Sort Newest to Oldest
2026/2/26
As Moore’s Law approaches its physical limits, the bottleneck in chip scaling is no longer confined to transistors—it increasingly lies in power delivery. Traditional front-side power delivery networks (FSPDN) ultimately constraining performance and power efficiency.To sustain progress in advanced process nodes, TSMC, Intel, and Samsung have each moved forward with backside power delivery network (BSPDN) technologies. This article analyzes the fundamentals of backside power delivery, introduces TSMC’s Super Power Rail (SPR), and compares it with Intel’s PowerVia and Samsung’s BSPDN solutions. It further explores two critical steps in implementing backside power delivery—wafer thinning and reclaimed wafers—and examines how these technological shifts could reshape the market and supply chain.
# Stocks
# USA
# Taiwan
# Manufacturing Industry
# Editor's Pick
2026/2/10
Building on the AI-driven bull run in technology stocks in 2025, U.S. equities entered early 2026 amid record-high sentiment, even as doubts began to surface over the efficiency of capital spending by major tech companies. Alphabet, Microsoft, Meta, and Amazon have sharply expanded AI-related CapEx. Despite strong revenue and earnings performance, the market is no longer willing to buy into the growth narrative alone. At the same time, rapid progress by Anthropic and AI agents is reshaping the valuation framework of the SaaS software industry. This article examines how AI capital expenditure, cash flow pressure, and breakthroughs at the application layer are collectively influencing the valuation direction of U.S. technology stocks in early 2026.
# Service Industry
# Fundamental Analysis
2026/2/5
As AI and HPC chip sizes continue to expand rapidly, advanced packaging is facing multiple challenges, including area utilization efficiency, warpage control, and cost structure. Although traditional CoWoS technology has established a critical position in the high-performance computing market, the physical limitations of circular wafers and ABF substrates are gradually becoming more apparent, making panel-level packaging an inevitable next step. This article explains the development background and technical characteristics of CoPoS packaging, while also comparing the differences between CoPoS and earlier CoWoS and CoWoP technologies. By adopting square glass panels as the core of its CoPoS architecture, CoPoS addresses the needs of larger and more highly integrated packages through higher unit throughput and improved structural stability, while also driving a new wave of growth momentum across equipment makers and the broader supply chain.
2026/2/3
After missing the surge in AI computing demand and facing setbacks in process-node execution, Intel is positioning its 18A node as a pivotal turning point to re-enter the advanced-manufacturing race. With core technologies such as RibbonFET (GAA) transistors, PowerVia backside power delivery, and High-NA EUV lithography, 18A not only carries hopes for a revival of Intel’s in-house processor roadmap, but is also viewed as a key bargaining chip for Intel Foundry Services to win tier-one customers. This article outlines the major technical elements and applications of the 18A process, compares Intel’s approach with TSMC and Samsung in the 2nm landscape, and assesses the challenges Intel may face going forward.
2026/1/29
Uber’s joint debut of a Robotaxi with Lucid and Nuro at CES 2026 not only signals that autonomous driving is entering a new phase of commercialization, but also clearly illustrates Uber’s strategic shift from in-house development toward becoming a “Robotaxi ecosystem integrator.” This article examines how Uber is leveraging a highly asset-light partnership model to connect electric vehicle manufacturers, autonomous driving technology providers, and its global ride-hailing platform, building a hybrid network in which human drivers and Robotaxi coexist to further improve vehicle utilization and platform pricing power. It also explores how, amid continued expansion by competitors such as Waymo and Tesla, Uber can still maintain a critical position in the autonomous mobility landscape by capitalizing on its massive user base and cross-market dispatch capabilities.
2026/1/27
As subscriber growth slows and platform scale gradually approaches a ceiling, Netflix’s operational focus is shifting away from user expansion toward enhancing ARPU (subscriber base × average revenue per user) and rebuilding its content pricing power. This article focuses on the growth constraints currently facing Netflix and argues that, compared with advertising and AI monetization—both of which still carry unproven outcomes—acquiring globally recognizable, long-lived IP through M&A and extending its cross-media monetization lifecycle may represent a more certain strategic path. The article further analyzes Netflix’s proposed acquisition of Warner Bros. Discovery, highlighting the structural advantages of IP portfolios. At the same time, it examines the key uncertainties surrounding the transaction, including highly leveraged financing, subscriber overlap, and regulatory scrutiny.
# News
2026/1/19
This article focuses on the key industry under the 2026 AI outlook—memory. From the angles of supply contraction and shifts in demand structure, it analyzes how NAND, DRAM, and HBM—driven by AI data centers, high-performance computing, and enterprise storage—are moving into an industry cycle of broad-based undersupply. With manufacturers maintaining conservative capex, DDR4 entering EOL, the crowding-out effect from HBM intensifying, and AI-driven storage demand growing far beyond expectations, memory pricing has entered an upcycle since 2025 and is set to remain strong into 2026, making memory one of the most structurally advantaged core industries within AI infrastructure.
# Investment
# Financial Planning
# Investment Strategy
Against the backdrop of rapidly rising AI compute demand and continuously increasing chip TDP, the thermal management industry is entering a critical technological inflection point. This article is the third installment of fiisual’s 2026 Outlook Series, focusing on the core structural changes in AI server cooling architectures and providing an in-depth analysis of how liquid cooling—driven by GB300 GPUs and AI ASICs—is officially becoming the mainstream cooling solution. The article compares the performance differences between air cooling and liquid cooling, explains the rising value content of key liquid-cooling components such as cold plates, QD, and CDU, and explores the growth drivers, technology roadmaps, and investment opportunities within Taiwan’s supply chain for the thermal industry in 2026.
2026/1/15
Continuing from the previous article, as AI enters its fourth year of development, the market is shifting from optimistic expectations to a more rigorous evaluation of actual performance. In 2026, investment focus is no longer just about whether a company is "related to AI," but rather about who can deliver irreplaceable structural value amid the surge in inference demand and increasing scrutiny on capital efficiency. This article will focus on the ASIC industry, analyzing which segments of the supply chain are most likely to become key beneficiaries in the next phase as AI moves from training to inference. It will also highlight the roles that relevant Taiwanese companies are expected to play in this transformation.
2026/1/13
In 2026, the global economy is entering a new phase of “multi-speed and uneven” growth, with AI expected to remain a core engine driving macro momentum.In the U.S., consumption and employment continue to diverge. Sticky inflation and labor market restructuring are eroding the spending capacity of middle- and lower-income households, while credit risk is gradually surfacing. With limited room for rate cuts, market focus is shifting from valuation expansion to profitability and capital efficiency. AI investment is also moving from a “scale-first” to a “return-first” approach, with capital increasingly concentrated in players demonstrating sustainable, long-term competitiveness.
# Investment Analysis
# Technical Analysis
# Chip Analysis
