Yih Lin Teh left three continents of consulting to build Axross, a startup that runs the invisible energy systems inside Southeast Asia's factories. The frontier of climate tech, it turns out, hums quietly in the background.
Every conversation about artificial intelligence seems to circle the same handful of images: chatbots writing emails, models painting pictures, code that writes itself. Yih Lin Teh went looking somewhere quieter. Somewhere with no glamour at all - the air conditioning humming inside a semiconductor factory, running twenty-four hours a day, drinking electricity nobody is watching closely. That hum is where the carbon lives. It is also where the money leaks. And it became the whole idea behind Axross.
Teh is the co-founder and CEO of Axross, a Singapore-based startup that went through Y Combinator's Winter 2022 batch. The company builds AI-driven autonomous HVAC control systems - software that decides, in real time, how a building's heating, ventilation and cooling should behave to keep conditions exactly right while spending as little energy as possible. In a chip plant, where humidity and temperature are not comfort settings but manufacturing requirements, that decision runs continuously, and the stakes are high.
The self-description Teh keeps returning to is deceptively plain: "econometrician turned climate tech entrepreneur." It is worth pausing on, because it explains almost everything about how Axross works. An econometrician does not guess. An econometrician measures - every degree of cooling, every kilowatt, every ringgit - and looks for the relationship hiding in the numbers. Point that instinct at a factory's energy bill and you get a company.
Before the startup, there was a career that looked nothing like a garage. Teh studied economics at the London School of Economics and did graduate work connected to MIT, then spent years inside the machinery of global problem-solving. There was time at Strategy&, part of the PwC network, where economics meets corporate strategy. There was Dalberg, advising governments and multinationals on turning green technology into something that actually turned a profit. There was Modern Electron, closer to the hard edge of energy itself. Three continents of it.
It is the kind of resume that usually keeps compounding in one direction - bigger firms, bigger titles, further from the factory floor. Teh went the other way. After working across three continents, the decision was to come back to Southeast Asia, specifically to make the region's manufacturing sector more energy-efficient and more competitive. Not the abstract version of climate work. The version with a service technician, a chiller, and an electricity meter attached.
That line is the entire Axross thesis compressed into a sentence. Build for the region, not the demo. It is also a tell about how Teh thinks - less about winning a single account, more about the standard a whole industry could reach if the tooling got good enough.
Teh has a favorite way of puncturing a comfortable assumption. Your phone says "designed in California, made in China." The reality is messier and more interesting: thousands of its components pass through Southeast Asian facilities - in Singapore, Malaysia, Vietnam, Indonesia - for testing, assembly and packaging. A quiet, enormous supply chain that most people never think about, holding up a global industry.
The argument that follows is both patriotic and practical. Small countries, Teh contends, cannot compete independently. In a world of shifting tariffs and rearranging alliances, regional cooperation across ASEAN is not a nice idea - it is the only way the bloc stays central to how the world's electronics get built. And the way to stay competitive is to manage resources wisely, which is exactly where the boring, load-bearing problem of energy comes back in.
There is a reason most climate startups reach for the photogenic stuff - solar arrays, electric vehicles, gleaming batteries. HVAC is the opposite of photogenic. It is ducts and chillers and a control panel in a basement. But heating, ventilation and cooling is one of the largest energy loads in any industrial building, and in the tightly-controlled environments of advanced manufacturing it never switches off. Shave a meaningful percentage off that, continuously, and the carbon and the cost both fall in a way a press release cannot fake.
Axross treats the factory's climate as an optimization problem rather than a fixed setting. Machine learning forecasts cooling demand before it arrives. The system sequences equipment - deciding which chiller runs, when, and how hard - to hit the required conditions with the least energy. It is the difference between a thermostat that reacts and an engine that anticipates.
Axross did not begin as an HVAC company. Its earliest life was tangled up with bipi, a venture aimed at Indonesia's warungs - the small corner stores that hold up much of the country's retail. The idea there was to let shopkeepers access brand promotions, earn rewards and manage their operations from a phone. Teh co-founded that effort with David Chew, a consultant-turned-entrepreneur with a first-class degree in aeronautics who now serves as Axross's CTO.
The pivot from serving shopkeepers to serving semiconductor plants is a big leap on paper. But the underlying instinct is consistent: find the unglamorous, overlooked system that a whole category of businesses depends on, and make it work better with technology. First it was the corner store's operations. Then it was the factory's energy.
Axross is not a giant. It is a lean team, headquartered in Singapore with hiring and engineering in Kuala Lumpur, that has grown to roughly a million dollars in annual revenue while staying small enough that everyone can see the whole board. The open roles tell you where the effort goes: data engineers, data scientists, IoT engineers - the people who turn sensor readings into decisions. It is a company built by people who like the readings.
There is something quietly contrarian in the whole setup. The prevailing story of ambitious technology says you build in Silicon Valley and export outward. Axross runs the opposite play: build climate AI in and for Southeast Asia, adapted to local conditions, aimed at local customers, contributing to local industry debates. Not the world's leftover solutions retrofitted onto ASEAN factories - the region's own tooling, made on purpose.
What makes Teh interesting is not a single dramatic bet. It is the steadiness of the direction. An economist's training that never left - the habit of measuring before believing. A career that walked toward the factory floor instead of away from it. A product aimed at the least fashionable, most load-bearing system in the building. And a conviction that the region doing the world's quiet manufacturing work deserves technology built with it in mind.
The path to a lower-carbon industry is often imagined as something visible - panels on rooftops, turbines on ridgelines. A good deal of it will be invisible instead: the air conditioner nobody thinks about, running smarter because software finally learned its rhythms. Yih Lin Teh made that invisible frontier the work. It is not the loudest corner of climate tech. It may turn out to be one of the most useful.