Alex Yeh, Founder and CEO of GMI Cloud
Founder & CEO · GMI Cloud

Alex
Yeh

Building the GPU backbone of the AI internet - one data center, one sovereign nation, one impossible bet at a time.

GPU Cloud NVIDIA Partner AI Infrastructure San Francisco
$93M Total Raised
7 Data Centers
5 Countries
6 NVIDIA Ref Partners

The Pivot That Rewired Everything

In mid-2023, Alex Yeh was running three Bitcoin mining data centers in Arkansas and Texas. The racks were humming, the cooling was adequate, and the margins were thinning. Then he noticed something in the market: developers, researchers, and enterprises were hunting for GPU compute, and almost nobody had it. He had 30 days. That was enough.

Yeh converted the entire operation to AI cloud infrastructure in a month. Not a partial pivot. Not a hedge. The whole thing. Today, GMI Cloud is one of only six NVIDIA Reference Platform Partners on earth - a list so short you could fit it on a business card with room for a bad joke. The company raised $82 million in its Series A in October 2024 and is now operating seven data centers across five countries, with a $12 billion sovereign AI infrastructure commitment in Japan and an AI Factory in Taiwan co-developed with NVIDIA.

The move was faster than most companies do a rebranding. But Yeh's career has always moved that way: materials science engineering at Johns Hopkins, then Singapore banking, then Taiwanese semiconductor family offices, then Web3 venture capital, then entrepreneur-in-residence at HOF Capital in New York. By the time he founded GMI Cloud in 2021, he had managed a $200M private equity portfolio, backed blockchain projects across Asia, and accumulated enough scar tissue to recognize a generational infrastructure opportunity when it materialized.

I see artificial intelligence as the 21st century's latest 'gold rush,' with GPUs and AI servers serving as the 'pickaxes' for modern-day 'prospectors.'

Alex Yeh, Founder & CEO, GMI Cloud

GPU Cloud, Stripped to the Metal

GMI Cloud does not apologize for being infrastructure. Where other companies layer AI products over rented compute, Yeh chose vertical integration - owning hardware down to the NVIDIA H100, H200, and incoming Blackwell processors, and building software on top. His framing borrows from Marc Andreessen's playbook but applies it one layer closer to silicon: "Vertical integration down to the hardware level is how legendary businesses are built."

The product stack runs from bare-metal GPU servers through Kubernetes orchestration, serverless inference APIs, and production-ready LLM endpoints. NVIDIA H100 time goes for $2.00 per GPU-hour; H200 at $2.60. The customer base spans AI startups and Fortune 2000 enterprises. The pitch is blunt: stop managing infrastructure, start building products.

$2.00
H100 / GPU-hour
110+
Employees
2021
Founded
$12B
Japan AI Init.

In January 2026, GMI Studio launched - a visual AI workflow platform targeting creators and developers who have never written a line of infrastructure code. The ambition is explicit: make building AI applications "as simple and streamlined as building a website on Wix or Shopify." If you squint, it looks like the same strategy AWS used to commoditize servers, applied to the GPU layer a decade later.

Funding Breakdown - Series A ($82M Total)

Equity
$15M
Debt
$67M
Prior Rounds
$11M

Lead investor: Headline Asia. Strategic: Banpu Next (power infrastructure), Wistron Corporation (product co-development)

Sovereign AI and the Asia Play

Yeh's most audacious bet is not a product launch or a funding round. It is a $12 billion, 1-gigawatt sovereign AI infrastructure initiative in Japan - an AI Factory in Kagoshima announced in late 2024, just weeks after closing the Series A. The numbers are large enough to require a second read. The country-level infrastructure ambition puts GMI Cloud in territory more commonly occupied by hyperscalers than by 110-person startups in Mountain View.

The Taiwan AI Factory, launched with NVIDIA in 2025, adds another node to the pattern. Yeh - with a background spanning Taiwan's semiconductor ecosystem, Japan, Singapore, and the United States - is building a company whose geography is its competitive advantage. Sovereign AI demand, the desire for nations to own their own AI infrastructure rather than pipe data to US hyperscalers, is a tailwind few infrastructure players are positioned to capture. Yeh is positioning GMI Cloud directly in front of it.

I want to build an invisible GPU cloud network - a silent enabler that empowers startups and enterprise innovators to bring their AI ideas to life.

Alex Yeh

From Materials Science to GPU Infrastructure

Yeh studied materials science engineering at Johns Hopkins University, not computer science, not business. The NSF Graduate Research Fellowship he received there suggests a research orientation that later showed up as the habit of studying systems before touching them. The progression from engineering to finance to venture to operations is less random than it looks: each stop added a different vocabulary for how capital and technology interact.

The Credit Suisse stint in Singapore in 2015 taught capital markets. The Mesh Ventures analyst role in Taiwan in 2017 connected him to early-stage companies. Managing $200M in CDIB's China Life Private Equity Fund gave him the operator's lens on portfolio companies. At Infinity Ventures Crypto - the Web3 arm of the $4.5B Headline Group - he backed blockchain projects in the $250K-$1M range and built pattern recognition for technology adoption curves. By the time Yeh landed as Entrepreneur-in-Residence at HOF Capital in New York in 2019, he was synthesizing a decade of watching other people build companies. In 2021, he stopped watching.

Career Milestones

2015
Investment Banking Summer Analyst, Credit Suisse Singapore
2017
Analyst, Mesh Ventures, Taiwan
2018-2020
Managed $200M PE portfolio at CDIB's China Life Private Equity Fund
2019-2021
Entrepreneur-in-Residence, HOF Capital, New York
2020-2021
Partner, Infinity Ventures Crypto (Headline Group, $4.5B AUM)
2021
Founded GMI Cloud - initially Bitcoin mining data centers in Arkansas and Texas
2023
Pivoted entire company to AI/GPU cloud infrastructure in 30 days
Oct 2024
Closed $82M Series A led by Headline Asia with Banpu Next and Wistron
Nov 2024
Announced $12B sovereign AI initiative in Japan (Kagoshima AI Factory)
2025
Launched AI Factory in Taiwan with NVIDIA
Jan 2026
Launched GMI Studio - visual AI workflow platform for creators and developers
Jun 2026
Speaking at NVIDIA GTC Live @ COMPUTEX TAIPEI on AI infrastructure

Quotes

"Our ultimate goal is to make building AI applications as simple and streamlined as building a website on Wix or Shopify."

"Vertical integration down to the hardware level is how legendary businesses are built."

"We're at a launching point in our company's journey. This funding empowers us to enhance the performance, security, and accessibility of our platform, helping businesses around the world scale even the most demanding AI workloads."

"The priority should be on building and refining their products, not spending valuable time on infrastructure management."


Interviews & Appearances

GMI Cloud CEO Alex Yeh on Building AI Internet

GMI Cloud CEO: I'm Building AI Internet (Part 1)

Under $1M to $40M in 2 Months — Alex Yeh

Under $1M to $40M in 2 Months - Why AI Workflows Are the Future

Voices in AI: Alex Yeh

Voices in AI: Alex Yeh, Founder and CEO of GMI Cloud

Delivering Scalable AI Infrastructure

Delivering Scalable AI Infrastructure - Ep. 3

Evolution of MLOps Keynote

The Evolution of MLOps - Alex Yeh, GMI Cloud (Keynote)


The Details That Don't Make the Deck

Studied materials science at Johns Hopkins - then ended up managing NVIDIA GPUs

Pivoted an entire data center company from Bitcoin to AI in roughly 30 days

NSF Graduate Research Fellowship recipient

Was youngest partner in Web3 VC before the AI pivot

His $12B Japan AI commitment is larger than many national AI budgets

Managed a $200M PE portfolio before the age of 30

GMI Cloud is one of only 6 NVIDIA Reference Platform Partners worldwide - a designation that puts it alongside hyperscalers and infrastructure giants, despite being a 110-person company founded in 2021.

Links & Profiles

Why the Bet Is Different This Time

Most GPU cloud companies were founded by engineers who understood compute before they understood business. Yeh came from the other direction. He had spent years inside private equity watching companies struggle to translate technical capability into operational scale, and inside venture capital watching early-stage founders confuse product with infrastructure. When he built GMI Cloud, he built it to solve the problem he had watched from the outside: every serious AI company eventually runs into a wall when they need to scale GPU workloads, and almost none of them want to run data centers.

The NVIDIA Reference Platform Partnership is the detail that separates GMI Cloud from the dozens of GPU cloud competitors that appeared after 2023. Six companies in the world hold that designation. It means co-engineering, early access to hardware, and a level of integration that you cannot replicate by simply reselling cloud capacity. Yeh's background in semiconductor family offices and deep Taiwan tech relationships was almost certainly a factor in securing it. In infrastructure, the relationships that look like luck are usually the ones that took the longest to build.

GPU Cloud AI Infrastructure NVIDIA Sovereign AI LLM Inference Cloud Computing Founder Series A High-Performance Computing Kubernetes Data Centers Asia Pacific Startup Venture Capital AI Factory Taiwan Japan Machine Learning

Share This Profile