Presented by Supermicro
Three decades ago CERN launched the world wide web into the public domain, and Supermicro CEO Charles Liang launched a motherboard and power supply company in Silicon Valley.
The company targeted PCs out of the gate, but soon turned its technological expertise to cornering the rapidly exploding server market — and it has grown steadily ever since. It was a historic convergence of right place, right time, right technology; today, Supermicro has hit the bullseye once again. The company saw record revenue and 50% year-over-year growth in fiscal year 2023, as the compute demands — and competitive necessity — of newly democratized generative AI takes center stage across industries.
Recently Liang sat down with VentureBeat Editor-in-Chief Matt Marshall at the Supermicro headquarters in San Jose to discuss what’s behind the company’s impressive trajectory. “Every year we are growing about 40 percent or more, and profitability is more than double year over year thanks to the generative AI opportunity,” Liang says. “We provide not just system design, but rack plug-and-play for our customers, who just have to plug in a power cable and a data cable, and they are ready to run.”
The company was already positioned as a leading supplier of end-to-end, application-optimized building-block solutions for a broad array of workloads and vertical industries, and providing complete rack level system integration and total solutions with plug-and-play offerings has already changed the game for customers of every size.
The building blocks of success
“Thirty years ago, when I designed our product based on building-block architecture, I did not know it would be as powerful as it is today,” Liang says. “It’s 30 years of improvement day after day, year over year. Now, with our building-block solution, we standardize all our parts and subsystems to easily customize for different customer applications, data centers and workloads, we can easily optimize for customers, while optimizing for cost, performance, maintenance, inventory and time to market.”
These building blocks are part of a comprehensive portfolio of processors, memory, GPUs, storage, networking, power and cooling solutions that have been designed, optimized and tested rigorously. Close collaboration with vendors, many of which are situated right next door to the company’s San Jose headquarters (such as Nvidia, Intel and AMD), offers time-to-market and cost advantages.
This modularity is a crucial competitive advantage at a time where the supply chain crunch has undermined more than one technology company. The ability to swap out less available parts for more readily obtainable hardware, as well as bring new and improved technology to market as quickly as it’s developed and tested, has been a huge benefit for customers in the race to derive value from AI.
For instance, the company is actively working to accelerate the performance, scalability and reliability of its rack scale AI solutions to support computationally intensive generative AI, large language model (LLM) training and high-performance computing (HPC) applications while handling ever-growing model sizes. To that end, they recently announced support for the new NVIDIA HGX H200 built with H200 Tensor Core GPUs. Its AI platforms, including 8U and 4U Universal GPU Systems, are now drop-in ready for the HGX H200 8-GPU, 4-GPU.
Compared to the NVIDIA H100 Tensor Core GPU, it offers nearly two times the capacity and 1.4 times higher bandwidth HBM3e memory, using NVIDIA NVLink and NVSwitch high-speed GPU-GPU interconnects at 900GB/s — up to 1.1TB of high-bandwidth HBM3e memory per node. Plus, its NVIDIA MGX systems support the upcoming NVIDIA Grace Hopper Superchip with HBM3e memory.
But whatever the supplier, the building-block solution supports chips from all the major manufacturers, Liang adds.
“Our only limitation is market size — and that will only continue to grow as generative AI and other AI technology and intelligent edge solutions evolve,” he says. “I have confidence in our vendors, our engineering excellence and the capacity of our capabilities.”
When green technology is also cutting edge
Right now, data centers consume up to 3% of total global energy which is expected to increase to as high as 8% by 2030, but Supermicro has made reducing that energy consumption a central tenet of its business.
Not only are servers designed to require less power at a given performance level, but the modular building block features mean that individual subsystems like CPU, memory and storage can be upgraded without tossing the entire chassis. Plus, because many companies still prefer to use free air cooling in their data centers, servers are designed for maximum airflow, but a growing range are ready to be used with liquid cooling. That reduces fan speed as well as the need to run high-powered HVAC systems.
It also paves the way for increasingly high-density servers. In fact, Supermicro recently announced the industry’s most compact high performance GPU server: NVIDIA HGX H100 8-GPUs systems in a liquid cooled 4U system, using the company’s proprietary liquid cooling solution. It enables data center operators to reduce footprints and energy costs while offering the highest performance AI training capacity available in a single rack.
“This liquid cooled server doubles the density of our rack scale AI solutions and reduces energy costs to achieve green computing for today’s accelerated data center,” Liang says. “It saves anywhere from 10 to 30 percent in power consumption.”
A true location advantage
Building block architecture along with rack plug-and-play are cornerstones of the company’s strategy and competitive advantage. Of course, close competitors have the ability to create plug-and-play systems, but one of the things that sets Supermicro apart from the crowd is its location.
“We are the only platform or hardware company so close to NVIDIA, Intel and AMD, so our people work together from early morning to evening,” Liang explains. “And some of our major customers have headquarters in San Jose or close by. We work together from design stage to validation stage to deployment stage, which helps accelerate time to innovation and time to market for customers in a wide range of industries and for any workload.”
The company can also design, build, validate and deliver thousands of fully populated racks of the latest servers, storage, networking systems and management/security tools to customers worldwide within a week, from manufacturing facilities on multiple continents.
“We’re in the hundred-meter dash,” Liang adds. “Everyone can run, but we simply have trained longer and run faster.”
Don’t miss the video interview with VentureBeat Editor-in-Chief Matt Marshall and Supermicro CEO Charles Liang right here.
Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact
