【Tianfeng Securities Overseas Technology】 Takeaways from an Interview with SanDisk IRD 1. Pricing remains a slightly larger contributor to FY27 Q1 earnings growth, but it is reasonable to estimate that pricing and volume each account for roughly half. 2. Volumes under long-term agreements are also fixed, with no room for adjustment. In many cases, volumes are specified not only annually but also quarterly or even monthly. Product types—including TLC/QLC and process nodes—are also explicitly stipulated. 3. HBF is not taking away HBM’s addressable market; it represents an additional memory tier. It is specifically designed for inference rather than training. In inference workloads, read operations significantly outnumber write operations. 4. YMTC is expected to capture an 80% share of the Chinese market. However, China’s market is growing so rapidly that even this would not be enough to satisfy demand. The impact of YMTC’s capacity expansion is material, but it is unlikely to become a major risk to global supply and demand over the next several years. 5. Asian investors are concerned about price declines in Q4 and Q1. However, the company has seen no signs of pricing deterioration to date and does not expect a material change in the 2027 supply-demand balance. The market remains tight. In fact, NBM customers are returning to request higher volumes or longer contract durations. 6. Finally, the company was asked, “What is the market’s biggest misconception about SanDisk?” The response remained extremely bullish. The market is still adjusting to a “new reality”—the sustainability of the business has fundamentally changed. Despite the company providing a range of supporting metrics, the market continues to question the sustainability of its business model.“The market is essentially telling you that the company’s terminal value is zero.” $SNDK

【Tianfeng Securities Overseas Tech】 SK hynix CPO At first glance, SK hynix’s CPO technology roadmap appears to have one axis pointing toward HBM and another toward optical communications. The underlying logic, however, is that AI competition is shifting from the performance of individual chips to the efficiency of data movement across the entire system. Over the past several years, HBM has addressed the problem of GPUs not receiving data fast enough. Vertically stacking multiple layers of DRAM and placing them next to the GPU enables extremely high bandwidth. However, as more HBM stacks are placed around each GPU, packaging area, interposer-edge space, power delivery, and cooling are all approaching their limits. At the same time, AI clusters have already scaled to thousands or even tens of thousands of GPUs. No matter how fast each GPU can compute, the entire system will still hit the “bandwidth wall” if data cannot move efficiently between GPUs and racks. SK hynix’s vision is to extend optical interconnects to memory. Low-latency, high-bandwidth local HBM would remain next to the GPU, while optical fiber would connect it to a larger shared memory pool. This would eliminate the need to fit all memory capacity within a single GPU package and enable horizontal scaling at the rack level. The investment implications are as follows. This does not mean that HBM will be replaced anytime soon. Instead, a new memory hierarchy is more likely to emerge. The most frequently accessed data would remain in local HBM, while larger volumes of less frequently accessed data would be stored in an optically interconnected memory pool, HBF, or SSDs. SK hynix is repositioning itself. It is moving beyond selling standardized memory chips toward jointly designing HBM, controllers, advanced packaging, and system-level memory architectures with customers. If this roadmap materializes, SK hynix could strengthen customer stickiness, increase product value-added, and improve its ability to secure long-term contracts. It could also position the company to respond proactively to the potential disruption that future memory disaggregation may pose to the traditional HBM business model. At the supply-chain level, areas likely to benefit over the long term include silicon photonics chips, optical engines, lasers, fiber coupling technologies, and advanced 2.5D and 3D packaging. That said, this concept still appears to be at an extremely early stage—essentially just a roadmap. I interviewed someone involved in TSMC’s packaging operations today, and they were completely unaware of this initiative.

【Tianfeng Securities Overseas Tech】 After reviewing all the various AI related earnings, we believe the AI bull market is returning on the back of strong fundamentals One very clear trend is emerging from the recent earnings of companies across the AI industry chain. Growth in cloud businesses is accelerating again, and compute demand continues to exceed supply. Customers are showing a willingness to accept price increases and even to prepay, and the unit economics of AI infrastructure are improving. Market attention is shifting away from concerns about excessive capital expenditure (CapEx) and back toward the belief that demand and profitability will actually materialize. Large cloud: The slope of cloud business growth turned upward this quarter, and the scale of new bookings in the quarter was overwhelmingly larger than in any of the past 20 quarters. Google has effectively no choice but to go all in. The mindset is that failure means the end. Even with negative free cash flow (FCF), every $1 of incremental cloud revenue converts into roughly $0.54 of new operating profit. On a simple calculation, the payback period is about 2 years. AWS's margin of 39.4% is very strong. This provides an answer to the questions about return on investment (ROI). AWS will ultimately become a business with $1 trillion in annual revenue, with roughly 5x of growth headroom over the long term. The payback period for the cloud business is also under 3 years. Microsoft: Azure's growth reaccelerated, providing an answer to concerns about the durability of that growth. Neocloud: Earnings, demand, pricing power, and financing capability have all been validated across the board. CoreWeave's (CRWV) recent compute resources are effectively all sold, with multiple customers competing to secure GPU volume. It added about 500MW of active power in a single quarter, and its year end target has been raised to more than 1.85GW. Its customer base is also broadening from large AI model companies and cloud giants to industrial, financial, life sciences, government, and traditional enterprises. AI is moving from a training centered frenzy toward a phase of adoption across every industry. CoreWeave raised prices across all SKUs by about 25% in July. The profit contribution rate on new contracts rose by 5 to 10 percentage points. This shows that customers judge the returns they earn from AI to be sufficient to absorb higher compute prices. Even the A100, launched in 2020, is being signed into high priced contracts running through 2029. This also refutes the core bear case that GPUs are rapidly rendered obsolete within 3 to 4 years and that their residual value goes to zero. Nebius (NBIS) saw total new contract value in the second quarter grow roughly 4x versus the prior quarter, with new customer contracts up more than 9x. Price tolerance on the demand side is so strong that, under current terms, it could sell all of its 2027 capacity today. Annual revenue on new contracts reaches $20 million to $25 million per MW. Pricing on prior generation GPUs also rose more than 30% versus the prior quarter, and short term total bookings reach as much as $40 million to $50 million per MW. Customers have begun prepaying the cost of building AI infrastructure. About 70% of new contracts include prepayment provisions, which can cover 50% to 60% of the related CapEx. Expected project payback periods have shortened from 2 to 3 years previously to 1 year and 10 months. This means AI infrastructure is not simply a business that scales while burning cash, but one that has already secured strong unit economics. Optical communications: The FCC related disruption is limited and the practical impact should be minimal. Lumentum (LITE) reported results above market expectations and raised guidance. EML demand exceeds supply by about 30%. The shortage of ultra high power lasers also continues to widen. Innolight's laser products are effectively sold out, and it plans to expand capacity roughly 4x over the coming quarters. The bottleneck remains delivery capability. Quarterly OCS revenue clearly exceeded $100 million for the first time. NPO is not a technology that replaces CPO but a newly added intermediate architecture. Key customers' CPO plans are unchanged and demand signals are getting stronger. AI architecture is becoming increasingly optically integrated. Memory and storage: It is now nearly certain that a bottom is forming. The core logic is to take time to secure room for recovery. The absence of price increases after the share price decline is already reflected in market expectations, and visibility on long term agreements (LTA) and shareholder returns is steadily improving. Compute: The market narrative is shifting back toward AI training and open source. We are positive that Nvidia (NV) will challenge its prior high, and a narrative around RSI is also gradually forming.

Tiana Musarra got emotional and broke down in tears after Kai Cenat accepted her into Streamer University 2026 🥺❤️ “Sh*t never works out for me and now I’m finally f*cking doing it. I’m so happy.”

Tiaohua cross-stitch embroidery is a traditional handicraft passed down through generations by women of Huayao, a branch of Yao ethnic group, in Longhui County. The origin of Tiaohua cross-stitch embroidery dates back to the Han Dynasty (202 B.C. - 220 A.D.) and reached maturity during the Ming (1368 - 1644) and Qing (1644 - 1911) dynasties. The embroideries feature a unique "Qingsha" stitching skill, and are improvised with silk threads on blue homespun cloth without pre-drawn designs. Even an ordinary embroidered tube skirt may contain hundreds of thousands of stitches, requiring several months or even years to complete. In 2006, Longhui Huayao Tiaohua embroidery making skill was listed in the first batch of the national intangible cultural heritages. Since the Huayao people have no written language, Tiaohua embroidery has played an important role in preserving their cultural and historical heritage, earning it the title "a wordless history book worn on the body." "Every pattern has a spiritual meaning behind," said Shen Yanxi, a provincial-level inheritor of Tiaohua. Over the centuries, Tiaohua cross-stitch embroidery has changed in style, with patterns and colors being renewed to better suit contemporary aesthetics. In recent years, Longhui County has promoted the preservation and innovation of Tiaohua by leveraging cultural tourism, digital preservation, and projects introducing intangible cultural heritage into campuses, bringing this traditional skill, once secluded in remote areas, to a wider audience around the world.