After announcing its foray into the LLM (Large Language Model) IP market, Skymizer recently unveiled a series of hardware and software solutions centered around LLM IP.These offerings are designed to bring more imagination to the LLM application service market. Skymizer's IP solution series is codenamed EdgeThought, with the first market-ready solution named ET2, capable of efficiently handling all current edge devices requiring LLM, including the recently released Llama3 with a parameter scale of up to 8 billion.Before introducing the IP solution, Skymizer's primary solutions were in the compiler space, bridging the gap between chips and software. This background has endowed the company with extensive experience in overall system hardware and software integration and optimization. As the demand for LLM rises, Skymizer, leveraging its solid market foundation, has entered the IP market to cater to diverse vertical application needs.Launch of integrated hardware-software platform: ET2 features edge computing, LLM, and AI inferenceAccording to Skymizer Executive Vice President William Wei, ET2 encompasses three key elements: edge computing, LLM, and AI inference. Besides accommodating various LLMs in the market, ET2 can flexibly expand computing resources to meet client needs.If the parameter scale of the LLM to be processed is too large, expansion can achieve the required computing power, naturally increasing memory capacity and power consumption. In addition to the existing IP solutions, Skymizer also launched the SkyGenie SDK (Software Development Kit).The SDK can address various categories of LLMs, including general, domain-specific, and private, assisting various industry applications. This enables software developers to create corresponding applications based on different LLM types, optimizing overall system performance. During COMPUTEX 2024, Skymizer will further demonstrate applications such as smart factories' Autonomous Mobile Robots (AMRs), Drive-thru ordering smart assistants, and smart automotive scenarios using ET2 and other hardware-software solutions.Wei emphasized that Skymizer's comprehensive hardware-software system development experience allows the broad tech industry ecosystem to benefit from Skymizer's complete platform solutions. He revealed that while the market currently sees edge AI GPUs performing at about 20 tokens per second, Skymizer's ET2 tests show around 32 tokens per second, with implementation costs at just 1/100 of Edge AI GPUs. This high cost-performance ratio makes ET2 an ideal choice for cost-sensitive end applications.First Chip with ET2 solution to debut at CES 2025 In the semiconductor domain, Wei candidly shared that Skymizer is open-minded and actively collaborating with domestic and international design services and IP firms. He also revealed that ET2 is highly expandable, from small IoT MCUs to high-performance Edge Servers.When paired with higher bandwidth memory interfaces, it can function as a server-level inference engine for multi-user, multi-batch processing. The first chip adopting ET2 is expected to debut at CES 2025, positioning ET2 as a game-changer for edge device LLM inference.ConclusionSkymizer's launch of its first IP solution, ET2, marks a new milestone for the company. With rich hardware-software integration experience, Skymizer not only provides high-performance, cost-effective edge computing solutions but also equips developers with powerful tools through the SkyGenie SDK to meet diverse LLM application needs.These innovations enhance the performance of different vertical market applications and create new possibilities for smart factories, autonomous mobile robots, smart automotive, and more. Skymizer's comprehensive platform solutions are poised to be a significant driving force in the future LLM application market.Skymizer not only provides LLM silicon intellectual property solutions but also offers the SkyGenie SDK, supporting various types of LLMs. This makes it easier and faster for AI application developers to create applications, enabling hardware chip partners to achieve higher integration and better meet market demands

The applications of AI are becoming increasingly widespread in our daily lives. With the continuous iteration of generative AI and large language models, various innovative applications are emerging. These include, such as smart home appliances, autonomous vehicles, and robots to VR/AR applications.These edge devices are equipped with AI processors that compute on the device directly, further accelerating the launch of innovative Edge AI applications. However, these devices also face several challenges, including excessive GPU power consumption and high costs limiting the widespread adoption of Edge AI products.DEEPX won three CES Innovation Awards in 2024Photo: DEEPXDEEPX, an AI chip startup from Korea, won three CES Innovation Awards 2024 in January for its unique AI chip ultra-gap source technology. The awards were in the categories of Computer Hardware, Embedded Technology, and Robotics, and Computer Hardware & components.Their NPU processor boasts low power consumption and cost-effectiveness and addresses the issue of insufficient accuracy found in existing NPUs on the market. This technology was recognized by lots of semiconductor companies during CES 2024 and is scheduled for mass market release in late 2024.DEEPX CEO Lokwon KimPhoto: DEEPXLearn from ARM to build DEEPX into a leading On-device AI company"The main battlegrounds in the AI era will move to the 'edge'. Just as ARM dominated the CPU market with smartphones, the semiconductor company that dominates the edge will dominate the AI market," DEEPX CEO Lokwon Kim said.Kim, a former senior researcher in Apple's Application Processor (AP) design, returned to South Korea to establish DEEPX after gaining extensive experience in semiconductor design at renowned companies like Broadcom, Cisco, and IBM T.J Watson. His goal is to build DEEPX into a leading AI company in the era of on-device AI, reminiscent of how ARM revolutionized the CPU market with its low-power technology.On-device AI, which processes information within a mobile device without connecting to a server or cloud, is a burgeoning field. ARM broke Intel's dominance in the CPU market with its efficient, low-power processors, which are now prevalent in smartphones and expanding into PCs and servers. Kim aspires for DEEPX to have a similar impact on the AI semiconductor industry.Kim identified a weakness in Korea's semiconductor ecosystem, particularly in system semiconductors. Drawing inspiration from Morris Chang, the founder of TSMC, who returned to Taiwan to establish a leading foundry after learning from the American semiconductor giant TI, Kim saw an opportunity to address these gaps in the South Korean market.Chang predicted a demand-driven semiconductor market, leading to the creation of TSMC, which produces semiconductors on consignment. Similarly, Kim believes that the opening of the AI semiconductor market offers a chance for innovation and growth.Democratizing AI TechnologyKim opposes the monopolization of AI semiconductor technology through strategic investments (SIs) that often lead to mergers, acquisitions, or alliances, resulting in technology being controlled by a single entity. His experience at Apple motivated him to advocate for more democratic and accessible AI semiconductor technology. He believes that AI semiconductors should be universally available and not restricted to proprietary use, emphasizing the importance of technology that everyone can utilize.To pursue this vision, Kim established DEEPX in Pangyo, South Korea's hub for fabless companies. He aims to create an AI semiconductor company with a unique value, promoting innovation over exclusivity.DEEPX is positioned not as a competitor to global semiconductor giants but as a complementary force, enhancing the global semiconductor landscape by providing accessible and advanced AI technology.DEEPX aims to be the world leader in on-device AIPhoto: DEEPXAnalyzing the Difference Between NPUs and GPUs: DEEPX's Unique SolutionThe necessity for developing specialized AI semiconductors, such as NPUs, stems from the limitations of GPUs, which have traditionally been used for AI computations. GPUs, originally designed to process graphic data, excel at handling large amounts of data simultaneously, making them suitable for AI learning tasks.However, their high power consumption and operational costs present significant drawbacks. This makes them less ideal for "edge AI," which involves running AI applications directly on devices like controllers, robots, and self-driving cars - collectively referred to as the "edge."NPUs (Neural Processing Units) are modeled after the human brain, offering the benefits of lower power consumption and reduced production costs. However, existing NPUs have struggled with accuracy and support for the latest AI algorithms. DEEPX stands out among fabless companies by addressing the core challenges of AI semiconductors in one comprehensive solution.Unlike typical NPU vendors that release a single chip, DEEPX recognizes that different electronic devices require varying levels of semiconductor capabilities. For instance, AI for closed-circuit television (CCTV) primarily needs to analyze video, whereas AI for robots involves far more complex computations.To address this, DEEPX has developed from low-end to high-end performance four chips at once: one that can connect a single electronic device for AI computations, and another that can link three or four devices for broader AI tasks. This universality across devices is a key reason DEEPX received the Innovation Award at CES 2024.DEEPX's commitment to high performance at lower power and cost also earned them the Innovation Award. The company's latest AI algorithm, Yolo7, runs on their semiconductor DX-V1, produced using Samsung's 28nm process.This algorithm was previously incompatible with conventional NPUs. In addition, the DX-M1 chip boasts a design area one-third the size of other NPUs, and its manufacturing cost is similarly reduced by one-third. Combining low unit costs with high performance in a low-power NPU, DEEPX's products are poised to lead the AI semiconductor market.NPUs are categorized into data center-based NPUs, which handle large-scale inference, and edge-type NPUs, designed for use in electronic devices such as robots, smart cameras, smart factories, consumer electronics, etc. DEEPX targets both data center and edge NPU markets. DEEPX's NPUs overcome the common shortcomings of existing NPUs by providing high accuracy and efficiency.DEEPX's innovation lies in creating NPUs that are not only small and cost-effective but also achieve accuracy comparable to, or even better than, GPUs. This success is attributed to DEEPX's pioneering work in two core technologies: IQ8 (an INT8 model compression technology), and Smart Memory Access(minimizes D-RAM usage). DEEPX leads the market for low-power AI solutions, achieving the world's highest power-to-performance ratio through proprietary advancements in hardware and software optimization.Advancing Towards Mass ProductionDEEPX has successfully demonstrated its original technology with sample units and is in the final stages of preparing its mass-production chip. Scheduled for market release in late 2024, the widespread adoption of products featuring DEEPX's chips could establish the company as a technology leader in the on-device AI market by 2025.The essence of on-device AI lies in low power consumption and the seamless integration of hardware and software. Since AI must operate on small devices, minimizing power consumption is crucial while maximizing AI performance within limited computing power.Kim, drawing from his experience at Apple, emphasizes the importance of designing hardware and software with equal priority from the outset. Unlike Apple, which develops its own devices and services, DEEPX engaged with approximately 700 customers during product development to understand their needs and find the optimal development point. This customer-centric approach has been pivotal in refining DEEPX's products.Four companies from Israel and the U.S., including DEEPX, are vying for dominance in the emerging on-device AI market. As the market begins to flourish beyond the server market, this competition is crucial for setting future momentum.Kim is confident in DEEPX's strategy. He notes that while some competitors prioritize rapid product releases, DEEPX focuses on price and optimization first. Given the rapid evolution of AI application services, Kim believes it is more important to align with market trends and offer high-performance features at a competitive cost rather than rushing products to market.DEEPX's innovative approach to developing versatile, efficient, and cost-effective NPUs positions the company as a formidable contender in the on-device AI market. By prioritizing customer needs and maintaining a balanced focus on hardware and software optimization, DEEPX is set to lead the industry with cutting-edge AI semiconductor solutions.DEEPX CEO Lokwon KimPhoto: DEEPXDEEPX Company's Philosophy: Valuing Technology Over Short-term ResultsDEEPX is not focused on making AI semiconductors for autonomous cars or smartphones. Instead, DEEPX aims to advance the integration of AI into everyday life. The company's products are designed to bring AI to areas such as CCTV and robots, pushing these technologies beyond outdated algorithms. DEEPX's mission is to pioneer advancements that should have already been made in the AI semiconductor industry.At the entrance of DEEPX, there is a note to employees that emphasizes the value of technology over monetary gains. It quotes Carl Sagan's "pale blue dot," reflecting the idea that life should be lived with value rather than chasing money and power. The message encourages the 70 employees to love their work and find it meaningful, aligning with the company's ethos of creating technology for the greater good.DEEPX believes in 'technology that everyone shares.' From a management perspective, AI is seen as one of humanity's final inventions, marking the endpoint of human evolution. The goal is not to monopolize technology or chase profits but to lead the way in making AI accessible and beneficial for all. DEEPX's AI semiconductors are the cornerstone of this vision, enabling widespread adoption of AI.Future Plan: Global Expansion and InnovationDEEPX's visionary CEO, Kim, has outlined a strategic plan for the future:- Global Market Entry: Starting from the second half of this year, DEEPX plans to aggressively enter the global market with its first-generation product, consisting of four AI chips. This move is set to usher in the era of "AI Everywhere."- Technological Innovation: DEEPX aims to develop new technologies that enable super-scale AI services with power consumption of less than 5W. This innovation will make advanced AI technologies more accessible and practical for widespread use.- Leadership in AI: DEEPX is committed to becoming a leading comprehensive AI chip company globally. By focusing on power and cost efficiency, the company seeks to provide core technologies that transition giant AI advancements from the realm of science to everyday applications.DEEPX's mission is to integrate AI into everyday life, advancing technology in meaningful ways. By valuing innovation over profit and aiming for global leadership, DEEPX is set to play a pivotal role in the future of AI. The company's commitment to shared technology and global expansion highlights its dedication to making AI accessible and beneficial for humanity.Join DEEPX's Early Engagement Customer Program (EECP), and don't miss out on their innovative products designed to enhance your AI capabilities. Discover how over 100 global companies, including Hyundai Kia Motors Robotics Lab, POSCO DX, Supermicro, and Dell, leverage their hardware and software to power their next-generation AI products.For more information, you can follow DEEPX on social media or visit their official website.

SK Hynix held a press conference titled "Memory, the Power of AI" on the sidelines of CES 2024 in Las Vegas where its CEO Kwak Noh-Jung laid out the company's vision in the AI era.At the press conference, attended by both domestic and international media, Kwak said that the importance of memory will grow further as generative AI becomes widespread. He also said that "SK Hynix is providing products from the world's best technologies to the ICT industry, leading the "Memory-Centric AI Everywhere."Kwak laid out the company's plan to introduce the "Custom Memory Platform" to provide customized AI memory solutions as demands for diverse memory products are growing.Below are the key points mentioned during the press conference.The ICT industry evolved dramatically through the PC, mobile, and now cloud-based AI era. Throughout, data in diverse types and massive amounts are being generated and communicated.Now we enter a new era of AGI built on all that data amassed. The new era will thus move towards a market where AGI constantly generates data and repeats learning and evolution.In the AGI era, memory will play a pivotal role in processing data.The role of memory is even more crucial from a computing system perspective. Before, systems were basically an iteration of dataflows from the CPU to memory and then back to the CPU in a sequential manner, but such a structure is not suitable to handle the massive data generated through AI.Now, AI systems are connecting large counts of AI chips and memory in a parallel fashion to accelerate massive data processing. This means that AI system performance depends on stronger and faster memory.The direction for memory in the AI era should be handling data at the fastest speed, in the most effective way, and in higher volume. That aligns with the past century of memory development that has improved density, speed, and bandwidth.The memory-centric AGI era has a clear leader: SK Hynix. We are providing diverse products with ultra-high performance such as HBM3 and HBM3E, the world's best and widest sought-after products, to markets and industries, TSV DIMM, the industry's largest-capacity server memory, LPDDR5 Turbo, the world's fastest mobile memory, and DIMM, the best-in-class performance server.SK Hynix is leading the "Memory-Centric AI Everywhere" in various industries ranging from AGI, conventional data centers, mobile, and PC systems.Building on our lineup, we will deliver a bandwidth-focused HBM4 and HBM4E, a power-improved LPCAMM, a capacity-expanding CXL and QLC storage, as well as compute-capable PIM and hone our technological leadership in the AI era.SK Hynix's AI memory solutions are delivering customers a new experience.Yet, AI systems are advancing at a breathtaking pace, leading to customer demands for diverse forms of memory performance.While some customers require ever larger capacities or more power-efficient products, others want higher bandwidth and the addition of data computing capabilities.So, we are planning to launch SK Hynix's proprietary "Custom Memory Platform."We will bring to AI systems our memory-specialized technology and R&D expertise for optimal convergence with customers' needs.This platform will transcend conventional ways and deliver an entirely new value proposition, catering to every customer with the most optimal memory solution for their needs.SK Hynix is preparing over 4.15 million square meters of a new memory manufacturing base worth over 120 trillion won (92.7 billion USD) investment.This massive capacity will allow us to continue serving existing customers while responding to the explosive growth of AI memory demand with the world's best products delivered in time.Today, I took the opportunity to share SK Hynix's readiness for a total AI memory provider with a clear plan for technology, customers, and production base.Please join us in the new AGI future we shall open and expand with you.SK Hynix CEO Kwak Noh-Jung at the Press Conference