Over the past six months I've met a lot of AI companies based in Singapore. Most of them, frankly, have nothing to do with Singapore — the team is Chinese, the customers are overseas, they're incorporated here because the structure is convenient. Swap Singapore for Dubai or London, and the company logic wouldn't change at all.

I recently caught up with Ling, the founder of ShipLinker, on their latest progress. They build an AI Agent for maritime chartering. And I think this might be the most "Singaporean" AI company I've encountered.

Why? Because the industry they've cut into, their customer acquisition context, and the industry density they require could almost only have sprouted in Singapore. To understand this, you need to understand what chartering in maritime actually involves.

Chartering: the core transaction of the maritime world

The central activity in maritime trade is chartering — renting ships. Simply put, it's a lease contract between cargo owners and shipowners to move goods from A to B.

But "renting a ship" is far more complex than it sounds. Based on how much control the charterer has over the vessel, the industry has three models:

Voyage Charter — priced per voyage. The shipowner handles delivery to destination, bearing all operating costs including fuel and port fees. The easiest to understand — think of it as a maritime Lalamove. You only worry about the cargo; fuel and driver wages are the owner's problem.

Time Charter — rented by duration, six months or a year. The shipowner provides the vessel and crew, but the charterer bears variable costs during the voyage — fuel, port fees, agency fees. Like long-term renting a car with a driver. You decide where to go; you pay for gas.

Bareboat Charter — you rent just the hull. Crew, fuel, maintenance, insurance — all on you. Like renting an empty car and driving it yourself.

The risk allocation across these three models is completely different. A charterparty contract involves laytime, demurrage, despatch, time charter equivalent (TCE) — every clause represents a negotiation between risk and profit.

And the people who broker all of this are a highly specialized intermediary group: chartering brokers.

80% of transactions still happen over the phone

The most counterintuitive thing about this industry: it's 2026, and over 80% of global maritime transactions are still completed through offline brokers.

A broker's daily routine looks like this: receive a cargo requirement, make phone calls, send emails, use experience and connections to find suitable ships. They need to simultaneously evaluate a vessel's real-time position, speed, estimated arrival time, port congestion, route fuel consumption, emissions standards, and whether contract terms match — an extremely complex combination of variables, all changing in real time.

A single broker's information bandwidth is limited. They might know certain routes and certain shipowners well, but they can't possibly review hundreds of thousands of ships globally.

Commission? 3-4%. On a voyage worth hundreds of thousands to millions of dollars, 3% is serious money.

But this industry can't be understood simply as "inefficient matching." I visited ShipLinker's founder Ling alongside a friend who had invested in Manbang (China's trucking platform). Our biggest takeaway: maritime and road freight are structurally completely different. You can't apply Manbang's logic here.

Manbang faces long-tail drivers and long-tail cargo owners — fragmented, high-frequency transactions where a platform doing information matching can work. Maritime participants are highly concentrated, relationship networks are stable, individual transaction amounts are enormous, and decision chains are long. You can't build a platform where a cargo owner clicks to place an order — this isn't ride-hailing.

So maritime's pain point isn't "can't find the counterparty." It's "found them, but can't crunch all the numbers." Too much information, too many variables, not enough human bandwidth. This is precisely the most natural application scenario for an AI Agent.

Data can't fix deals. Agents can.

What ShipLinker does, in one sentence: AI Broker for maritime.

It doesn't replace brokers' relationships and negotiations. It automates the repetitive but critical work — screening vessels, calculating costs, predicting arrival times, assessing route risk. You give it a requirement profile, it runs global vessel data, and tells you: these deals are recommendable, do you want to pursue them?

Humans make the call. The Agent does the math.

Maritime already has several data companies — industry leaders that package global vessel, route, and port data for sale. But Ling said something that stuck with me:

"Data can't help you fix a deal. An Agent can."

Having data is useless on its own. Data is static, passive — you still need to filter, calculate, and judge. An Agent is different — it combines your requirements with real-time data and gives you an actionable recommendation. The gap between information and decision — the Agent walks that road for you.

This is why ShipLinker doesn't call itself a "data platform" or "SaaS tool" — it calls itself an AI Broker. What it's doing is automating the most time-consuming part of a broker's work — information collection and option screening — while preserving the industry's core assets: relationships and judgment.

A concrete example: when Middle East tensions escalated, the traditional offline broker network suddenly broke down — routes changed, risks changed, many established relationships stopped working. ShipLinker suddenly gained a wave of "unusual clients" reaching out proactively, because they couldn't get deals done through traditional channels, but the Agent could access global data and surface options that traditional networks couldn't reach.

$1,000 to start, then camels pull camels

The customer acquisition strategy is worth unpacking.

First tier: $1,000/month, providing vessel data and basic matching. Low barrier, easy for clients to start. Once they use it and find the Agent's recommendations genuinely useful, upgrade demand emerges naturally.

ShipLinker then launched a $3,000/month tier with full vessel history, port call records, and buying/selling information — not just helping with chartering, but with vessel purchase and sale decisions.

The logic is clean: brokers do both chartering and vessel trading, just like real estate agents do both rentals and sales. One group of people, two revenue streams. Once you pull them in, their needs naturally expand.

Currently about 20% of clients are actively upgrading from $1,000 to $3,000. MRR is approximately $26,000, growing 5x since launch last May, covering clients in 16 countries, tracking 200,000+ vessels globally — roughly 90% coverage. Cash flow has turned positive.

B2B acquisition works through what Ling calls "camels pulling camels" — one cargo owner uses it and likes it, pulls their partner shipowner onto the platform. One broker has a cargo they can't handle, pulls in another broker to co-arrange. The maritime industry is inherently a highly networked business where upstream and downstream naturally pull each other in. Once embedded in daily workflow, switching costs are high.

Industry veterans wire money in two weeks. VCs need a half-hour tutorial.

The fundraising process reveals something telling.

ShipLinker's fastest angel investors: a former trader at America's largest dry bulk trading firm — closed in two weeks. The former CFO of Danone Southeast Asia — his own company faced the same logistics decision pain points, understood immediately. A British industry veteran with 40+ years in global supply chain logistics.

What they have in common: you don't need to explain what dry bulk is, what chartering brokers do, or what TCE means. Understanding leads to speed.

With VCs, Ling says the industry tutorial alone takes half an hour. Afterward, they "roughly understand what you do."

This isn't unique to ShipLinker. Vertical industry AI startups inherently face a fundraising paradox: the deeper your product goes into the industry, the less VCs understand; but precisely because they don't understand, the moat is real.

The most "Singaporean" — but can Singapore's capital keep up?

Back to the opening question: why is ShipLinker the most "Singaporean" AI company?

Singapore is one of the world's most important maritime hubs. Shipowner, trader, and broker density is globally top-tier. MPA (Maritime and Port Authority of Singapore) is actively pushing maritime digitization. A complete maritime finance and arbitration system is right here. ShipLinker is already in discussions with MPA to enter their ecosystem with official introductions to the shipowner side. There are even maritime-focused funds locally — Motion and ShipFocus — small but indicating a vertical ecosystem forming.

This type of company isn't transplanting a Chinese model to Singapore. It exists because Singapore's industry density enabled it to grow. It needs physical proximity to enough clients, real industry demand and feedback, and an investor ecosystem that understands maritime.

But this is also where the contradiction lies.

Singapore local funds' DPI (distributed to paid-in capital) hasn't exceeded 1 since 2016. The best is around 0.6 — LPs have gotten back at most 60 cents on every dollar invested. The entire local VC ecosystem hasn't made money for LPs in the past decade.

This underlying reality shapes their investment behavior: valuations capped tightly, generally not exceeding $30M post; due diligence processes painfully slow — some founders endure 6 months of DD without receiving a term sheet; meetings take two weeks minimum to schedule, email replies in three to four days are considered fast.

For a project like ShipLinker, you need investors who both understand the industry and can decide quickly — but local VCs generally lack both qualities. So Ling's strategy is to secure angel and industry capital first. VCs can join if they want; if not, wait for the next round.

This is actually a very typical microcosm of AI entrepreneurship in Singapore: the industry soil is good, the startup direction is right, but the local capital ecosystem is far from mature enough to support these companies' rapid growth. The smartest money comes from industry. The fastest money comes from industry too. VCs are actually the slowest, most hesitant link.

For deep-industry AI companies, perhaps the most pragmatic path is: build the business and client relationships in Singapore, but don't limit fundraising to local sources. Angels from within the industry, institutional capital from global funds.

The best AI companies in Singapore might be precisely the ones that Singapore's local capital finds hardest to understand.

This is a Field Note from StolenChat「离线时间」. If you're building in AI and thinking about going global, I'd love to hear from you.