Taiwan's Startup Ecosystem: What Three Founder Calls Revealed

I was on vacation this past week with some friends, rapidly tapping away at my laptop at about 10:00 PM on a Tuesday while they watched some mid-2000s reality TV. (Tabatha Takes Over — strangely addictive in a way that is nostalgic for 2010…)

“What are you doing?” a friend asked me.

“I keep getting introduced to these startups in Taiwan that want to meet with me, and the time difference is killing me.”

His reply: “Wow. Well, that is a sentence.”

So how did we get here?

Well, in actuality, it started when I was somewhere between unemployed and not fully yet self-employed. That’s how any good portfolio career gets its start. I’d been invited to an evening founders talk at Silicon Valley Bank in New York. (Yes, it still exists…) And I met a real estate private equity associate over mini-hamburger sliders. When I told him I worked in DeepTech, he said I should meet someone he played with on his college soccer team who now worked at Plug and Play Tech Center in Sunnyvale. DeepTech is oddly specific in that way…

So one thing led to another, and I became a Startup Mentor for Plug and Play, duly submitting my details into the netherworld of their AirTable mentor database.

And I waited…

Six months later, I got a nice inbound from a Program Manager at Plug and Play in Taiwan. Some founders had seen my profile and wanted to meet with me. I was flattered! (I’ve never even been to Taiwan.) So I met with the founders, provided some guidance, and signed an NDA or two. I even did some work for them at the time. And then I went about my life, got a new role here in New York… c’est la vie.

Well, last week, a different Program Manager in Taiwan reached out (the last one had moved to the Netherlands) because some more Taiwanese founders had seen my profile in this database, and wanted to meet with me. (It must be some profile, right?)

This is what I call synchronicity, and I’ve written about it before. You plant a flag somewhere — a mentor profile, a byline, a conversation at an event you almost didn’t attend — and it sits dormant for a few months or even years, and then it produces a meeting that you could never have engineered on purpose even if you’d tried. Twice now, a soccer-team introduction from a man I met once in Midtown has resulted in founders on the other side of the planet setting their alarms to talk with me. (And me setting mine! The 12-hour time difference works both ways, and neither direction is kind.)

So who did I meet with this week at 7:00 AM ET? Let me tell you about three of Taiwan’s most interesting startups — and why you should care:

The first call was with a chip-design startup that spun out of ITRI — Taiwan’s government-backed Industrial Technology Research Institute. If you don’t know ITRI, here is the shortest possible version: it is the research engine that helped seed Taiwan’s entire semiconductor industry. (TSMC traces its lineage there.) When a team spins out of ITRI, it does not spin out with a pitch deck and a dream. It spins out with years of applied R&D and, in this case, paying customers already in hand.

Their thesis is delightfully contrarian in a moment when the entire world is genuflecting before the GPU: for a certain class of AI workloads, you don’t want a GPU at all.

Their architecture runs algorithms directly on FPGA hardware — at the circuit-design level — rather than through the software layers a GPU pipeline requires. Strip out the software mediation and the batch processing, and inference latency drops from seconds to microseconds. That’s not incrementally faster. It’s a different unit of measurement.

Two second-order effects made me sit up on this call:

1. First, the IP protection angle: Software-defined AI is, structurally, easy to copy — your model is a file, and files travel. When the algorithm is embedded in the very circuit design itself, reverse-engineering it becomes extraordinarily difficult. For customers in semiconductors and other sensitive industries who lie awake worrying about IP theft, that feature is the product.

2. Second, the power math: Workloads running directly on hardware don’t need racks of servers just to keep up. In a world where data-center power consumption has become a national security issue, “we do it in microseconds on a fraction of the hardware” is a sentence with a very long commercial tail.

And here is where the portfolio career pays its dividend: their U.S. go-to-market thinking centered on semiconductor equipment and defense. Reasonable. But I’ve spent time inside a finance, so I raised my hand and offered a third vertical: high-frequency trading. Firms east of Silicon Valley spend staggering sums every year to shave microseconds off execution, and they treat performance budgets the way the rest of us treat rounding errors. Microsecond determinism on a smaller hardware footprint is practically a love letter to that industry.

The CTO, it turned out, had prior experience combining trading algorithms with real-time risk controls. The vertical wasn’t a stretch. It just hadn’t been on their map — because nobody in the room had lived inside that world.

(This is the actual value of a non-linear career, by the way. Not the resume. The map and the territory.)

The second call was with a two-year-old company working on something unglamorous and enormous: solar panels get dirty, degrade, and underperform, and the industry mostly shrugs about it.

Their entry point is an anti-soiling coating that restores and protects panel performance — applied on-site, panel by panel, with payback measured in months rather than years. Useful, but coatings are not why I’m writing about them.

Two things are.

1. The team: The CEO spent seventeen years at Epistar, the Taiwanese LED chip manufacturer, helping scale it to global #1, through an IPO, and through a series of mergers. Every co-founder has prior CEO experience. This is not a first-rodeo team; this is a team that already sold one company’s assets and deliberately pivoted into a bigger problem. At seed stage, you are backing people — and founders with this kind of operating history routinely undersell it. (I told them so, in the gentlest way I know how, which any former colleague will confirm is still fairly direct...)

2. The data: Everyone else inspects solar farms with drones and thermal cameras — surface-level snapshots from a distance, with surface-level blind spots. This team physically touches every panel they treat. Which means they are collecting micro-scale defect data nobody else has, at the exact moment the industry needs training data for automated inspection. The coating business funds the data collection business; the data then trains the AI; robotics scales the whole thing. A service business quietly assembling a data moat — my favorite kind of company: the one whose second act is hiding in plain sight inside its first.

They’re already running pilots with utilities across the Gulf and Europe. The market will figure this one out eventually…

The third call was with the founder of an AI infrastructure company building what he calls a runtime stability layer — software that sits inside AI systems and enforces behavioral boundaries while they run. Observability, containment, and active response: not just watching what your AI agents do, but being able to stop them mid-misbehavior.

If you have ever watched an agentic AI system spiral into a retry loop at 2 AM, burning your precious tokens like a teenager with their parent’s credit card and last year’s iPhone, you’ll understand the problem immediately. If you are a regulator, a bank, or anyone deploying AI in an environment where “oops” is not an acceptable incident report, you’ll understand it even faster.

Three things stood out:

1. He built the entire stack himself over two and a half years: Not an engineer by training — a cybersecurity and go-to-market person who taught himself what he needed, then hired engineers to productize it. Three exits behind him. This is what obsession looks like when it compounds. (Like recognizes like).

2. He’s filing patents like it’s a part-time job: In a category where every major platform vendor offers some flavor of “AI observability,” defensible IP is the difference between a company and a deprecated feature.

3. The market-entry logic is the tell: He chose Japan as his first market because it is the most heavily regulated AI environment in Asia. Most founders run from regulation; he’s running toward it, on the theory that the customers with the strictest requirements are the ones who will actually pay for the governance. That is a strategy formed by someone who has sold into enterprises before.

You knew it was coming.

Three companies, three completely different sectors — chips, solar, AI infrastructure. But squint, and I’ll show you how they rhyme:

1. Taiwan’s ecosystem is compounding in public, and almost nobody in New York is watching: We flatten Taiwan into one four-letter company. But the machinery around it, the government research institutes that spin out real companies, deep supply-chain relationships, founders with decades of hardware operating experience, is producing startups that arrive with revenue, customers, and IP on day one. While American seed rounds fund pitch decks, ITRI spinouts show up with paying customers. That contrast should provoke more discomfort than it does. I wrote about the AI funding bubble — startups raising before product-market fit — earlier this year. These teams are the photographic negative of that essay. They’re the real deal.

2. Hardware is eating its way back up the stack: For fifteen years, the valley catechism was “software is eating the world,” and hardware was the commodity underneath. Every one of these calls suggests the pendulum swinging back: AI advantage increasingly lives in the physical layer, in circuits, in coatings, in sensors, and in power budgets. The countries and companies that never stopped being good at atoms are suddenly holding better cards than the ones that specialized exclusively in bits and bytes…

3. The judgment layer travels farther than you think: None of these founders needed me to explain their technology — they all understand it at depths I never will. What they wanted was translation: which U.S. verticals, which investors, how an American buyer thinks, what a pitch signals to a sophisticated audience. That is judgment, context, and orchestration — the exact things I keep arguing the AI economy makes more valuable, not less. And it’s why a mentor profile in a database can sit dormant for six months and then summon you to a 7:00 AM call: the scarce input wasn’t my expertise. It was a particular set of rooms already inhabited.

The uncomfortable version of that thesis, for you, dear reader: your technical knowledge is depreciating faster than ever. Your accumulated map of rooms, industries, and people to whom you can claim a plausible connection is appreciating exponentially, by the hour. Build that second one deliberately, and with renewed purpose.

As for me — of course I’ll keep taking these calls. The founders are all excellent, the problems they’re solving are real, and the flag I planted in a database two years ago keeps generating meetings I could never have planned. (The one this morning had a translator… that’s how you know its the real deal.)

And that, my friends, is why you should always go to networking events at formerly bankrupt financial institutions. RIP Lehman Brothers.

-Alex

If you’re building a career where the byline is the storefront and the network compounds while you sleep (ideally with fewer 7:00 AM calls), that’s what this newsletter is for — subscribe and join 2,100+ readers.

Alex Randall Kittredge writes the Substack ARK Strategy and is the founder and Managing Director of APR Strategic Consulting. He has advised 9+ CEOs, integrated workforces, and redesigned organizations across hedge funds, startups, and industrial companies. He is a graduate of Columbia University and the University of Cambridge, and mentors entrepreneurs & startup founders through Oxford Entrepreneurs Network, CamEntrepreneurs, and Plug and Play Tech Center. He is a Director of The Oxbridge Method Ltd, and is the author of the forthcoming book, How Your Side Hustles Will Save You: Creating a Durable Career that Transcends the Corporate Ladder.

Compliance Statement: All views and opinions expressed are solely my own and do not reflect those of any current or former employer, client, or organization with which I am or have been affiliated.