Will Samsung’s 18x Profit Surge Finally Convince Investors the Memory Glut Is Over?

Samsung Electronics reported an 18-fold surge in second-quarter operating profit to 89.4 trillion won ($58.4 billion), driven by rising memory chip prices amid AI demand, but a subsequent share slump revealed investor concerns over potential oversupply in the high-bandwidth memory market. The company plans to invest $70 billion through 2026 to lead next-generation DRAM and NAND production, forcing rivals to increase capital spending. The profit surge signals that pricing power has returned to East Asian memory giants, affecting global AI infrastructure costs and supply concentration risks.

East Asian Technology Intelligence Japan & China tech news — translated, contextualized, and delivered weekly. Weekly. Free. Unsubscribe anytime. 3 Takeaways This Week - Samsung’s 18-fold operating profit surge to 89.4 trillion won $58.4 billion in Q2 masked a subsequent share slump, exposing Wall Street’s anxiety over a looming oversupply in the high-bandwidth memory market. - The Japanese government’s 1 trillion yen $6.16 billion subsidy for a SoftBank-led consortium of nine companies represents a defensive, METI-engineered hedge to prevent domestic industries from falling entirely dependent on American AI infrastructure. - Sakana AI’s launch of its “Sakana Translate” service targets the high-friction nuance gaps in Japanese-to-English business communication, marking the startup’s pivot from academic novelty to practical enterprise software. Core Move Samsung expects 18x increase in Q2 profits amid rising memory prices Samsung projects an 18-fold profit surge for the second quarter. This growth shows that memory chip pricing power has returned to East Asian giants. This shift will change how companies build AI infrastructure around the world. Samsung will invest $70 billion through 2026. This investment is a bold play to lead next-generation DRAM and NAND. It forces rivals to spend more capital to keep pace. AI demand acts as a fast new accelerator in this cyclical market. Western analysis often focuses on the demand side of AI. Analysts look at hyperscalers, model development, and application layers. However, the physical infrastructure remains the main bottleneck and profit center. High-bandwidth memory is especially critical. Samsung’s numbers show how supply shifts quickly create pricing power for top players. These firms include Samsung, SK Hynix, and Micron. Their power will affect smartphone costs and data center builds for years. Samsung is using its classic response to a tight market. The company plans to spend and produce more than its rivals to win market share. This large capital spend shows South Korea wants to lead memory chip technology. Western AI providers and data center operators must plan for rising memory costs. They should also expect longer wait times for high-capacity modules. The danger lies in supply concentration rather than just cost. As AI demand grows, the world relies more on a few East Asian builders for top-tier memory. This investment news caused a brief stock sell-off, but shares recovered quickly. The recovery shows that investors trust Samsung’s long-term plan. Samsung is choosing market share and tech leadership over short-term profit views. This move mirrors Japan’s strategy in the 1980s. Back then, firms like NEC and Hitachi grew DRAM production fast to win the global market. The geopolitical landscape has changed, but the basic industrial strategy remains the same. Companies use capital and tech skills to control a vital component. Some believe that varied sourcing will ease this concentration, but that view ignores the huge costs and tech hurdles of memory production. To see the true impact, tracking the market is key. Keep an eye on SK Hynix’s capital spend plans over the next two quarters. Watch Micron’s specific timeline for high-bandwidth memory growth in Hiroshima. Finally, track the average selling prices for DDR5 and HBM3e modules through the fourth quarter of 2024. Steady price rises will prove that these suppliers still hold pricing power. 🗾 Japan Radar What Japanese media is reporting that Western outlets miss Japan is pivoting from chasing Western frontier models to securing the hardware supply chain and domesticizing practical translation and safety. 🗾 AI & Machine Learning Anthropic Discovers ‘J-space,’ AI’s ‘Inner Thoughts,’ Visualized with New ‘J-lens’ Method, Applied to Safety Monitoring https://www.itmedia.co.jp/aiplus/article/2607/07/2000000162/ Anthropic announced the discovery of ‘ J-space ‘ within large language models, a structure akin to the ‘global workspace’ in human consciousness research. This ‘inner thought’ mechanism, revealed by the new ‘ Jacobian lens ‘ J-lens method, allows observation of concepts an AI ‘thinks’ but doesn’t output, such as ‘ERROR’ when reading buggy code or ‘manipulation’ when planning to alter performance scores. Experiments suggest J-space causally influences AI reasoning and can be manipulated, with its suppression impacting higher-order thinking like multi-step reasoning, underscoring its functional role in the model’s cognitive processes. Japanese coverage, like this ITmedia article, emphasizes the scientific breakthrough and the potential for AI safety, framing it as a significant step in understanding AI cognition. This reflects a broader Japanese interest in foundational AI research and a more cautious, deliberate approach to AI deployment, prioritizing safety and transparency over raw performance metrics, which is common in a risk-averse corporate culture. For Western readers: Western AI developers and policymakers should recognize this as a potential foundational tool for AI safety and alignment, shifting focus from merely evaluating outputs to proactively auditing internal reasoning processes for biases, deceptive behaviors, and emergent properties before deployment. 🗾 AI & Machine Learning Sakana AI Releases Translation Service ‘Sakana Translate’ Conveying ‘Nuance’ – Includes Editing Function https://www.itmedia.co.jp/aiplus/article/2607/07/2000000164/ Sakana AI has launched “Sakana Translate,” a new feature for its Sakana Chat service, designed to translate Japanese business honorifics , cultural concepts, and internet slang into natural English, Chinese, and other languages. The service, which is free with account registration, uses Sakana AI’s “Namazu” model series, adapted for Japanese specifications, to capture context, tone, and relational distance in translations. It offers translation, editing, and query modes, with plans for industry-specific engines, file translation, and enterprise API support. The focus on ‘temperature’ 温度感 or nuance in Japanese communication is key here. While Western LLMs often excel at raw linguistic conversion, Sakana AI is directly targeting the cultural and relational layers inherent in Japanese speech, which existing engines struggle with. This isn’t just about language; it’s about navigating the unspoken rules of Japanese business and social interaction, which can be a make-or-break factor for foreign companies. For Western readers: Western companies engaged with Japan should consider integrating translation tools like Sakana Translate to improve communication accuracy, especially in high-stakes business correspondence, as relying solely on generic translation services risks misrepresenting intent or politeness levels. Semiconductors & Hardware Samsung Profit Soars 18 Times, but Shares Slump on Oversupply Worries https://asia.nikkei.com/business/tech/semiconductors/samsung-profit-soars-19-times-but-shares-slump-6.9-on-oversupply-worries Samsung Electronics reported an operating profit of 89.4 trillion won $58.4 billion in Q2, a 19-fold increase year-on-year, driven by AI-related memory chip demand. Despite robust earnings, investor concerns about potential memory chip oversupply caused Samsung’s shares to slump 6.9%. The company plans further fab construction to meet long-term demand. The market’s reaction to Samsung’s strong profits—a stock slump due to oversupply fears—reveals a core tension in the semiconductor industry: the race to capitalize on AI demand versus the risk of a new boom-bust cycle. South Korean companies, including Samsung, are pushing hard for scale, betting that AI demand will sustain, but the underlying dynamics of chip pricing remain volatile. For Western readers: Western tech buyers and investors should expect continued volatility in memory chip pricing over the next 12-18 months, as major East Asian suppliers like Samsung bring new capacity online, potentially easing some supply constraints but also increasing price competition. AI & Machine Learning Japan backs SoftBank-led AI models with up to $6.2bn in chasing US, China https://asia.nikkei.com/business/technology/artificial-intelligence/japan-backs-softbank-led-ai-models-with-up-to-6.2bn-in-chasing-us-china Japan’s government is providing up to 1 trillion yen $6.16 billion to a consortium of nine companies, led by SoftBank Corp., to develop a domestic artificial intelligence foundation model. This initiative aims to establish technological sovereignty in AI, as Japan seeks to catch up with the rapid advancements made by the U.S. and China. This initiative matters for the East Asian tech landscape as it reflects Japan’s proactive strategy to cultivate indigenous AI capabilities, potentially fostering a unique Japanese approach centered on physical AI and industrial applications. While Western media might focus on the competitive aspect with the US and China, the local framing emphasizes national technological resilience and industrial strength. For Western readers: Western AI companies and investors should watch for potential new partnerships or competitive challenges in specialized AI applications, particularly those integrating physical hardware and robotics, as Japan accelerates its domestic development. Policy & Regulation Japan to launch council to overhaul legal frameworks governing AI use https://www.japantimes.co.jp/news/2026/07/07/japan/ai-new-council/ Japan’s government is establishing a new council to review and update legal frameworks related to AI, aiming to balance innovation with ethical and safety considerations. This initiative comes as Chief Cabinet Secretary Minoru Kihara emphasizes the need for a comprehensive approach to AI governance. The formation of this council indicates Japan’s intention to develop its own AI regulatory path, rather than simply adopting frameworks from the EU or US. This approach aims to foster domestic AI development while addressing public concerns, which is a common strategy for national ‘champion’ initiatives. For Western readers: Western tech companies operating or planning to enter the Japanese market should anticipate a distinct, Japan-specific AI regulatory environment emerging, potentially affecting data handling, model transparency, and liability. 🇨🇳 China Watch China’s technology moves, framed for Western readers China circumvents Western curbs by prioritizing compute substitution and physical manufacturing depth over raw frontier model benchmarks. Semiconductors & Hardware Huawei Updates Tao’s Law Paper, Discloses Detailed LogicFolding Process Parameters for First Time https://pandaily.com/huawei-tao-law-v2-logic-folding-process-parameters-jul2026-v2 Huawei has updated its ‘Tao’s Law’ paper, publicly disclosing detailed process parameters for its ‘LogicFolding’ technology, which aims to improve chip performance through advanced architectural design. This move signals Huawei’s strategy to address semiconductor manufacturing constraints through innovation in chip design rather than solely relying on fabrication process nodes. The updated paper details the use of 3D integration , heterogeneous architectures, and advanced packaging to optimize power, performance, and area PPA for computing systems. Huawei’s detailed parameter disclosure, especially around 3D stacking and heterogeneous integration, indicates a shift in focus towards architectural and packaging innovation to circumvent process node limitations. This isn’t just about faster chips; it’s about making existing fabrication methods more competitive for AI workloads. It highlights how Chinese firms are pursuing parallel paths to enhance computing power under sanctions, prioritizing domestic design over accessing the most advanced foreign foundries. For Western readers: Western companies relying on advanced process nodes for performance leadership should recalibrate competitive assessments, recognizing that Chinese firms like Huawei are investing heavily in design and packaging innovations that can yield substantial performance gains from less advanced silicon. Expect more competitive products from China built on this approach, particularly in areas where 3D integration offers advantages. AI & Machine Learning China’s Domestic AI in the First Half of 2026: From Compute Substitution to Training Closure https://pandaily.com/china-domestic-ai-first-half-2026-compute-substitution-training-closure-jul2026 In the first half of 2026, China’s domestic AI industry prioritized compute substitution , driven by US sanctions , shifting from an emphasis on large-scale model training. This strategic pivot aims to reduce reliance on advanced foreign GPUs and foster a self-sufficient AI ecosystem within China. China’s shift from training large models to compute substitution indicates a critical strategic adjustment in its AI roadmap, directly influenced by US export controls on advanced semiconductors, which impacts global AI development trajectories. For Western readers: Western semiconductor companies face sustained pressure and reduced market access in China, while Western AI developers will observe China’s unique, more constrained path to AI advancement. AI & Machine Learning Tencent Launches Hunyuan Hy3 AI Model, Integrates Across Products https://technode.com/2026/07/07/tencent-launches-hunyuan-hy3-integrates-model-across-multiple-products/ 📊 Featured Chart Source: TechNode Tencent has launched Hunyuan Hy3 , a new reasoning model leveraging a Mixture of Experts MoE architecture with 295 billion total parameters and a 256K token context window. The model is now integrated into several key Tencent products, including WorkBuddy/CodeBuddy and Yuanbao, and its API is available via Tencent Cloud . Tencent’s approach of integrating Hy3 deeply across its internal product suite reflects China’s broader strategy to monetize large language models through ecosystem control, rather than solely through standalone API access. This contrasts with some Western strategies that prioritize open access or specialized vertical applications. For Western readers: Western cloud providers should assume that Chinese competitors like Tencent will continue to build out highly integrated AI services within their existing product ecosystems, making direct competition on API-only models increasingly difficult in the China market. Cross-Regional Analysis To boost manufacturing, India need not choose between Japan and China https://www.scmp.com/opinion/asia-opinion/article/3359473/boost-manufacturing-india-need-not-choose-between-japan-and-china?utm source=rss feed 📊 Featured Chart FY2025-26 data; FDI for 2025-26 Japanese Prime Minister Sanae Takaichi recently led a business delegation to India, aiming to expand Japanese investment, which already totals over US$48 billion this century. Despite Japan being a significant investor, its trade with India is modest compared to China, which is India’s top trading partner. The article argues that Japanese industrial investments in India often rely on Chinese components, suggesting a pragmatic need for India to leverage both East Asian powers for manufacturing growth. The persistent interdependence of Japanese and Chinese supply chains, even as Japan seeks to build alternative manufacturing hubs, is often understated in Western commentary. Tokyo’s efforts to cultivate India are a long-term hedge, not an immediate replacement for China’s industrial base; Japanese firms will continue to source from China for cost and efficiency, even when investing in India. For Western readers: Western companies planning to onshore or friend-shore manufacturing outside of China must account for the reality that ‘diversified’ supply chains, particularly in countries like India, will likely still depend on China for critical components and raw materials for the foreseeable future. Policy & Regulation Talk of US-China decoupling is getting loud – but neither side is ready for a clean break https://www.scmp.com/economy/china-economy/article/3359613/talk-us-china-decoupling-getting-loud-neither-side-ready-clean-break?utm source=rss feed Despite escalating rhetoric and actions from both Washington and Beijing aimed at economic separation, the underlying financial and trade ties between the US and China remain deeply intertwined. Analysts suggest a complete unwinding is impractical due to extensive mutual dependencies, with China fortifying its financial systems while the US restricts capital flows. Beijing’s initiatives to diversify from the US dollar and build alternative payment systems , while framed by Western media as a challenge to dollar hegemony, are primarily a strategic move to secure its financial infrastructure against potential sanctions or disruptions. This allows China more autonomy in funding its domestic technology champions and managing trade without direct US leverage. For Western readers: Western businesses operating in or with China should assume a bifurcated financial system is emerging, requiring adaptation to yuan-denominated transactions and non-SWIFT payment rails for operations involving Chinese entities or domestic capital markets. 🔺 The Triangle Where US, Japan, and China technology interests intersect AI hardware demand is shifting bottleneck focus from cutting-edge logic to mature-node capacity and specialized memory supply chains. Semiconductors & Hardware2 STORIES Samsung’s 1,800% Profit Surge Proves AI Hardware Demand Remains Bulletproof https://www.electronicsweekly.com/news/business/samsung-expects-18x-increase-in-q2-profits-2026-07/ Driven by skyrocketing prices for DRAM and high-bandwidth memory chips, South Korean giant Samsung projected an 18-fold increase in Q2 profits. Despite growing Western anxieties over a potential software-driven AI bubble, the massive financial windfall and Samsung’s planned $70 billion production expansion show that global demand for physical AI silicon remains concrete and growing. This hardware boom is occurring alongside a pragmatic shift by some Western companies toward cost-effective Chinese AI models, adding a new layer of competition to the market. Why it matters: Samsung’s reported profit surge isn’t just a financial footnote; it reflects concrete supply-demand dynamics in the memory chip market. Western observers often miss how quickly these supply-side shifts translate into pricing power for the major East Asian memory players like Samsung, Hynix, and Micron, influencing everything from smartphone costs to data center buildouts. The $70 billion investment signals a commitment to maintaining market leadership, which for companies like Samsung, often means outpacing competitors in process technology and fab capacity to capture demand. For Western readers: If you are a Western enterprise AI provider or data center operator, anticipate continued upward pressure on memory component costs and potentially extended lead times for high-capacity modules, as East Asian memory giants prioritize market share and invest heavily in next-gen production. Semiconductors & Hardware AI Inference Pushes Memory Beyond DRAM and HBM https://www.eetasia.com/ai-inference-pushes-memory-beyond-dram-and-hbm/ Growing AI inference workloads are exposing severe memory capacity limitations , driving interest in high-bandwidth flash as a scalable alternative to traditional DRAM and HBM. The current memory architectures, optimized for consistent, cache-friendly access, are becoming inefficient for read-heavy, latency-tolerant AI inference with large, deterministic data access patterns. The article points to fundamental architectural issues with DRAM and HBM for AI inference, not just supply chain kinks. East Asian memory manufacturers, particularly the Korean giants, have focused on optimizing HBM for raw bandwidth, but this piece argues that capacity and latency tolerance are becoming equally critical for LLM inference, potentially shifting product development priorities and market share. For Western readers: If you are designing AI inference systems, assume that HBM supply for your capacity needs will remain constrained and inefficient; prioritize development on architectures that leverage high-bandwidth flash or other alternatives within the next 18-24 months. Semiconductors & Hardware AI Component Capacity Squeeze, Foundry Output Cuts to Extend Mature-Node Price Increases in 2027 https://www.eetasia.com/ai-component-capacity-squeeze-foundry-output-cuts-to-extend-mature-node-price-increases-in-2027/ Increased demand for AI-related products, coupled with production cuts by TSMC and Samsung in mature nodes, is tightening 8-inch and 12-inch foundry capacity and driving price increases. This shift has led to some high-voltage HV process orders moving from Taiwanese foundries to Chinese suppliers as capacity utilization reaches critical levels, particularly in 8-inch fabs. TrendForce expects these price increases to continue into 2027. The re-prioritization of advanced nodes by major players like TSMC and Samsung is a calculated move to maximize returns on leading-edge technology, pushing older nodes to second- and third-tier foundries. For Western firms, this means that even ‘mature’ components crucial for AI infrastructure like power management ICs are now subject to the geopolitical and supply-chain realities of East Asia, where Chinese fabs are becoming increasingly critical suppliers. This isn’t just about price; it’s about control over fundamental components. For Western readers: If you are a Western firm building AI infrastructure, assume that mature-node components like PMICs and power discretes will face persistent supply constraints and rising prices, requiring diversified sourcing strategies that include Chinese foundries. Cross-Regional Analysis Google boosts video commerce tools in Southeast Asia as YouTube shopping videos hit 6M https://technode.global/2026/07/07/google-unveils-video-commerce-tools-for-southeast-asia-as-youtube-shopping-videos-hit-6m/ Google has launched new AI-powered advertising and video commerce tools for Southeast Asia, citing a five-fold surge in YouTube video commerce since 2022, now accounting for 25% of the region’s e-commerce gross merchandise value. The initiative, piloted with Shopee, aims to enhance creator partnerships and direct high-intent shoppers from YouTube ads to checkout pages. Google’s strategy of leveraging AI in Search and the YouTube creator ecosystem for commerce directly challenges the established live commerce and social selling models refined by Chinese players like Douyin and Pinduoduo. This isn’t just about ads; it’s about owning more of the purchase funnel, and doing it in a region where Chinese platforms have a significant head start. For Western readers: Western e-commerce platforms and brands should analyze Google’s Southeast Asia pilots for scalable strategies against Chinese social commerce leaders, especially regarding the integration of creator content with AI-driven purchase paths. 🧩 Pattern This Week Japan: METI-backed consortium funnels $6.2B into SoftBank-led domestic AI models China: Domestic AI strategy shifts toward localized compute and training closure Korea: Samsung profits surge 18-fold amid skyrocketing AI-driven memory demand East Asian nations are rapidly decoupling from Western AI infrastructure by building self-sufficient, vertically integrated domestic ecosystems, which challenges the assumption that Silicon Valley will maintain its global monopoly on foundational AI technologies. AsiaAI.FYI https://asiaai.fyi · Written by Dick Weisinger · Subscribe https://asiaai.fyi