SKT is driving cooperation with Penguin Solutions, a leading American integrated AI data center solutions provider. At CES 2025 held in January, SKT signed a cooperation agreement with SK hynix and Penguin Solutions for joint R&D and business promotion of AI data center solutions [see article here]. In July of last year, SKT made its largest AI investment to date amounting to 200 million USD[see article here].
SKT Newsroom interviewed Mark Seamans, Global Marketing VP of Penguin Solutions, on the company’s business strategy, the outlook on AI infrastructure, and the significance of the cooperation with SKT. Mr. Seamans has deep technical mastery and leadership experience in designing and implementing complex enterprise infrastructures such as cloud computing, AI, big data storage, high-performance computing (HPC), and business process management. Drawing from his extensive technical experience, he is responsible for developing and driving Penguin Solutions’ global marketing strategy.
* This article contains the opinions of other companies and may differ from SK Telecom’s official position.
Infrastructure is key to AI growth, even more critical in enterprise markets adopting AI for competitive advantage
Mark Seamans, Global Marketing VP of Penguin Solutions
Q. The AI ecosystem is rapidly taking shape. How do you expect AI to change the industry in the future?
AI is driving fundamental change within virtually every industry based on its ability to drive efficiency and effectiveness across product and services segments. Within the product space, AI is advancing the way that researchers pursue the exploration of new breakthroughs ranging from drug discovery to personalized healthcare—and product teams are benefitting from both concept refinement and actual design acceleration.
Up to this point, many organizations have been focused on exploring AI capabilities, identifying potential use cases within their markets, and beginning to understand how and where to source the critical enterprise AI data that is required to train and finetune models that will support initial deployments. These initial steps tend to occur somewhat slowly at first, but they provide the foundation for accelerated deployments in the very near future. AI adoption is expected to evolve from this experimental stage to become a core business tool and we believe that enterprises will soon see end-to-end control of their own sovereign AI infrastructure as a competitive requirement.
For example, we are engaged with early-adopter customers in the financial services marketplace who began their AI journey when the technology first entered the spotlight. To date they have identified many use cases across their business where the technology can drive improvements that range from incremental to transformational.
Running AI training and inferencing processes on generic virtualized environments within the public cloud may not provide the level of performance, responsiveness, and control that enterprise IT leaders desire. As AI technology becomes a core corporate asset, enterprises will come to view bespoke AI infrastructure as a key strategic advantage. However, many aspects of AI infrastructure deployments—particularly when done at scale—present new challenges for enterprise IT teams given that many of the hardware, software, and process elements are new and different.
Given these trends and potential business impacts, we expect that AI will change the enterprise IT infrastructure delivery industry by opening a significant new opportunity for specialists in high-performance computing, accelerated networking, and AI-specific system management. This initial phase will focus on enabling enterprises to deploy their near-term core infrastructure, but the pace of change and innovation within the marketplace will also create opportunities for solution providers such as Penguin Solutions that can continually consume the latest capabilities and translate them into technology advantages for our customers.
Q. What role do you think AI infrastructure will play in this ecosystem? Specifically, how do you think the size or trends of the AI DC market, which will play a key role, will change?
Infrastructure is at the heart of AI growth—particularly within the enterprise market, where corporations will look to AI for differentiation and competitive advantage. If we think of AI as a car, then infrastructure is the engine and transmission that powers the driving experience.
At Penguin Solutions, we have consistently spoken about AI emerging in a series of “waves” as different market segments invest in the technology over time.
The first wave involves large-scale foundational model development, with companies making large investments in AI training systems to build out the core models that enable a new level of AI capability. Penguin Solutions has worked with leaders such as Meta to participate in the design, build, deployment, and ongoing management of the clusters that have been used to train major industry AI models such as Llama. These deployments can be very large and Penguin Solutions has significant experience in how to optimize deployments at scale.
However, most enterprises will not need this massive level of infrastructure. The second wave of AI adoption is about enterprise-level deployments and enabling organizations to build on models developed during the first wave to create application-optimized models for specific use cases. These models can be trained and finetuned with mid-sized or large-scale infrastructure, using insights from specific enterprise data to power automation and more efficient business operations. The design, build, deployment, and management of these customer-specific solutions is a specialty area for Penguin Solutions.
The third wave of adoption focuses on AI inference at scale, leveraging models from the first two waves – making every AI-driven organization a part of Wave 3 at some level. Some use cases will require modest inferencing configurations. Others will involve massive, mission-critical“Mission critical” refers to a characteristic or condition that, if not functioning properly or destroyed, could have a fatal impact on the entire business operation and cause disastrous impact to the organization or society. This includes the telecommunication system of an online business company, a disaster communication network, and an aircraft operation control system. systems. The system architecture, processor, and network options for inference solutions are far more varied in this wave. Customers may elect to deploy some systems at “the edge” close to where data generation is occurring.
Helping customers understand and navigate this broad set of AI infrastructure decisions is a core part of Penguin Solutions’ value proposition to its customers. Our goal is to become a trusted advisor on AI infrastructure and long-term partner for our customers so that they can leverage AI for their competitive advantage.
Penguin Solutions participated in the global big tech exhibition at the ‘SK AI Summit 2024’ held on November 4 and 5, 2024. Penguin Solutions’ exhibition booth
Penguin Solutions participated in the global big tech exhibition at the ‘SK AI Summit 2024’ held on November 4 and 5, 2024. Penguin Solutions’ exhibition booth
Q. Penguin Solutions, Inc. is expanding its business into the field of AI Factory. What made you decide to move into the space? And what is the role of the AI Factory business in Penguin Solutions’ portfolio?
Penguin has been focused on the design, delivery, and management of high-performance computing solutions for over 25 years. Many of these solutions have been very large and custom-made for organizations and researchers working across a variety of domains. Use cases have ranged from wind tunnel simulation for the design of efficient cars to weather forecasting and climate simulation to seismic data analysis for oil and gas exploration.
In 2017, Penguin Solutions did its first work on a large-scale GPU-accelerated cluster that was targeted for use in AI model training. After successfully completing this effort and going on to work on several additional AI training configurations, it became clear that Penguin Solutions could bring significant value and repeatability to the AI market by packaging a solution architecture for AI factories that could be efficiently purposed, deployed, and managed for our customers.
Based on this realization, Penguin Solutions began work on its “OriginAI” solution platform, which provides an end-to-end infrastructure environment for customers at all scales of deployment. When it comes to “end-to-end”, Penguin Solutions truly provides a complete and managed experience that includes all the hardware, GPUs, networks, storage, software, and expert Penguin services needed to enable rapid implementation and seamless management of the new AI solution.
With its OriginAI platform, Penguin Solutions solves the complexity of AI infrastructure for our customers so that they can focus on their core business. For this reason, OriginAI has become a core part of our portfolio and its frequently the starting point for our conversations with new customers.
Q. Recently, you have proven your capabilities by winning the AI cluster project for Voltage Park. What are the differentiated advantages of Penguin Solutions for AI data center integration solutions?
Penguin Solutions’ relationship with Voltage ParkNext-generation GPU cloud service provider based in the US is a very strong example of how we can be both creative and adaptive in our work with customers. Voltage Park had already procured a large set of hardware infrastructure components—including approximately 24,000 of the latest NVIDIA GPUs—but they needed a partner who could work with them to transform that powerful set of components into a production-grade GPU-as-a-Service cluster spanning four separate data center locations. They needed to get to production quickly and deliver an outstanding experience for their users.
After a series of visits to each of the sites, Penguin Solutions proposed an approach that combined our differentiated ClusterWare software platform with comprehensive end-to-end services for both system deployment and production.
The collaboration between Penguin Solutions and Voltage Park has been very successful because of our experience and capabilities, but also because we view this relationship as much more than just a standard deployment project. We continue to deliver operational services daily to ensure that Voltage Park customers have an outstanding AI infrastructure experience.
Penguin Solutions’ ClusterWare was originally developed as a solution over 20 years ago as the foundation for powerful compute clusters at NASA (U.S. National Aeronautics and Space Administration). It was later acquired by Penguin Solutions to serve as the basis for our commercial HPC cluster platform offerings. In fact, Penguin Solutions was inaugurated into the United States Space Technology Hall of Fame for our work over the years in delivering advanced compute technology, and Penguin Solutions’ ClusterWare played a big part in that. Over the years, the platform has been significantly enhanced by our software engineering team, including the addition of many AI- and GPU-specific features.
From a services perspective, the Penguin Solutions team differentiates itself from others through our focus on and experience with high-performance computing (HPC) and AI clusters. Rather than delivering generic IT services, our team has deep capabilities and proven processes for GPU-based deployments, NVIDIA networking technologies, parallel file system storage, and proactive system management.
Q. There is a forecast that AI infrastructure investment will decline due to concerns about the current AI bubble. What do you think about this?
While no one can predict the future, Penguin Solutions is not seeing any decline in customer excitement about and interest in AI. Our firm belief is that we are in the very early days of AI adoption and that there is a robust future ahead for the design, build, deployment and management of AI infrastructure.
As we briefly discussed earlier, AI is happening in waves. We are just now entering the second wave, where enterprise customers will build on the significant work of foundational models such as GPT, Llama, and others. In the months ahead, we expect that advances in sovereign AI models from specific countries and locales will be tuned for target users, use cases, and regulatory frameworks.
For every customer use case in these early waves of adoption, there will be a need to deploy inference clusters on which the production AI models will run. Just one year ago, discussions with customers around the design and delivery of AI inference clusters occurred infrequently—now they happen every day at Penguin Solutions as customers look ahead to production deployments. The inference environment also opens opportunities to leverage a broader set of technologies such as the Rebellions NPU processors built specifically for AI inference. These additional choices create new opportunities for Penguin Solutions to deliver value and provide expert advice to our customers as they navigate the complexity of AI.
In short, as the shape and approach to AI usage continues to evolve rapidly in the coming years, it seems clear that virtually every enterprise will be looking for ways to leverage AI to increase their business efficiency and differentiation. This in turn will bring more customer investments in AI—and more opportunities for Penguin Solution.
Cooperation with SKT to create broad and strategic value… Maintains strong communication and partnership
Justin Kim, AI Infrastructure President of SK hynix (left), Ryu Young-sang, CEO of SK Telecom (center), and Mark Adams, President and CEO of Penguin Solutions, pose for a commemorative photo during the agreement ceremony at CES 2025 in Las Vegas on Jan. 9, 2025.
Q. What is the significance and expected effect of this cooperation with SKT?
Penguin Solutions’ collaboration with SKT spans numerous areas of the AI spectrum. We have formal teams and workstreams focused on each of these elements.
Within the software workstream, both Penguin Solutions and SKT bring powerful capabilities around AI cluster deployment and management. These capabilities fit together nicely as part of a more complete AI infrastructure software stack that can transform bare metal AI hardware into powerful AI compute clusters capable of running customer workloads.
For the AI processor workstream, Penguin Solutions’ expertise in the development and delivery of AI compute servers blends well with SKT’s investments in Rebellions creation of advanced accelerator cards for AI inference use cases.
As the size and complexity of AI models continues to increase, memory plays an increasingly important role in AI system architectures. The SKT and Penguin Solutions memory workstream is focused on identifying where our SMART Modular division can expand its already strong relationship with SKT hynix to bring innovative, cost-efficient solutions to customers.
Finally, the globalization workstream is focused on business collaboration planning and identifying where Penguin Solutions and SKT can execute and grow given SKT’s global presence and Penguin Solutions’ strong brand and customer base in the U.S. marketplace.
We see our collaboration with SK as having broad and strategic value for both organizations. Communications between the two firms and the spirit of the partnership have been very strong.
Q. What is your assessment of SKT’s approach to the AI industry?
SKT’s commitment to the AI industry is clearly going to have benefits for both customers and for Penguin Solutions. We look forward to continued collaboration as we pursue new areas for growth.
From a vision perspective, SKT has been clear that it views AI as a transformational technology that will be at the core of its product offerings and customer value proposition. Even the core messaging around “Building the AI Infra Super-Highway” positions SKT’s core strengths in network and communications as a foundation for AI workloads and use cases.
SKT is making exciting investments in technology, including its work on the AI Infra software stack. These advances are focused on adding customer-driven capabilities such as more efficient AI cluster utilization, multi-tenant cloud environments, and sovereign AI models. These can all be differentiators as customers evaluate partners and potential technical directions.
Finally, SKT’s approach is squarely focused on its customers and on finding ways to deliver AI-related value as a partner of their journey. This focus on customer value and long-term customer relationships offers a foundation for consistent revenue growth even as the AI market evolves over time with the introduction of new technologies and breakthrough capabilities.
