently, is not the bottleneck.

Trust in AI, pointed at the front office

AI is the main engine of this optimism. Fifty-eight per cent of female entrepreneurs name AI and broader technology advances as their top reason for feeling positive. Fifty-one per cent place “a great deal” of trust in AI to support their business, seven points ahead of men (44 per cent). The pattern carries into personal finance: 86 per cent of women trust AI tools for portfolio analysis and wealth planning, against 83 per cent of men.

That trust stands out in a region where tech leaders are learning to trust AI agents, but not with production. More telling is what the women plan to do with it.

Both sexes rank employee productivity as AI’s main job. Men, however, lean harder on AI as a cost lever. Forty per cent want fatter profit margins (37 per cent of women), and 27 per cent want a leaner workforce (24 per cent of women). Women tilt towards revenue. Thirteen per cent use AI to improve customer service and experience, versus 8 per cent of men. Ten per cent use it to mine consumer insights, versus 7 per cent.

Also Read: Gender gap in GenAI skills is narrowing, but progress remains uneven, Coursera finds

The usual health warning applies. These are single-digit shares, and a five-point gap on a base of 13 per cent is a nudge, not a revolution. Still, the direction matters in markets where winning customers is expensive and loyalty is fickle. An AI deployment that keeps a customer is worth more than one that trims a salary line, provided it actually works. That proviso is why SEA’s retailers are being urged to protect AI-powered customer experience with proper monitoring.

The scepticism that cuts both ways

Higher trust does not mean blind faith. Women are twice as likely as men to worry that AI will erode critical thinking and decision-making, although the figures are 8 per cent versus 4 per cent. They are also more anxious about falling behind rivals on AI adoption (13 per cent versus 10 per cent).

Both worries are rational. The first echoes a debate already running through the region’s talent pipeline. The junior developer dilemma has employers asking who will audit machine output once entry-level grunt work disappears. The second is a reminder that ambition alone does not ship products. Singapore SMEs are discovering that messy data keeps AI pilots stuck.

What the survey cannot tell us is why male sentiment fell so sharply. The cause may lie in the sectors men dominate, the size of their firms, or their exposure to a turbulent geopolitical year. The report presents the divergence without explaining it, which leaves a gap worth closing in next year’s edition.

What investors should take from it

Strip away the cheerleading, and three practical points survive.

First, women founders appear more inclined to deploy AI where it touches customers. Investors weighing AI-era startups should therefore look harder at retention and lifetime value, not just headcount savings. That lens happens to favour the founders the industry has historically overlooked.

Second, guardrails are not a sign of timidity. Founders who worry about AI degrading judgement are more likely to set policies before a customer finds the problem. The same logic explains why Singapore’s SMEs need rules, not bans, as staff bring their own AI tools to work.

Also Read: AI could redefine women in the workplace, and companies must act now

Third, the confidence data undercuts one of the industry’s laziest explanations for the funding gap: that women simply don’t pitch boldly enough. On this evidence, at least among owners who have already built wealth, women are the more optimistic half of the room.

Whether that optimism turns into term sheets is a question for the people writing them.

The post Male founders are losing confidence. Women are doubling down on AI appeared first on e27.

Before I used Mailchimp, my monthly newsletter was one of those tasks I kept postponing. Not because it was complicated. Because it was annoying.

I was running a small business, which meant the newsletter was always competing with more urgent things. Customer messages, fulfilment, invoices, product updates, random admin. By the time I got to the email, I was usually rushing.

The old process was clumsy. I would copy email addresses into BCC, write the message in a normal email editor, hope the formatting looked decent enough, and send it without really knowing what would happen after that. I did not know who opened it. I did not know which links were clicked. I did not have a proper unsubscribe flow. And every month, I had to repeat the same messy setup from scratch.

At the time, I thought the problem was just that I needed a better email tool. That was only part of it. What actually changed things was moving from a one-off sending habit to a more structured email workflow.

Mailchimp happened to be the tool I started with because the free plan was enough for my list size. But the real shift was not “I switched software.” The shift was that the newsletter stopped being a manual blast and started becoming a repeatable process.

Also Read: Hong Kong builds a test track for robots that keep failing outside the lab

The first obvious improvement was formatting. My emails stopped looking like something thrown together at the last minute. Templates helped. Headers helped. Buttons helped. Even basic layout made the newsletter feel more deliberate.

The second improvement was list management. I no longer had to keep a messy list somewhere and wonder whether it was current. Subscriptions and unsubscribes were handled properly. New sign-ups could be kept in one place. That alone removed a surprising amount of friction.

The third improvement was visibility. For the first time, I could see whether anyone was actually engaging with what I sent. Not in a deep analytics sense. Just enough to stop guessing. Which subject line got opened more? Which link got clicked? Which section of the newsletter seemed to get attention?

That was the point where the workflow changed. Before, sending the newsletter felt like completing a chore. After, it started feeling like a small loop I could improve.

I could look at what got opened, what got ignored, and what people clicked. Then I could decide what to keep, what to shorten, and what to change in the next issue. It was still simple, but it was no longer blind.

That was also where ChatGPT became useful, though not in the way people usually pitch it. The useful part was not asking it to write the entire newsletter from nothing. That usually gave me something too polished, too generic, or too eager. What worked better was using it inside the workflow.

I would gather the rough pieces first. A product update, one customer question that kept coming up, a useful article, maybe one thing I wanted to promote. Then I would use ChatGPT to help turn that pile into a cleaner draft by reading that read for context before giving it all the necessary files via the Dropbox connection. It could suggest a tighter subject line, rewrite a clumsy paragraph, shorten a section that dragged, or give me two versions of the intro depending on whether I wanted the tone to feel more direct or more relaxed.

That made the task lighter, but it still stayed mine. I was not handing over the newsletter voice to a machine. I was using it more like an extra pair of hands inside a workflow that already made sense. More like a sounding board or someone who can arrange some things before I take over.

Also Read: Employers are buying AI tools. Their managers are being left behind

The lesson for me was simple. The hardest part of a small-business newsletter is often not writing one email. It is building a repeatable way to gather updates, draft them, send them properly, and learn something from the result.

Once that workflow existed, ChatGPT became more useful too. It had something to work with. It could help refine a draft, not pretend to replace the whole process. Unfortunately, ChatGPT does not have a Mailchimp app yet. Else I could do this end to end and have more of the workflow handled by AI. Automaid could use my browser but seeing it trial and error from clicking on the wrong buttons while my token usage climbed was too much for my liking. (Perhaps next what I would do is create and submit a community card that is much more descriptive so that the agent knows exactly which button goes where!)

I would not really describe this as a Mailchimp review, even though Mailchimp was the trigger for the change. The real improvement came from structure.

Before, the job was: remember to send something. After, the job became: collect the updates, draft the email, send it through a proper system, check what happened, and improve the next one.

That is a small operational change, but for a small business, those are often the ones that take up time and are easy to neglect.

I still do not treat newsletters like a grand strategy exercise. But I no longer treat them like an annoying monthly scramble either.

That changed more than the template did.

—

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 changed when I stopped sending newsletters manually appeared first on e27.

In Singapore, a car is less a vehicle than a ten-year licence wrapped in steel. Carro now wants institutional money to treat that licence as collateral.

The Singapore-headquartered automotive marketplace has set up a S$450 million (about US$350 million) asset-backed securitisation facility backed by Singapore car loan receivables. In a statement on Thursday, Carro called it a key step in diversifying its long-term funding sources and said it is the first securitisation of Singapore automotive loan receivables.

Also Read: Carro acquires CarPlace: what it means for SEA, Australia markets

Securitisation sounds exotic, but the mechanics are simple. A lender pools loans it has already made and sells them to a separate entity, which raises money from investors against the borrowers’ future repayments. The lender gets cash today to write more loans, and investors get a stream of instalments.

In Carro’s case, that entity is a bankruptcy-remote special purpose vehicle (SPV), a legally separate company built so the assets stay protected even if the parent runs into trouble. It buys and securitises auto loan receivables originated by Genie Financial Services, Carro’s wholly owned lending arm. The structure is approved under the Monetary Authority of Singapore’s (MAS) SPV regime.

Why the COE matters

Carro’s pitch rests on a quirk of Singapore’s car market. The company says the deal creates a new asset class for hire-purchase securitisation because the loans are secured by vehicles, and the Certificate of Entitlement (COE) system, which licenses vehicle ownership, can give that collateral a floor value and regulatory certainty.

The logic holds up to a point. A car in Singapore cannot be driven without a valid COE, and the certificate retains residual value over its lifespan, which makes repossessed cars easier to price than in markets where used-car valuations are a guessing game.

But a floor is not a fixed number. COE premiums swing sharply with quota supply and demand, and a certificate’s value erodes as its ten-year life runs down. Investors will be underwriting a cushion, not a guarantee. A more reliable buffer may come from MAS’s own motor-financing caps, which limit how much of a car’s price buyers can borrow and force them to put real equity into every purchase.

The plumbing behind the pitch

The facility can run as a multi-funder structure, letting lenders participate through capital markets notes or direct loans on equal footing within the same vehicle. That flexibility matters: banks wary of a new asset class can lend directly, while bond investors can buy notes.

Carro also says the transaction includes fraud-risk safeguards and asset-protection measures. Independent third-party calculation agents will prepare reports, while automated data controls and a planned API integration will allow real-time performance reporting and monitoring of transaction triggers, the pre-set thresholds, such as rising arrears, that can change how cash is distributed or halt new loan purchases.

Also Read: Carsome posts tenth profitable quarter as SEA’s used-car race matures

Those safeguards are not decoration. The 2025 collapse of US subprime auto lender Tricolor, amid allegations that collateral had been pledged to more than one lender, reminded credit markets how badly things go when nobody verifies what sits inside the pool.

Aaron Tan, Carro’s co-founder and CEO, said the facility shows the strength of the company’s AI-enabled platform, data infrastructure and underwriting, and gives it the financial flexibility to expand across the region.

What the announcement does not say is equally telling. Carro has not named its funders, disclosed pricing, said whether the notes carry a credit rating, or shared delinquency data on the Genie book. S$450 million is a ceiling, not cash in the bank, and the market will judge the deal by how much is drawn and at what cost.

A lender wearing a marketplace’s clothes

The deal underscores what Southeast Asia’s used-car platforms have quietly become: lenders with showrooms attached. Ernest Chew, Carro’s CFO, said the company plans to diversify funding through bank loans, block discounting (selling hire-purchase receivables to a financier at a discount for upfront cash) and securitisation.

Its closest rival, Malaysia’s Carsome, has made the same shift, attributing its record quarterly EBITDA to a larger share of retail and financing services. In Indonesia, Moladin has targeted the used-car and auto-financing chain, while in the Philippines, Salmon recently bet US$140 million on secured lending that includes motorbikes.

Globally, the playbook is well worn. In the US, Carvana and CarMax regularly package auto loans into asset-backed securities, and both have shown how exposed the model becomes when credit cycles turn or used-car prices fall.

For Carro, the appeal is cheaper and more stable money. Lending businesses live or die on funding costs, and a platform reliant on a handful of bank lines is vulnerable when those banks tighten. The move also fits a wider regional appetite for structured debt, from private credit funds raising fresh capital to government-backed debt pools for tech firms. For readers new to the asset class, private credit has been steadily filling gaps banks leave behind.

The IPO subtext

There is also the listing question. Carro has said it is IPO-ready but would list only when conditions are favourable. A securitisation programme that survives independent scrutiny gives public-market investors something they rarely get from platform companies: third-party evidence of loan-book quality.

Also Read: From cash loans to motorbikes: Salmon bets US$140M on secured lending in the Philippines

Carro transacted more than 120,000 new and used vehicles in its 2026 financial year and operates in eight markets, including Australia following its acquisition of CarPlace.

The harder test lies beyond Singapore. The COE gives Carro an unusually clean collateral story at home. In Indonesia, Thailand or Malaysia, there is no such licence propping up residual values, and repossession is messier. If Carro wants this facility to fund regional expansion, it will eventually have to prove its underwriting works where the cars are cheaper, the borrowers riskier, and the floor much lower.

The post Carro turns Singapore car loans into a US$350M funding pipeline in a regional first appeared first on e27.

A few months ago, I was trying to understand why one high-protein formulation puffed well during low-moisture extrusion while another did not.

I gave ChatGPT the formulation details, expansion results, rheological measurements and observations from the extrusion trials. Then I started asking what might explain the difference.

One line of reasoning concerned critical strain — broadly, how much deformation a material could tolerate before its structure began to break down. A plausible hypothesis was that greater deformation tolerance might help a material expand.

There was a problem. One of the materials with the highest critical strain had performed poorly in expansion.

So I pushed back: if that explanation was right, why did the experiment show the opposite?

We reconsidered the interpretation, brought in other observations and explored alternative explanations. Eventually, the important lesson was not that one rheological measurement could predict puffing. It was that several properties and processing events had to be considered together.

That exchange made me realise that I was no longer simply using AI to retrieve information.

I was using it as a sparring partner.

Not just returning the ball

I used to play table tennis, and a good sparring partner does not simply return the ball comfortably.

One shot goes to your forehand, another to your backhand. Sometimes the ball goes to a corner you would rather not cover. That is how you discover the weaker parts of your game.

That increasingly describes how I work with AI.

I can put forward an explanation. ChatGPT offers another possibility. I question it. It challenges an assumption. I add another experimental observation. Sometimes the original explanation survives; sometimes it does not.

And sometimes the correct conclusion is simply that we do not yet have enough evidence to decide.

The value is not that AI gives me the answer.

It is that the exchange helps me reach better questions.

There is an understandable concern that easy access to AI-generated answers may make us think less. My experience has often been the opposite. Used this way, AI makes me interrogate my own thinking more.

Also Read: Hong Kong builds a test track for robots that keep failing outside the lab

Getting up to speed faster

This has become particularly useful in my advisory work.

I sometimes need to evaluate research and innovation projects spanning disciplines outside my own technical specialisation. Before I can assess the scientific basis of a proposal, what may be genuinely novel, or what questions should be put to the researchers, I first need to understand enough of the subject.

Previously, entering an unfamiliar technical area could be slow. I would search for background material, begin reading, encounter terminology or mechanisms I did not understand, search again and gradually assemble a mental map of the field.

AI has shortened that first stage considerably.

I can begin with elementary questions. If an explanation introduces another unfamiliar concept, I can ask immediately. I can compare mechanisms, ask why one interpretation is preferred over another, and keep probing until I know what I should be reading and questioning.

It does not evaluate the project for me. I still need to examine the original material and evidence and make the judgement myself.

But it gets me to the point where I can ask useful questions much faster.

For me, AI has reduced the cost of being curious.

When collaboration becomes trust

AI also saves time in more obvious ways. I use it to organise long documents, compare versions, restructure technical material and work through scientific manuscripts.

But efficiency creates another issue: trust.

Recently, I worked through many versions of a long document with ChatGPT. This was not a case of handing over the task and accepting whatever came back. I questioned arguments, requested specific changes and repeatedly refined the direction.

After each round, ChatGPT summarised what it had changed.

Gradually, I realised that I had begun trusting those summaries. If it told me that a particular point had been corrected, I did not always reopen the entire document and verify every amendment before proceeding to the next issue.

The collaboration had become useful enough that I trusted the account of what had been done.

Also Read: Before you use predictive AI, decide what decision it should change

I became conscious of the risk when I later reread part of a scientific manuscript. The writing was coherent, but an important part of my scientific intention was not sufficiently prominent.

Nothing was obviously wrong. But it was not quite telling the scientific story I intended.

That is a boundary I now pay much more attention to.

AI can help me construct an argument. It can challenge it, revise it and tell me what it has changed.

But if my name is attached to the work, I still have to ask: Is this accurate? Is this what I mean? Am I prepared to stand behind it?

AI cannot make that decision for me.

Covering more of the table

I therefore find myself asking a different question from when I first started using AI.

Not simply: What work can AI do for me?

But: What kind of thinking does working with AI allow me to do better?

Sometimes it helps me enter an unfamiliar field more quickly. Sometimes it helps me manage an overwhelming amount of information. And sometimes it sends a question back across the table that exposes an assumption I had not realised I was making.

That is why I keep returning to the table-tennis analogy.

A good sparring partner does not play the match for you. It does not remove the need for technique, practice or judgement.

It makes you cover more of the table.

The best use I have found for AI is not outsourcing my judgement. It is giving my judgement a better sparring partner.

—

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 How AI became my sparring partner appeared first on e27.