The story
1993-2009: Founding, Ablecom, and the Hyperion project
In 1993, Supermicro began as a five-person business operation run by Charles Liang, a Taiwanese-American, alongside his wife and company treasurer, Chiu-Chu Liu, known as Sara. Prior to founding Supermicro, Liang earned a B.S. in electrical engineering from the National Taiwan University of Science and Technology and an M.S. in electrical engineering from the University of Texas at Arlington. Liang holds several patents for server technology and was previously the president and chief design engineer of Micro Center Computer, a motherboard design and manufacturing company, from July 1991 to August 1993. Liang stated that the company was able to turn a profit after six months of operation.
The same year it was founded, Supermicro established its first European partnership with Boston Limited, a UK-based IT solutions provider, which launched Supermicro products into the UK and European markets. The continued collaboration has seen several Supermicro technologies introduced to Europe, including the P5EXTRA in the 1990s, the 1U Twin in 2006, and innovations such as the Supercool 1U Twin in 2007 and high-performance GPU computing systems in 2009.
In 1995, Supermicro released a dual-CPU motherboard for servers, becoming one of the first companies to do so. The motherboard was followed up by a four-CPU server board. In 1996, the company opened a manufacturing subsidiary, Ablecom, in Taiwan, which is run by Charles's brothers, Steve Liang and Bill Liang. Charles Liang and his wife own close to 31 percent of Ablecom, while Steve Liang and other members of the family own close to 50 percent.
In 1998, Supermicro opened a subsidiary in the Netherlands.
In 2004, Supermicro began to develop energy-saving servers, which was partially influenced by Liang watching the movie The Day After Tomorrow with his family. In 2004, it also became aware of a federal investigation into its sales of computer systems to Iran and set up an export-control program that same year. In 2006 the company pleaded guilty to a felony charge of violating the US embargo against Iran and paid a $150,000 fine.
In 2008, Supermicro was among ten computing companies brought on to the Hyperion project by the National Nuclear Security Administration's Lawrence Livermore National Laboratory. The companies helped to develop a testbed for high-performance computing technologies for the purpose of maintaining the U.S. nuclear weapons stockpile while avoiding underground nuclear testing, and improving the industry's ability to make petaFLOP/s (quadrillion floating operations per second) computing and storage more accessible.
2010-2019: Taiwan science park and San Jose headquarters
In May 2010, Supermicro further expanded into Europe with the opening of its system integration logistics center in the Netherlands.
In 2012, Supermicro debuted its new 2U and 4U/Tower platforms.
Read the full text
In 2012, the company opened its Taiwan Science and Technology Park, totaling $99 million in construction costs. The Taiwanese factory was expanded in 2019 and came online in 2021.
In September 2014, Supermicro moved its corporate headquarters to the former Mercury News headquarters in North San Jose, California, along Interstate 880, naming the campus Supermicro Green Computing Park. In 2017, the company completed a new 182,000 square-foot manufacturing building on the campus, which was designed to meet LEED gold certification.
In 2016, Supermicro sent 30,000 MicroBlade servers to a Silicon Valley data center with a claimed power usage effectiveness (PUE) of 1.06. While Supermicro did not name the customer, it was likely Intel, who opened a similar data center in November 2015 with a PUE of 1.06. In 2018, the company was briefly delisted from the Nasdaq after delaying to file financial reports by nearly two years.
2020-2023: Expansion, Rakuten partnership, and liquid-cooled servers
In April 2020, Supermicro announced the H12 A+ Superblade, a blade server based on the 2nd gen Epyc 1P family of CPUs. It was the first blade server platform to implement AMD's Epyc processors. The same year, it settled with the United States Securities and Exchange Commission over violations in accounting practices between 2014 and 2017, agreeing to pay $17.5 million in penalties.
In April 2021, Supermicro introduced over 100 application-optimized server product SKUs using (new at the time) 3rd Gen Intel Xeon Scalable processors, including Hyper, SuperBlade, the Twin Product Family (BigTwin, TwinPro, and FatTwin), Ultra, CloudDC, GPU, Telco/5Gand Edge servers. The company expanded its San Jose campus in September 2021 with a manufacturing facility for advanced storage and server equipment. Supermicro was reported to have 2,400 people working in San Jose.
In 2023, Supermicro partnered with Rakuten Symphony on high-performing Open RAN technologies and storage systems for operators of cloud-based mobile services. Later in the year, Supermicro debuted servers with liquid cooling, focusing on ESG policies. The servers save approximately 40% of the power expended on air-cooled data centers. In June 2023, Supermicro saw increased demand for its large language model optimized AI systems, featuring NVIDIA chips.
2024-present: xAI Colossus, DataVolt agreement, and US manufacturing expansion
In July 2024, VentureBeat reported that Supermicro would be providing half the servers for Elon Musk's artificial intelligence start-up, xAI, with Dell providing the other half. The project was completed in 122 days according to the company, resulting in the creation of a 750,000 square foot Memphis-based data center to host Colossus, a supercomputer. Supermicro also supplied servers for Tesla's Gigafactory Texas.
In October 2024, Supermicro announced a partnership with Fujitsu, a Japanese IT services company, where the two companies will jointly develop energy-efficient servers and liquid-cooling systems for high-performance computing, generative artificial intelligence, and data centers.
On December 25, 2024, Super Micro Computer announced a joint venture with Taiwanese development company Guo Rui to build an artificial intelligence data center powered solely by renewable energy at an undisclosed location in Taiwan.
In February 2025, Supermicro began building its third California-based manufacturing campus. The new campus is being developed with the intention to increase production of liquid-cooled services for data centers. The company produces a majority of its servers in California. Following a push for more state-side manufacturing by American President Donald Trump, Supermicro considered expanding server production in states like Mississippi and Texas. A few months later, in July 2025, Supermicro expressed its interest in expanding investment regarding manufacturing in Europe to meet artificial intelligence demand in the area.
In May 2025, Supermicro announced a $20 billion partnership with DataVolt, a Saudi Arabian data center company. Under the agreement, Supermicro will supply GPU platforms and rack systems for DataVolt's AI campuses in both Saudi Arabia and the United States.
In September 2025, Supermicro announced an artificial intelligence-based partnership with Nokia, focusing on integrated data center network products. In the following month, Supermicro would partner with Hitachi Vantara, each allowing the other company to offer their products through the opposing companies, focusing on artificial intelligence applications and data lakehouses.
In October 2025, Supermicro created a subsidiary focusing on American federal agencies, which would provide cloud-services and data center materials manufactured from its facilities in Silicon Valley, California. From November 2025 to June 2026, Supermicro added eight additional production facilities in the United States, including its Data Center Building Block Solutions campus in San Jose, which was launched at the end of April 2026. As of June 2026, the company planned to open two additional facilities in Taiwan. Supermicro also expanded its Netherlands-based operations. In addition, it operates a Malaysian facility, catering towards Southeast Asia and parts of Europe.
In November 2025, Supermicro reported fiscal first-quarter 2026 results under expectations, stating that "design win upgrades" on server projects, where $1.5 billion in expected revenue shifted due to a high-volume customer requesting last minute updates, which extended integration, testing, and validation times for new GPU-based systems. From the last two years until then, Supermicro's revenue had grown from $7 billion to $22 billion. Additionally, the company received $13 billion in orders on systems based on the Nvidia Blackwell Ultra-based GB300 architecture. In the same month, Supermicro displayed new products at the 2025 Supercomputing Conference, including the Nvidia GB300 NVL72 rack, 8U and 6U SuperBlade multi-node systems, 2U FlexTwin, HGX B300 4U rack, GB200 NVL4 1U server, and GB300-based developer workstation. Updates to the modular and blade families such as BigTwin, MicroBlade, and MicroCloud were also shown at the conference.
In January 2026, Supermicro announced it had expanded manufacturing capacity surrounding it's liquid-cooling technology in support of Nvidia's Vera Rubin (including the Vera Rubin NVL72 SuperCluster and the HGX Rubin NVL8) and Rubin AI platforms. In addition to Supermicro's close ties to Nvidia, the company also has relationships with AMD, Broadcom, Intel, Samsung and Micron.
In June 2026, during CEO Charles Liang's keynote speech at Computex, Supermicro demonstrated its VR200 NVL72 rack using Vera CPUs and Rubin GPUs, which featured liquid-cooling through the company's newly updated proprietary coolant. Also in that month, Supermicro released a series of ARM-based rack-scale AI systems, featuring ARM AGI CPUs. The series consists of the 2U Hyper Server, 5U GPU Server, 2U4N Liquid-Cooled Server, 2U Hyper-E Server, and the 1U 4N in an OCP ORW rack.
In September 2026, Supermicro released a workstation built around Nvidia's GB 300 Grace Blackwell Ultra superchip, containing 748 GB of RAM split between 498 GB of system memory on the Grace CPU, and 252 GB of HBM3e memory on the Blackwell Ultra GPU, intended for "research labs and enterprises running large models locally".
On May 6 2026, BNN Bloomberg reported that Supermicro and Nano Nuclear Energy signed a memorandum of understanding to provide power to AI data centers through the creation of on-site nuclear power. Later in the month, Supermicro announced a collaboration with Verda, with the main goal of producing rack-scale systems designed for AI cloud infrastructure.
From Wikipedia’s article on Supermicro, shared under CC BY-SA 4.0.8