From Silicon to Intelligence: What Taiwan's Founders Showed San Francisco at Founders After Hours

Fifteen Taiwanese startups closed out TechCrunch Disrupt week with a pitch night and demo evening in San Francisco. Here is why the format worked, who was on stage, and why the founder DNA that made Taiwan the heart of the semiconductor world is exactly what the next phase of the AI revolution calls for.

SAN FRANCISCO, October 13, 2026: Funders & Founders After Hours: Taiwan Innovation Meets Capital, an Official TechCrunch Side Event produced by Sparknify.

Fifteen Taiwanese startups commandeered a space at the iconic Phelan Building on 760 Market Street in San Francisco for an evening of Funders & Founders After Hours: Taiwan Innovation Meets Capital on October 13, 2026. A couple of blocks over, TechCrunch Disrupt was in full swing at Moscone West. The gathering was an Official TechCrunch Side Event at a venue set up to do one thing: place founders with hard physical problems to solve before the technology executives, corporate venture teams and investors who want to see them.

For two hours there were pitches from the stage, demos at the booths and the sort of after hours talk that comes when the formalities are done but no one has left. In this piece we review the format and what made it work, who was on stage and why the DNA of these founders is precisely what the coming stage of the AI revolution will call for.

Why the evening worked

There are two ways a pitch night can play out. It is either a procession of slides that are quickly forgotten or, as it should be, a means for a room of busy investors to home in on the three or four conversations they have any interest in having. This was the latter by virtue of its structure.

Judy Jiang, International Manager with Innovation to Industry (i2i), put the frame on the program from the outset. What was on offer was not an ideas showcase but companies that had been proven at home and were prepared to scale. The founders were given a narrow window to cover the problem, the product, traction and the ask. There was no time for warm up or filler.

After that the room changed character. With a numbered booth by the windows for each company, an investor could make ten steps after a good pitch and get their hands on the product. Lixel's spatial display was up and running with a 3D model of a human heart one could reach into. SleepWell, Aztron Medtech and Coherence had their devices out on the table next to each other. Brochures were not being perused; people were on to their second and third questions, the only indication of value at such an event.

The timing was also in its favor. As part of Disrupt week and coming on the heels of the Real World AI session of that evening, the crowd was thick with those already in deal mode. The founders did not need to make the case for being in San Francisco. That was understood by all in the room.

The semiconductor DNA

Any discussion of Taiwan and technology has to include semiconductors. The island is the maker of the vast preponderance of the world's most sophisticated logic chips, and one could say that all the cutting edge AI models in training today are running on silicon that has been through a Taiwanese fab. This kind of pre-eminence is not the work of a single firm or an act of fortune. It is the product of qualities put in place over forty years, and they were in evidence on the stage at 760 Market Street.

Depth rather than hype is the first of those qualities. In yield, process control, reliability and cost, the unglamorous side of things, Taiwan's semiconductor industry has been superior. The founders who made their pitch that night put it in the same terms. Wayne Tsai of Phasetrum was not speaking in abstractions about satellites; he was on about phase resolution on an RF chip and matters of yield and calibration time. Cole Wu of Coherence Biotech did not put forward "AI for drug discovery." His pitch was the machine that can do in one what would otherwise be twenty seven manual hand offs in a two week cell experiment.

Then there is the long view taken on capital and partnership. Public research bodies and an industrial policy with patience sowed the seeds of the chip ecosystem in Taiwan, which was then put on a scale by a close knit web of suppliers. That is the ecosystem from which the startups at Founders After Hours came to San Francisco: Taiwan's Small and Medium Enterprise and Startup Administration (SMESA), Startup Terrace, i2i, Innoventure and Sustainable Impact Capital. These are not founders working in isolation. They have the support of institutions, the imprimatur of Taiwanese hospitals and universities and, in some instances, the National Development Fund.

A manufacturing instinct is another hallmark. Much of the work done by these teams is hardware or software that is intimately tied to it. They know the way to get a device past the FDA, a chip to tape out, a benchtop system within a few percent of a lab result. Such an instinct is uncommon in the software first startup milieu and will be the deciding factor in the coming stages of AI.

Finally, there is a quiet global ambition. With a home market of only twenty three million, Taiwanese companies are accustomed to looking further afield. Each of the founders on stage had been proven in their own country and was laying out a straightforward route into Europe, Japan, the United States or Southeast Asia. None of them were seeking leave to go global. They wanted the right partner to do it more quickly.

Why this matters for the next AI revolution

Where the initial AI revolution was a matter of data centers and models, the coming wave is one of infusing intelligence into the physical world. It will be found in hospitals and power grids, in satellites and factories, in operating rooms and the devices in people's hands. This change is to the advantage of founders with an equal grasp of atoms and bits, which happens to be where Taiwan has its strengths.

Examine the fifteen companies and a pattern emerges. A few are constructing the infrastructure for AI. Others are making the interfaces that will allow people to engage with it off screen. Some have the real world data that AI models cannot obtain from the internet. And some are applying AI to medicine in a way that puts a premium on precision and regulatory discipline over speed. In aggregate they do not seem a random assemblage but rather a map of AI's next destination.

The founders on stage

Here is a closer look at several of the companies that pitched. They are grouped by the role they play in that map, rather than by industry label.

Building the infrastructure AI runs on

For the satellite and radar markets, Phasetrum is putting together RF chips and phased array antennas. The Hsinchu firm was set up in 2023 with support from Taiwan's National Development Fund to deal with a particular bottleneck: the cost of user terminals for low earth orbit constellations. About half of that expense is in the RF front end. With its patented architecture, Phasetrum marries the antenna to the RF IC and does phase calibration over several stages. The team claims this produces much better yields on the production line, calibrates quicker and runs on less power. A number of satellite operators are already lined up to evaluate it. In a way it is semiconductor DNA at its most basic, put to work on the physical layer of an AI connected world.

KiWi New Energy has an AI operating system for energy under construction. It takes real time data from renewable projects, smart meters and utilities to project when green power can be best used, then calls the shots on storage charging and discharging on a minute by minute basis. FamilyMart is among the partners deploying the system in thousands of its outlets; Decathlon is using it to get ahead on its RE100 obligations. With AI data centers driving electricity demand ever higher, such an intelligent grid layer is no longer just a matter of sustainability but infrastructure proper.

Autonomous machines need to see, connect and run, so LongLink makes the communication and vision systems for them. Jane Chien, founder and CEO, has put the technology to the test with clients in the U.S., Japan, India, Eastern Europe and Taiwan. Relying on Taiwanese know how for semiconductors and doing the assembly and testing in Guam ensures the product is in full compliance with the NDAA. In the case of drones and autonomous ground systems, where the market is expanding quickly, a trusted supply chain outside China is not optional. It is what closes the sale.

Building the interface for AI beyond the screen

Lixel put on a show with one of the night's most crowded booths. The company has an unadorned but ambitious pitch: if the PC brought us the mouse and the smartphone touch, AI calls for a spatial interface. To that end Lixel has a glasses free 3D display which allows a user to view and work over a floating 3D object by hand without the need for a headset. More than eighty international patents are held by the firm and its technology is in use at five large hospitals and four medical schools for such things as radiotherapy planning, bronchoscopy navigation and anatomy instruction. A visitor reaching into a rendered heart and turning it with a gesture was more effective in making the case than any slide could be.

Then there is Optiqb AI Technology, which takes on the problem from the consumer angle. An ordinary phone is made into a 3D device with their naked eye 3D screen protector and an app driven by AI to track the eyes of the viewer and render 2D video as 3D on the spot. Founder Jye Lin and his team have secured purchase orders for 150,000 units in Asia and an agreement with Singtel, and were first prize winners in the SwissTech Global Deep Tech Battle at VivaTech. It serves as a reminder of the surprises Taiwan's display and optics know how can put forth.

Collecting the data that AI cannot find online

DATAYOO Application makes of satellite data a new form of physical intelligence. Its SpectraEye technology is a kind of virtual AI lens for current satellites that draws on spectral information to determine soil moisture and organic carbon as well as forest volume and water conditions, even beneath the vegetation canopy and in the soil itself. There is no need for new satellites. Originally focused on precision agriculture with the FarmiSpace platform, the firm is moving into disaster response, land management and water resources in Taiwan and Southeast Asia.

Anya Cheng, founder and CEO of Taelor AI, has put together an AI layer for fashion brands based on post wear data from the real world. For the consumer there is an AI stylist and a menswear rental subscription with good retention rates. But it is the brand side that is of interest: over a hundred have signed on because each rental yields data on preference, fit and durability that one would not get from purchases alone. Taelor has been named an Inc. Best in Business winner and counts an early investor in TSMC among its backers, not uncharacteristic for a business that makes training data out of daily habits.

TwoSevenths is addressing a 2026 issue head on: when AI assistants are the ones making decisions on where consumers go, any business the models cannot see is left without the customer. Chien Hsun (James) Chu, a co-founder of KKday, has made a generative engine optimization platform for lifestyle brands that is rooted in first hand appraisals of actual visits. With more than two hundred paying clients the operation is profitable in Taiwan and TwoSevenths is now making its way into the US.

Putting AI into medicine

With Keystone, Coherence Biotech has put together a sealed benchtop system to conduct a multi day cell experiment in a single enclosure and on one timeline. Incubation, liquid handling and imaging are all part of the same record. The reason is straightforward: while AI can put a drug candidate on paper in a day, weeks of manual labor are required to test it, and preclinical experiments have a habit of not being reproducible. Backed by validation from an AI drug discovery partner and Academia Sinica, Keystone's results are within a hair of what is done manually and customers are already inquiring to make a purchase.

The case for KURA Care begins with its founder, KJ Yang. He did not simply come across the issue of post acute cardiac care, he went through it with cancer, chemo and heart failure, and the patient's quandary of whether it is safe to exercise. The company's agentic AI virtual coach uses information from wearables and electronic health records to set safe and individualized exercise objectives for cardiovascular and cancer patients. A randomized trial at UC San Diego was telling, with high levels of engagement and a drop in blood pressure that was meaningful. KURA Care is now working with providers in Southern California.

SleepWell Technology has set about turning any phone into a sleep health screen. There is no hardware to buy; a face scan of the geometry and blood flow will give you vitals and a sleep health score. They are also developing an EEG based sleep mask to measure and improve sleep at home, which has been put through its paces in a clinical study at Taipei Veterans General Hospital. Given the hundreds of millions of adults with undiagnosed sleep apnea and the lack of lab capacity to find it, having a screening tool in a phone is exactly the sort of leverage AI ought to provide in healthcare.

Xensur Medical is putting in place the data infrastructure for microsurgery down the road. Founder Alex Huang is a computer scientist as well as an eye surgeon and his view is that surgical robots need exact data on how experts do their job if they are to learn. Xensur's platform follows real instruments to under twenty microns of accuracy. The structured datasets so produced are used for training surgeons now, will offer real time guidance in the future and eventually train the robots.

Rounding out the medical cohort are Aztron Medtech and iPreg, two examples of Taiwanese device makers meeting the highest regulatory standards. Aztron's TendMIN system has FDA clearance for a minimally invasive tendon repair via an incision much smaller than usual. Published figures show less time in surgery and fewer complications and the system is in use at top US health systems. Then there is iPreg, whose FDA cleared biochip for sperm sorting relies on thermotaxis to select healthy sperm for fertility treatment. The firm is making inroads in the United States, Japan and the Middle East.

What the investors took home

What made Founders After Hours so useful was not any one pitch in particular but the pattern that put itself in evidence as the night wore on. All of the companies had a bona fide product, with real revenue or clinical data to show for it, or actual purchase orders. They were all validated by serious institutions back home and each was putting together something in the physical world, the very gap AI investors are at present set about filling.

There is no accident to that pattern. It is the result of a country devoting forty years to the manufacture of the most exacting objects known to man and then turning that culture toward the next frontier. The founders who have come to San Francisco bring that with them; they are patient and technically thorough, of a global mind and with no tolerance for hype.

Looking ahead

As the capstone to a three part series at TechCrunch Disrupt week, Sparknify put on Funders & Founders After Hours for SMESA. The reaction in the room was telling of one fact: there is a fast growing appetite in Silicon Valley for companies from Taiwan. Those investors not in attendance can still have a word with the companies listed above. And for the founder in Taiwan contemplating a move to the U.S., this group is evidence that such a route is possible.

Taiwan's silicon has been essential to the first AI revolution. The founders here are arguing that the same DNA will be of equal import in the next.

This article draws on the event program, the founders' presentations at Funders & Founders After Hours, and materials published by the participating companies. Photos courtesy of Sparknify and the featured companies.