Finance Associate

Pegatron to exhibit at OCP 2026, showcasing AI infrastructure strength 2026-10-13 13:28 Economic Daily News / Reporter Wu Kaizhong / Taipei real-time report Pegatron (4938) announced on the 13th that it will showcase the development and manufacturing capabilities of its next-generation AI infrastructure at OCP Global Summit 2026. Under the theme "Scaling innovation for the AI era," Pegatron will demonstrate how to transform advanced AI platforms into mass-producible, deployable systems, racks, and clusters, accelerating the introduction of AI infrastructure into data centers. As AI infrastructure expands from single servers to large-scale clusters, successful deployment depends not only on computing performance but also on system engineering design, liquid cooling technology, power architecture, open standards, production readiness, and rigorous validation capabilities. Through designs adopting OCP architecture standards, including ORv3 busbar, DC-MHS, and OCP 3.0 network interface cards, Pegatron builds modular and scalable infrastructure suitable for modern data centers. The core product exhibited this time is NVIDIA Vera Rubin NVL72. Pegatron's RA4803-72N3 is a rack-level AI platform integrating 72 NVIDIA Rubin GPUs, 36 NVIDIA Vera CPUs, sixth-generation NVIDIA NVLink, and a fully liquid-cooled rack-level integrated design. This platform demonstrates Pegatron's system design and manufacturing strength in complex rack-level AI solutions, as well as its integration capability to rapidly advance products from the design stage to cluster deployment. In addition to accelerated computing, Pegatron will also showcase the NVIDIA Vera CPU liquid cooling system and NVIDIA Groq 3 LPX, addressing the evolving infrastructure requirements of agentic AI. As a CPU purpose-built for AI agents, NVIDIA Vera can optimize tool calls, sandboxed code execution, and reinforcement learning loops required between model computation steps. In addition, Pegatron will also showcase AI-ready computing and storage infrastructure using AMD EPYC 9006 series processors. These include SS201-1A1, a 2U single-socket storage server supporting up to 32 E3.S NVMe SSDs, designed for high-throughput AI data storage requirements; and GS203-2A2 and GS205-2A2E, two 2U dual-socket platforms suitable for enterprise, cloud, high-performance computing (HPC), and data-intensive workloads. Pegatron General Manager and Co-CEO Jason Cheng said that the key metric for AI infrastructure is how efficiently it can move from engineering development to actual deployment. Pegatron combines deep system design expertise, vertically integrated manufacturing capabilities, OCP-compliant architecture, and pre-deployment rack-level validation capabilities to rapidly build and deploy AI servers, clusters, and data center infrastructure at scale. By linking digital twin simulation, thermal validation, automated manufacturing, and rack-level testing, we help customers accelerate AI infrastructure from design to mass production deployment.

KONST raises $30M Series B funding led by ADATA Technology to scale up AI data center buildout in Asia and advance its Token Factory strategy. Oct 09, 2026, 11:00 ET Funds will also advance KONST's Token Factory strategy as the company serves large enterprises through a three-layer model of compute supply, cloud delivery and AI governance TAIPEI, Oct. 9, 2026 /PRNewswire/ - KONST, a Taiwan-based next-generation AI compute operator, announced that it has raised $30 million Series B funding led by ADATA Technology (TPEx: 3260), with strategic participation from M Mobility, Pegatron Venture Capital, and several other corporate investors. KONST will use the proceeds to scale up its AI data center buildout across Asia and accelerate its Token Factory strategy, aiming to meet fast-growing enterprise demand for AI adoption. | Note: Investors are listed in alphabetical order by company name | Series B Round: Strategic Participation from Multiple Corporate Groups The round's principal investors come from memory and storage, electrification solutions, and electronics manufacturing, while other participants span energy, semiconductors, and thermal management - groups that are themselves on the demand side of enterprise AI adoption. KONST said that beyond the capital itself, it places even greater value on the synergies its investors offer across industry use cases and customer applications, which will help connect KONST's services with real-world deployments more quickly. "This round is about more than the amount raised. What matters most is that we and our investors share the same vision: AI must make its way into how enterprises actually operate, not stall at proof of concept," said Ben Chang, co-founder and chairman of KONST. "Once AI is part of operations, computing power is no longer a one-time project expense. It becomes essential infrastructure that enterprises rely on every day. Our investors understand this better than anyone, because they are on the demand side, too." From Data Centers to Usage Governance: KONST's Three Layers Turn Compute into a Service Enterprise demand for AI is shifting from trials by individual teams to routine use across departments, and compute demand is shifting with it, from one-off project procurement to continuous, fluctuating day-to-day consumption. Fixed-capacity provisioning is finding it increasingly difficult to keep pace with variable usage patterns. Meanwhile, as agentic AI matures, workloads are placing higher demands on computing performance, stability, and reliability. Demand is rising in both volume and quality at the same time, which is precisely why AI infrastructure has become so much more important. The Token Factory concept treats compute as production equipment and tokens as its output. KONST is building toward it in three layers: At the foundation is compute supply. Konstra AI takes projects from site selection and the buildout of mechanical, electrical, and environmental control systems through to GPU cluster tuning and acceptance, followed by ongoing operations and maintenance. Proceeds from this round will be used to expand its build capacity across Asia. In the middle is the conversion layer. Glows.ai breaks cluster compute into smaller units that enterprises access with per-second billing, paying only for what they use. At the top are governance and delivery. Horizon AI's (k) ATP Token handles this layer, determining who can use what, how much, and how it is charged. KONST believes this three-layer division of labor also shows how Taiwan's existing strengths in hardware and supply chains can extend further up the value chain - from manufacturing compute to delivering it. (k) ATP Token: One Key to Access Every Model and Manage Team Usage Governance is the most easily overlooked of the three layers, and it is also where enterprise AI adoption often stalls. The problem isn't that the models fall short; it's that no one can say clearly which department used which models, how much was spent, or what happened with each request. (k) ATP Token is designed around the principle of "One key. Every model." With a single project key, enterprises can connect to every authorized model, without having to negotiate contracts and build integrations with each vendor one by one. The platform's four-tier structure - organization, workspace, project and key - allocates quotas down the organizational hierarchy, sets model access at the project level and flags budgets before they are exceeded. The model and usage of every request are recorded in audit logs. For enterprises already operating at scale, (k) ATP Token goes a step further, bringing every model on the platform under one contract and one invoice, with custom contract pricing based on actual usage. KONST emphasized that (k) ATP Token is not a standalone product but the governance foundation built into every AI adoption project. From day one of a proof of concept, cost tracking, permissions, and request logs are already in place, so going live is simply a matter of scaling up traffic. "Buying GPUs is only the beginning," said Chang. "You need data centers you can build, compute you can break down finely, and usage you can keep under control. Only when all three layers are connected do you get what we call delivering compute." Focused on AI Adoption at Large Enterprises KONST said Horizon AI focuses on sizable business groups and large organizations. AI adoption at these customers typically involves multiple departments, integration with existing systems, and long-term governance requirements. What they need is not a point solution but a partner that can work alongside them through the entire adoption cycle. Horizon AI has completed deployments for several enterprises and institutions and is conducting needs assessments with large customers across multiple industries. As its business grows, KONST continues to expand its team, bringing in talent across a range of disciplines to meet increasingly diverse customer needs. To learn more about open positions, visit KONST's LinkedIn page: https://www.linkedin.com/company/konsttech About KONST KONST Co., LIMITED is a next-generation AI compute operator headquartered in Taipei, Taiwan. Through its One KONST brand and strategic partner ecosystem, KONST brings together the capabilities of three brands - Konstra AI, Glows.ai, and Horizon AI - to deliver the complete AI compute value chain as one integrated service. Konstra AI provides compute buildout services for AI data centers, covering end-to-end construction from mechanical, electrical, and environmental control systems to GPU clusters. It also provides equipment operations and maintenance, as well as compute delivery. Glows.ai provides an enterprise-grade GPU cloud platform with per-second billing. Horizon AI is dedicated to AI application adoption and model asset governance. Customers work with one team on one order, whether their needs sit at the compute, platform, or application layer. Building on Taiwan's engineering strength and supply-chain advantages, KONST combines capabilities in GPU cluster operations, data center operations and maintenance, cloud platforms, and AI services. Amid the global AI wave, it works with enterprises, developers, research institutions, and international partners to build compute supply that is available, reliable, and sustainably scalable. Media Contact KONST Co., LIMITED PR Contact: Jora Huang, Chief Public Relations Officer Email: [email protected] SOURCE KONST Co., LIMITED

Pegatron revenue hits record high, surpasses NT$200 billion... September surges 1.4-fold, growth tops domestic contract manufacturers 2026-10-09 01:16 Economic Daily News / Reporter Wu Kaizhong / Taipei report Pegatron (4938), benefiting from Apple's new iPhone 18 Pro series and the "dual engines" of AI servers, yesterday (the 8th) reported September consolidated revenue of NT$214.753 billion, a record monthly high, and a surge of more than 1.4-fold from August, the highest month-on-month increase among domestic contract manufacturers, also up 86.9% from a year earlier. Its performance this quarter remains promising and is expected to reach the year's single-quarter high. Pegatron's September performance was the first time it surpassed the NT$200 billion mark; third-quarter consolidated revenue was NT$389.073 billion, the strongest third quarter in history, up 41.8% quarter-on-quarter and 50.9% year-on-year; consolidated revenue for the first three quarters was NT$907.626 billion, up 13.8% year-on-year. Pegatron analyzed that in September, all three major product lines grew month-on-month. In addition to the communications business benefiting from the new models of a major customer, AI server growth was more obvious; among them, notebook shipments in September were 460,000 units, better than 400,000 to 450,000 units in August, with cumulative third-quarter shipments of about 1.3 million units. Institutional investors noted that Pegatron's September revenue soared due not only to a large contribution from AI servers, but also to the simultaneous boost from the launch of new iPhone 18 Pro series models. Based on the growth rate of Pegatron's September revenue, the server business now accounts for 10% of its revenue. Looking ahead, institutional investors said that as the iPhone 18 Pro series reaches its shipment peak this quarter and AI server momentum continues upward, Pegatron's performance this quarter is expected to climb to the year's single-quarter peak. As Pegatron pushes its AI server business, co-CEO Cheng Kuang-chih recently said that new orders from U.S. customers have begun shipping and will be reflected in September performance. In response to strong global demand, the team is accelerating capacity expansion, and AI servers may account for up to 10% of revenue this year. Market rumors said Pegatron received AI server orders from a major U.S. electric vehicle manufacturer. Cheng Kuang-chih responded that Pegatron's AI server progress this year is good, orders from U.S. customers have entered the shipment stage, mainly the L11 specification for complete rack systems. Cheng Kuang-chih said that Pegatron is currently shipping both Nvidia's GB200 servers and the new-generation GB300 servers. However, GB200 had a short market transition period due to complex initial design and high assembly difficulty; by contrast, GB300 is more mature, but due to total cost of ownership considerations, some customers still choose GB200 for its better cost-performance ratio. Cheng Kuang-chih believes that demand for AI servers remains very strong, and as more countries invest in sovereign AI, demand for hardware from enterprises continues to rise. Pegatron's AI server team currently has about 1,000 people and will continue to add personnel to related business teams, accelerating team expansion according to business scale. Pegatron previously said at an investor conference that this year's server business performance will exceed original expectations. Not only is it "very confident" of achieving the tenfold annual growth target, the growth rate may exceed tenfold. It estimates that within one to two years, servers may become one of the company's top three business divisions.