From AI to customers at scale, from small enterprises to large data centers— Vik Malyala#ChanakyaaEnglish · · Chanakyaa English
“These deployments are actually quite complex. You're talking about billions and billions of dollars going into each and every data center, and the one thing that people want is stuff to be up and running reliably over a period of time.”
On , Vik Malyala, Senior VP, MD & President of the EMEA Division at Supermicro, spoke about data center investment during AI to customers at scale, from small enterprises to large data centers— Vik Malyala#ChanakyaaEnglish on Chanakyaa English.
Vik Malyala, Senior Vice President, Managing Director and President of the EMEA Division at Supermicro, has been discussing the company's role in scaling AI infrastructure for a range of customers. He stated that Supermicro is working to "democratize this whole AI adoption" by making hardware accessible to developers and customers through partnerships with AMD and neo-cloud providers such as Vulture, Tensor Wave, Cruso, and Digital Ocean, as well as through a jump start program. Malyala noted that refresh cycles have shortened from three years to less than one year, requiring faster innovation. He highlighted Supermicro's vertically integrated approach to design, development, and manufacturing, and described the company's focus on bringing OCP-inspired designs to market, including a 2U design based on the B300 that can support up to 144 GPUs per rack. Malyala has also addressed the challenges of deploying AI infrastructure, including power, cooling, and data center readiness. He stated that data centers are not always equipped for the latest technology and that Supermicro works with customers to audit facilities for factors like floor tile weight capacity and power drops. He noted that efficient data centers are already reaching 200 kilowatts per rack and predicted that 500 kilowatts to a megawatt per rack could become common within a year to a year and a half. Malyala emphasized the importance of sustainability, stating that data centers consume 2 to 3% of total energy and that liquid cooling is being used to reject 75 to 85% of heat in AI infrastructure. He also discussed the company's consideration of "Make in India" initiatives, saying Supermicro is evaluating what makes sense to manufacture locally, either directly or through partners.