Technology trends have come and gone during the nearly 20 years Linda Yang, a Supermicro senior solution manager, has worked in Silicon Valley.
When it comes to AI, the former software developer and hardware product manager knows what AI means to service providers as well as infrastructure suppliers based on her experience at companies such as Cisco and ServiceNow. Speaking from all that valuable experience, Yang predicts AI is here to stay.
"AI has emerged as the new operating system and the cornerstone of almost all applications — existing and new," said Linda, who spoke during our talk about the vast changes occurring in AI.
As businesses race to develop increasingly more impactful AI applications, the types of models and workloads employed are rapidly evolving, she said.
Linda said three prevailing trends in AI development are driving the transformation:
- The expansion of the General Purpose model. She says the training of these massive models continues. They're notable for possessing a wide range of potential uses and tasks.
- The integration of the Expert Model into enterprise applications. These distilled, smaller, and domain-specific models are embedded into various enterprise applications to deliver specialized capabilities and enhance functionality.
- The development of new model architectures. Research is ongoing to discover innovations that go beyond the transformer architecture used in large engine models.
As a result, enterprise companies require AI infrastructure to keep pace with the change.
Suppliers have responded to the shifting AI landscape by creating new types of infrastructures. Linda said that some infrastructures are being built to accommodate massive models like OpenAI’s ChatGPT-5, and soon featuring trillions of parameters. Others are designed for specialized hardware and developed to produce optimized performance for specific tasks, particularly for AI-inferencing workloads.
Additionally, the development of innovative compute architectures is ongoing. Linda said she expects quantum computing to one day play an important role.
"[Customers are] adopting a tailored approach and choosing hardware that aligns with their specific workload requirements, whether it's a right mix of CPU, GPU, storage or networking components," Linda said. "As AI continues to revolutionize every industry, Supermicro's goal is to accelerate our customers' AI initiatives, reduce time to market and fully realize the potential of their AI investments in a sustainable way."
Sustainability is a theme Linda returned to several times during the interview.
To illustrate the need for greater efficiency in all areas of AI development, Linda cited several eye-opening statistics: In the last two years, data center energy consumption has recorded double digit growth due to AI; By 2027, consumption could reach 1000 terawatts, twice the amount consumed in 2022.
Because of the mind-boggling amounts of data processed during AI training and development, large amounts of energy are needed to power computation. The process also generates a lot of heat as well as cost.
Linda said producing energy efficient servers and hardware is critical for protecting the environment and helping companies achieve the lowest possible TCO.
"Efficiency and sustainability are essential to Supermicro's solution approach," Linda said. "We prioritize energy efficiency through our server design in partnership with AMD, reducing both power consumption and heat output."
Linda applauded AMD's advancements in server processor technologies and noted that Supermicro has integrated AMD EPYC server processors and the AMD Instinct GPU family into Supermicro's HPC and AI accelerator solutions.
"Over the years, [Supermicro and AMD] have super-optimized the integration process, manufacturing supply chain, and our go to market strategy, so we deliver the best value to our customers together," Linda said. "AMD's benchmarks consistently show their leadership in performance per watt. When combined with Supermicro's liquid cooling technology, we create a sustainable infrastructure for AI, cloud computing, data science, data analytics and even more.
Together with AMD," Linda continued, "we can offer sustainable solutions to meet growing demand from data center infrastructure."
