每日信息简报｜2026 年 8 月 6 日

DAILY BRIEFING · 2026.08.06

技术扩张正在遇到真实世界的约束

从训练基准、大型代码库智能体到算力供给与数据中心外部性：AI 的能力边界正越来越由工程、能源、治理和组织执行共同决定。

今日要点

NVIDIA Blackwell Tops MLPerf Training 6.0 with Industry-Leading Scale and Performance

MLPerf Training 6.0 是 MLCommons 的训练基准；NVIDIA 称 Blackwell 在本轮测试中取得规模与性能领先。基准成绩适合比较特定提交配置，不应直接外推为所有模型训练任务的成本或速度。来源：NVIDIA Developer，发布于 2026-08-05。

Meta launches Muse Code, an AI agent for large code bases

TechCrunch 报道 Meta 推出面向大型代码库的 Muse Code 智能体。关键不在“能写多少代码”，而在它能否在真实仓库中理解依赖、约束与回归风险。来源：TechCrunch AI，发布于 2026-08-05。

How China’s A.I. Is Surging Across Africa

纽约时报观察到，非洲技术社区正在使用价格更低、可自由获得的中国模型。模型扩散不只是能力比较，也取决于价格、部署门槛、语言支持与本地开发者生态。来源：纽约时报 AI，发布于 2026-08-05。

AI 与科技

A Deluge of A.I. Computing Power Is About to Come Online, Fueling Major Leaps

纽约时报的可视化聚焦正在上线的 AI 计算能力。算力扩张会推动模型与应用迭代，同时把电力、选址、网络和资本开支推到同一张约束表上。来源：纽约时报 AI，发布于 2026-08-05。

Google Names Demis Hassabis to New AI Role in a Leadership Shake-up

Google 在多名顶级研究者离开后调整 AI 领导层。组织变化的意义要看研究、产品与基础设施的决策权如何重新连接，而非只看职位名称。来源：纽约时报 AI，发布于 2026-08-05。

Trump’s Push for More A.I. Data Centers Will Mean Major Air Pollution

纽约时报报道，扩大 AI 数据中心的政策取向可能带来显著空气污染外部性。数据中心讨论应同时核算算力收益、供电结构、后备发电与周边社区承受的成本。来源：纽约时报 AI，发布于 2026-08-05。

Jeff Dean and other top AI researchers are leaving Google to launch their own startup

TechCrunch 报道 Jeff Dean 等研究人员将离开 Google 创业，目标是用 AI 推进科学发现。人才流动本身不是成果，后续值得跟踪的是研究路线、数据与实验闭环是否形成。来源：TechCrunch AI，发布于 2026-08-05。

Finland Tries to Protect Children Against Deepfakes, With Russia in Mind

芬兰尝试把深度伪造识别纳入儿童与社会的信息韧性建设。技术检测必要但不足，教育、平台分发机制和可信来源的训练同样是防线。来源：纽约时报 AI，发布于 2026-08-05。

Taming Outlier Tokens in Diffusion Transformers

Apple 研究团队讨论扩散 Transformer 中异常 token 的处理。此类工作价值在于把模型稳定性问题定位到可测量、可复现实验，而不是仅以总体指标判断。来源：Apple Machine Learning Research，发布于 2026-08-05。

值得深读

Third-party cyber evaluations involving OpenAI models

这则报道聚焦 Third-party cyber evaluations involving OpenAI models。它的意义不只在单一消息本身，也在于把技术能力放回成本、治理、组织与现实约束中观察；具体事实以原文及后续一手资料为准。来源：OpenAI，发布于 2026-08-04。

Elon Musk’s attempt at an AI Wikipedia hasn’t been updated in months

这则报道聚焦 Elon Musk’s attempt at an AI Wikipedia hasn’t been updated in months。它的意义不只在单一消息本身，也在于把技术能力放回成本、治理、组织与现实约束中观察；具体事实以原文及后续一手资料为准。来源：The Verge AI，发布于 2026-08-06。

NVIDIA Ising Decoding Cuts Color Code Logical Error Rates by Over 300x

这则报道聚焦 NVIDIA Ising Decoding Cuts Color Code Logical Error Rates by Over 300x。它的意义不只在单一消息本身，也在于把技术能力放回成本、治理、组织与现实约束中观察；具体事实以原文及后续一手资料为准。来源：NVIDIA Developer，发布于 2026-08-05。

快速浏览

NVIDIA Developer · 2026-08-05Train Models Faster with JAX and MaxText Using NVFP4 on NVIDIA Blackwell：保留原文，适合结合上下文继续阅读。↗ V2EX · 2026-08-06[DeepSeek] DeepSeek API 要涨价了！且蹬且珍惜。：保留原文，适合结合上下文继续阅读。↗ V2EX · 2026-08-06[问与答] WPS 明明有文档场景优势，为什么 AI 体验还是没拉开差距？：保留原文，适合结合上下文继续阅读。↗ V2EX · 2026-08-06[问与答] AI 编程都这么卷了嘛，都出了新范式 SDD+TDD？：保留原文，适合结合上下文继续阅读。↗ V2EX · 2026-08-06[问与答] 大模型写论文相关咨询：保留原文，适合结合上下文继续阅读。↗ V2EX · 2026-08-06[问与答] UI/UX 方面的 Agent：保留原文，适合结合上下文继续阅读。↗ BBC · 2026-08-06Iran says deal with Oman on Strait of Hormuz is in final stages：保留原文，适合结合上下文继续阅读。↗ The Guardian (RSS Bridge) · 2026-08-05US to shut five consulates as critics fear China could fill diplomatic vacuum：保留原文，适合结合上下文继续阅读。↗

以下为本窗口全部原帖的紧凑信息流；内容仅按发帖者陈述呈现，不构成事实核验或投资建议。

@aleabitoreddit · 请求账号 @aleabitoreddit · 2026-08-06 01:23:49 UTC · 原创原帖：Funniest callout of the year from $VIAV on CPO timelines: "Well, there's been a lot of industry talk [about delays], like, 'Oh, because the yield is going to be slower.' That's all nonsense. CPO and all that thing is moving forward" They stated that Yes there's always issues, but the process is being improved and things are getting better. To follow up on that, the CEO stated: - Viavi has POs for CPO-testing activity - “We’re already getting [CPO revenue] this quarter, and probably in December it will start accelerating.” On a side note, they also stated, 1.6T is "ramping very quickly", should reach parity with 800g next year, and signaled longer tail for 800g demand. TLDR: Seeing CPO test revenues hit for $VIAV to $FORM around now. Nvidia also stated switch-side CPO for scale out and scale across has entered mass production earlier. As for scale up timelines, early production should be H2 2027 for some players, with other CPO players Ayar stating 2028 is the volume inflection period. Basically, all the timelines seem to be the same as before and revenue is starting to hit testing players first.↗ @thsottiaux · 请求账号 @thsottiaux · 2026-08-06 01:22:26 UTC · 引用原帖：I don't really think these legends needed a pitchdeck. Congrats on taking a crazy, but inspiring leap. I recommend you explore /goal in Codex, it's a pretty powerful loop with GPT-5.6 Sol.↗ @jukan05 · 请求账号 @jukan05 · 2026-08-06 01:14:22 UTC · 原创原帖：Google today feels like Meta when it was assembling MSL, while Meta today feels like Google when it was rolling out Gemini 3.↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 23:58:49 UTC · 原创原帖：SNDK's gross margin plateau is a welcome move. It means they are not squeezing customers to the extreme, and that they are looking at a long term vision instead. The disciplined, balanced capex execution I have been arguing for all along, the kind that prevents a company from investing to extremes at the top of the cycle and taking severe damage in the downcycle, requires SanDisk's NBM model. There is no way around it. So is now the right time to buy SanDisk? The bears argue that the gross margin plateau marks the peak of this cycle. That is also why the stock is falling right now. As I wrote in my SK Hynix earnings review(You can find it in the Citrini Substack chat), this is the point where we should be looking at Q, not P. As long as you believe in the AI supercycle, as long as big tech keeps spending, as long as compute companies like Nvidia remain in short supply, there is no scenario where the shortage eases for memory alone. That is the corresponding effect of LTAs (NBM) and disciplined capex. Even if gross margins stall, as long as Q expands and the floor set by those LTAs prevents an excessive collapse, the total profit pool will grow enormously. Layer massive shareholder returns such as buybacks on top of that, and memory companies should be rerated, and they will be rerated. From cyclical stocks to essential infrastructure of the AI world. $SNDK↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 23:26:57 UTC · 原创原帖：Taiwan's GlobalWafers: "Supply constraints have emerged in 12 inch wafers" Taiwan's GlobalWafers has formally acknowledged a shortage in silicon wafers. At its second quarter earnings conference call on the 4th, GlobalWafers said an imbalance between supply and demand in the semiconductor wafer industry is already underway, adding that its 12 inch (300mm), 8 inch (200mm) and 6 inch (150mm) lines have effectively reached maximum utilization and that supply constraints have already emerged for some advanced 12 inch products. Orders have surged on rising demand from AI semiconductors, high bandwidth memory (HBM), silicon photonics and advanced packaging. The company explained that the shortage has become a reality as logic makers, not only memory makers, move to secure long term capacity. The shortage is appearing across the wafer industry. Germany's Siltronic said in its earnings release on the 30th of last month that inventories at memory and logic customers are normalizing and that customers have begun rebuilding wafer safety stock, adding that this restocking is expected to increase wafer demand further and that new customers as well as existing ones are requesting 300mm volumes. Japan's Shin-Etsu Chemical, the global number one, also said in its earnings release on the 24th of last month that 300mm wafer shipments grew at double digit rates both year over year and quarter over quarter. Japan's SUMCO likewise assessed in May that customer inventory adjustment was entering its final stage. SK Siltron is expanding capacity by bringing equipment at its new fab online in stages, but the pace of order growth is understood to be outrunning that. The way long term agreements (LTAs) are struck is also changing. In the past the purpose was for customers to secure stable volumes, but recently they have taken on a stronger character of locking up capacity ahead of a shortage. Contract terms are also lengthening from around three years to five to eight years. Last month GlobalWafers signed a 10 year LTA with Micron, the longest in the wafer industry. The contract begins next year and includes strategic financial support in the nature of a prepayment. Most of its revenue in the second half of this year and the first half of next year is also expected to come from LTAs. Shin-Etsu Chemical said it currently supplies all of its 300mm wafers under LTAs. Siltronic supplies two thirds of its volume through LTAs. Memory customers are requesting volumes for delivery two to three years out. Moves to raise prices are also emerging. With supply tight, prices have begun to rise first in the non LTA (spot) market. Siltronic said prices for 300mm non LTA wafers have started to rise. GlobalWafers also expects spot prices in the second half of this year to be higher than in the first half. Price increases are expected to spread to LTAs as well. GlobalWafers said it is discussing price increases with customers to reflect higher raw material, energy, logistics and labor costs, adding that while LTAs already signed will maintain their contracted prices, new contracts will reflect not only price and volume but also price adjustment mechanisms, product mix and market changes. Siltronic likewise stressed that with some LTA renegotiations coming up in 2027 and 2028, prices need to recover to levels that allow reinvestment, and that broad and meaningful price increases are necessary. The wafer industry had previously gone through roughly two years of simultaneous shipment declines and price pressure as customer inventories built up sharply from 2023. Customer inventories of 300mm wafers then turned lower from late 2025, and months of inventory also declined from the first quarter of this year, completing the inventory correction. The industry is understood to have turned from the second quarter of this year. A deeper shortage and full scale price increases are expected to appear from the second half of 2027. Rising utilization at existing semiconductor lines is currently driving shipment growth, but once new memory fabs from Samsung Electronics, SK hynix and Micron come online, wafer demand is expected to rise to roughly double current levels.↗ @AndrewYNg · 请求账号 @AndrewYNg · 2026-08-05 17:33:12 UTC · 引用原帖：Congratulations @JeffDean, @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix! Knowing you guys, this will be amazing. Cheering you on and wishing you well on your important work!↗ @Brad_Setser · 请求账号 @Brad_Setser · 2026-08-05 17:22:44 UTC · 引用原帖：Thanks, but getting back to the 2008-2010 level v the USD isn't a victory -- want to see a signal that the PBOC will let the CNY continue to appreciate ...↗ @demishassabis · 请求账号 @demishassabis · 2026-08-05 16:04:58 UTC · 原创原帖：I’ve been working towards AGI my whole life, and as we enter this pivotal moment, I’m stepping into a new role as Chair of Google DeepMind & Chief Scientist of Alphabet. This will allow me to focus on long-term strategy, and accelerating scientific breakthroughs, including leaning into my work at Isomorphic to help cure disease. I’m excited that @koraykv will be stepping up to lead GDM as SVP, alongside @joshwoodward and our exec team. I could not be more excited and confident about our amazing next chapter! 🚀↗ @Claudia_Sahm · 请求账号 @Claudia_Sahm · 2026-08-05 15:51:15 UTC · 引用原帖：“The Fed’s work is only as effective as the public’s understanding of it, and you, the press, are essential to keeping the public informed about what we do and why. The people we serve benefit from your careful reporting.” Jay Powell at his last presser.↗ @michaelxpettis · 请求账号 @michaelxpettis · 2026-08-05 15:46:47 UTC · 原创原帖：Good WSJ piece on the dynamics behind any potential appreciation of the Chinese RMB. Much of the analysis is from Mark Sobel. @Lingling_Wei @sobel_mark↗ @michaelxpettis · 请求账号 @michaelxpettis · 2026-08-05 15:41:00 UTC · 原创原帖：Most people might not guess that this is Beijing, but this is what I saw earlier today as I cycled along Liangma in the northeast part of the city.↗ @Claudia_Sahm · 请求账号 @Claudia_Sahm · 2026-08-05 15:01:00 UTC · 引用原帖：my contribution to the dozen ...↗ @Claudia_Sahm · 请求账号 @Claudia_Sahm · 2026-08-05 12:59:34 UTC · 引用原帖：Watching the steady stream of Fedspeak now reminds me of my piece after the June meeting. The search for the shadow Fed Chair is on ...↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 12:57:26 UTC · 引用原帖：$AVGO↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 09:30:17 UTC · 引用原帖："For Chinese and Indian fabless companies that need to secure leading-edge-class performance amid U.S. sanctions on China, Samsung Foundry is a realistic alternative" LMAO↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 09:29:01 UTC · 原创原帖：Samsung Foundry’s 4nm Capacity Fully Booked as 5nm Orders Rise, Expanding into AI and Servers Samsung Electronics' foundry business is stepping up its sales push for its 5nm process. The move is read as offering 5nm as an alternative now that utilization of the in-demand 4nm node has hit its ceiling. According to semiconductor industry sources on the 5th, Samsung Foundry's 4nm process is expected to remain at full capacity through next year. "As Samsung Electronics' HBM4 order volumes have surged, 4nm line utilization has shot up," one industry source said. "The foundry's largest customer has become the memory division." On top of that, volumes for U.S.-based Nvidia's inference LPU "Groq3" have reportedly exceeded expectations, further contributing to the 4nm full-capacity situation. In response, Samsung Foundry is presenting 5nm as an alternative to customers seeking 4nm. The strategy is to pitch the 5nm process — where yield and performance are already proven — as a value-oriented substitute for fabless customers who find it difficult to move up to the 2nm and 3nm nodes given the higher cost burden versus 4nm. "The 2–3nm leading-edge nodes carry a heavy burden for customers in terms of yield and cost, and the 4nm line is already full," the source said. "Samsung Electronics is actively proposing its yield-stabilized 5nm process as an alternative, and fabless companies are viewing it favorably." The 5nm process is classified as a dedicated automotive line. Hyundai Motor's ADAS chip targeting mass production in 2030, Bos Semiconductor's SoC, and Telechips' automotive AP "Dolphin7" are all slated for volume production on SF5A, the automotive-specific process. More recently, its scope of application has been broadening into server and high-performance AI/NPU chips. "The 5nm process is also well suited to implementing high-performance server chips, so licensing agreements with overseas fabless firms have been coming in one after another," another industry source explained. "Over the past three to four months, customer demand for 5nm has clearly increased versus the past." Overloading on Taiwan's TSMC leading-edge lines and global geopolitical factors are also working in favor of Samsung Foundry's 5nm order intake. As supply-demand on TSMC's 3nm, 4nm and 5nm lines tightens, Chinese and Indian fabless firms are turning to Samsung's 5nm line both to meet their fine-node requirements and to circumvent sanctions. "For Chinese and Indian fabless companies that need to secure leading-edge-class performance amid U.S. sanctions on China, Samsung Foundry is a realistic alternative," a Samsung Foundry supply-chain partner said. "Samsung is absorbing even the smaller-volume customers pushed out of TSMC, so it is reaping spillover benefits on the 5nm process." Meanwhile, alongside the pragmatic 5nm-centered sales activity, global customers are also continuing to sound out volumes on the 2nm process. "I understand inquiries about 2nm keep coming in to Samsung Foundry," an industry source said. "What matters now is nailing the yield."↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 09:26:38 UTC · 引用原帖：• MI500 expected to outperform Nvidia’s Rubin Ultra in hardware performance 👀↗ @michaelxpettis · 请求账号 @michaelxpettis · 2026-08-05 08:55:02 UTC · 引用原帖：I would make Tordoir's point more broadly. German experiences provide an especially good set of examples on whether it is "innovation" or mercantilist labor policies are more likely to drive manufacturing competitiveness. Germany switched from small deficits and lower manufacturing competitiveness before 2003-05 to a rapid surge in trade surpluses and competitiveness thereafter. During the late 2010s, however, both once again started to decline. Given that the 2003-05 labor reforms resulted in a sharp decline in German wages relative to productivity, was the subsequent surge in Germany's manufacturing competitiveness driven by the growing gap between wages and productivity, or was it driven by a sudden increase in German innovation skills and a greater innate German ability to invent more advanced technology? If the latter, why did it last only a decade or so? By the middle of the 2010s, tight German labor markets resulted in a gradual recovery of wages relative to earlier productivity growth. Was the subsequent gradual loss of Germany's manufacturing competitiveness the result of the closing of the gap between wages and productivity, or did Germans around this time began to lose their ability to innovate and to invent more advanced technology? The point is that innovation and advancements in technology are important because they cause overall productivity to rise, and they allow certain manufacturing sectors, where the innovation is concentrated, to become relatively more competitive than other manufacturing sectors in that country. But they do not necessarily lead to all manufacturing sectors in a country becoming more competitive than the manufacturing sectors of that country's trade partners. That only happens when wages in that country are lower relative to productivity than they are in the rest of the world. Technological innovation is of course a great thing, but while it can explain sectoral trade surpluses, it has never explained systemic trade surpluses.↗ @Brad_Setser · 请求账号 @Brad_Setser · 2026-08-05 07:17:11 UTC · 引用原帖：I enjoyed this piece; recommended↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 07:09:39 UTC · 原创原帖：Korean media reports that Solidigm, a subsidiary of SK hynix, is pursuing a pre-IPO funding round ahead of a potential Nasdaq listing. The company is reportedly targeting a valuation of KRW 50 trillion in the fundraising round. Solidigm is a U.S.-based subsidiary established in 2021 to house Intel’s NAND flash and SSD business, which SK hynix agreed to acquire for approximately KRW 10 trillion in 2020.↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 06:16:09 UTC · 原创原帖：China's CXMT Plays Hardball — Strengthening Samsung and SK's DRAM Bargaining Power Apple, which had recently been evaluating China's Changxin Memory Technologies (CXMT) as a new supply-chain source in a bid to cut costs, is understood to be struggling in negotiations to secure further price reductions. Paradoxically, with the cheap-Chinese-component workaround now blocked, Samsung Electronics and SK hynix have been relieved of the burden of shipping commodity DRAM and are concentrating their production lines on high-value AI memory such as high-bandwidth memory (HBM) — creating a market dynamic in which they now hold complete control over global memory pricing power. ◆ "More expensive than Samsung" — CXMT rebuffs Apple as DRAM market leadership flips According to semiconductor industry sources on the 5th, Apple recently negotiated with CXMT over supply pricing for mobile DRAM including LPDDR5X, seeking to ease manufacturing cost pressure on its next iPhone and smart-device lineups, but was reportedly refused a price cut. CXMT instead held firm at prices equal to or higher than those quoted by Samsung Electronics and SK hynix. In effect, Apple — long a byword for buying power in the finished-goods market — has been stonewalled by a component supplier's hardball pricing. Behind CXMT's willingness to flatly reject Apple's demands is the wholesale volume sweep by Chinese IT giants such as Huawei and Xiaomi. With Washington's semiconductor sanctions on China still in force, Chinese device makers pursuing domestic self-sufficiency have locked up CXMT's DRAM output in advance through high-priced long-term contracts — leaving CXMT with no reason to bend to an Apple that demands both exacting quality qualification and lower prices. Analysts say the solid captive demand of China's domestic market is providing CXMT with a powerful pricing shield. The "cross-diversification into cheap Chinese components" playbook favored by Apple and other global device giants for beating down component prices has been rendered useless amid the commodity DRAM shortage. Finished-goods divisions have historically deliberately dangled Chinese panels or memory as bait to pressure affiliates and incumbent suppliers, thereby seizing the upper hand in annual pricing contracts. But as the supply-demand imbalance across the memory market accelerates, the balance of power is seen as having shifted decisively from set makers to memory manufacturers. ◆ Big Tech's cheap workaround closes — the shortage spreads to commodity DRAM This phenomenon is read as a knock-on effect of the sharp drop in absolute capacity for commodity DRAM (DDR5, LPDDR5X) caused by the buildout of HBM-dedicated production lines throughout the first half. As major memory makers cut wafer input for commodity DRAM and concentrated on HBM packaging processes to meet demand from next-generation AI servers and accelerators, even commodity DRAM prices have surged, entrenching the "memflation" phenomenon. With Chinese memory makers — once viewed as the alternative — now demanding full price as well, Big Tech has lost its detour around rising costs. The market's read is that this episode is feeding a virtuous cycle delivering record practical windfalls to Korean semiconductor giants Samsung Electronics and SK hynix. Because CXMT is absorbing large volumes of China-focused commodity DRAM, Korean firms have been fully freed from the shipment pressure of clearing low-priced commodity volume and from the associated price-decline risk. Into the space vacated by commodity DRAM oversupply fears, Samsung Electronics and SK hynix are aggressively pushing forward ultra-high-value AI memory lineups such as HBM4, LPCAMM2 and eSSD. Having reengineered their product mix toward high-value products, the two now hold unrivaled pricing power heading into second-half long-term pricing negotiations with global Big Tech. "An unusual situation has emerged in which CXMT is holding the price floor and thereby firmly underpinning commodity DRAM pricing as well," one industry official said. "With the commodity DRAM floor intact and Samsung and SK hynix monopolizing leadership in high-value AI memory like HBM4, operating margins and global market control for Korea's semiconductor camp should rise all the more steeply in the second half."↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 06:14:01 UTC · 原创原帖：SOUTH KOREAN MEDIA: APPLE RECENTLY NEGOTIATED WITH CXMT OVER SUPPLY PRICES FOR MOBILE DRAM, INCLUDING LPDDR5X, IN AN EFFORT TO EASE MANUFACTURING COST PRESSURES ON ITS NEXT-GENERATION IPHONE AND SMART DEVICE LINEUP, BUT CXMT REJECTED ITS REQUEST FOR PRICE CUTS SOUTH KOREAN MEDIA: CXMT INSTEAD INSISTED ON PRICES THAT WERE HIGHER THAN OR COMPARABLE TO THOSE OFFERED BY SAMSUNG ELECTRONICS AND SK HYNIX $AAPL↗ @michaelxpettis · 请求账号 @michaelxpettis · 2026-08-05 05:52:09 UTC · 原创原帖：Financial Times: "Beijing is stepping up efforts to attract foreigners to its sovereign debt by building out futures contracts and other bond market infrastructure."↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 05:49:02 UTC · 原创原帖：It appears that Victory Giant has resolved the issue.↗ @michaelxpettis · 请求账号 @michaelxpettis · 2026-08-05 05:45:53 UTC · 原创原帖：1/6 Financial Times: "China has launched a global hunt for hundreds of billions of dollars in unpaid taxes going back decades as Beijing seeks to fill a deepening fiscal hole by targeting the ultra rich."↗ @thsottiaux · 请求账号 @thsottiaux · 2026-08-05 04:29:28 UTC · 引用原帖：Better Cyber. Excited to welcome Halvar Flake to the team soon.↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 04:11:01 UTC · 引用原帖：👀👀↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 03:11:14 UTC · 引用原帖：Counterpoint Research published a worthwhile piece on the reported U.S. restrictions on imports of Chinese optical modules. "In an AI supercluster, a missing batch of optical transceivers renders billion-dollar GPU allocations idle. Constraining transceiver imports risks delaying the commercial turn-up of next-generation 800G/1.6T clusters by multiple quarters." The US restriction on imports of Chinese optical transceivers is nothing short of self inflicted damage. US data centers will literally grind to a halt.↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 02:24:46 UTC · 原创原帖：REUTERS: SAMSUNG AND SK HYNIX ARE EXPLORING THE ADOPTION OF CHINESE-MADE EQUIPMENT AS U.S. SANCTIONS COULD MAKE IT MORE DIFFICULT TO BRING EQUIPMENT INTO THEIR CHINA FABS According to Reuters sources, the two companies began testing AMEC’s etching equipment about two years ago, although no decision has yet been made on broader deployment. The aim is not to expand production capacity in China, but to keep Chinese suppliers in reserve in case U.S. restrictions are extended to the servicing, repair, or replacement of existing Western equipment. Samsung operates a NAND plant in Xi’an, while SK Hynix operates a DRAM plant in Wuxi and NAND facilities in Dalian. These plants currently rely heavily on etching equipment from U.S. suppliers such as Applied Materials and Lam Research. AMEC’s equipment is already used by Chinese chipmakers including YMTC. Chinese tools are about 20% to 30% cheaper than comparable foreign equipment, while the technology gap has also narrowed in areas such as etching, deposition, cleaning, and planarization. Approval from Samsung or SK Hynix would provide AMEC with significant technical and commercial validation and could, over the longer term, pose a threat to established global equipment suppliers such as Applied Materials, Lam Research, and KLA.↗ @jukan05 · 请求账号 @jukan05 · 2026-08-05 01:06:37 UTC · 原创原帖：TSMC Expands Outsourcing of Key Packaging Step CoW… Expected to Ease AI Chip Manufacturing Bottleneck TSMC is expanding outsourced production of a major process step for artificial intelligence (AI) chips. The move is seen as a step toward resolving the bottleneck in AI chip manufacturing. Much of the volume TSMC had been handling in-house is set to go to outsourced semiconductor assembly and test (OSAT) firms, and a large-scale wave of equipment investment is expected to follow. According to industry sources on the 4th, TSMC has decided to place additional CoW (chip-on-wafer) packaging volume — part of the CoWoS process used to manufacture AI chips — with ASE and other OSAT providers. In response, major OSAT companies are moving ahead with equipment orders to build new production lines. Purchase order (PO) discussions are also reportedly underway with Korean materials, components, and equipment suppliers. The specific investment scale has not been disclosed, but capacity is expected to be expanded substantially versus current levels. CoWoS is TSMC's 2.5D packaging technology, used in building AI chips. A processor (SoC) for AI workloads, such as a GPU, sits at the center with high-bandwidth memory (HBM) placed around it, and an intermediate substrate called an interposer is used to connect the two. TSMC refers to the step of attaching the chips to the interposer as CoW, and the step of bonding that interposer to the main substrate as WoS (wafer-on-substrate). Until now, TSMC has mainly outsourced WoS, the bonding to the main substrate. ASE, Amkor, and SPIL have been producing it under technology licenses from TSMC. Some CoW work was also outsourced, but only in limited volumes. The decision to significantly expand CoW outsourcing this time is attributed to the persistent bottleneck in AI chip manufacturing. Demand has surged as not only global AI chip designers (fabless firms) led by Nvidia, but also AI data center operators, move to develop and deploy their own custom silicon. TSMC is estimated to hold roughly 90% of the global AI chip manufacturing market. With demand continuing to grow and TSMC's own capacity insufficient to meet it, the company is looking to increase outsourcing. "TSMC has continued to invest in packaging capacity for AI chip manufacturing, but it is still being criticized for bottlenecks at the packaging stage," an industry official said. "This expansion of outsourcing is an attempt to scale up capacity together with its key OSAT partners, as market demand keeps expanding." As OSATs receiving CoW volume from TSMC ramp up equipment investment, back-end materials, components, and equipment suppliers are also expected to benefit. Investment is seen concentrating in particular on dicing — cutting wafers and interposers — and on the bonding processes that join wafers to interposers. According to multiple sources in the Korean equipment industry, discussions on supplying CoW-related tools are underway in the laser processing and bonding segments. $ASX $AMKR↗ @Claudia_Sahm · 请求账号 @Claudia_Sahm · 2026-08-05 00:22:35 UTC · 原创原帖：#PuffiSays let me at ‘em … this little guy was rescued from kitty and relocated.↗ @aleabitoreddit · 请求账号 @aleabitoreddit · 2026-08-05 00:16:18 UTC · 原创原帖：$RKLB has been awarded a $397M contract by the US Space Force. To build out Flatellite satellites and launch them on their Medium Lift rocket, Neutron.↗ @jukan05 · 请求账号 @jukan05 · 2026-08-04 23:26:21 UTC · 原创原帖：👀↗ @Brad_Setser · 请求账号 @Brad_Setser · 2026-08-04 23:01:12 UTC · 引用原帖：I think there is something to do this -- though not in such a direct way ... Remember that the math around the $550b investment commitment is pretty fuzzy, and a lot of it can be funded by borrowing dollars (via JBIC, etc) 1/↗ @Claudia_Sahm · 请求账号 @Claudia_Sahm · 2026-08-04 20:54:24 UTC · 引用原帖：My heart goes out to the staff compiling the press clips distro this week …↗ @thsottiaux · 请求账号 @thsottiaux · 2026-08-04 20:27:31 UTC · 原创原帖：What is my title at OpenAI↗ @Brad_Setser · 请求账号 @Brad_Setser · 2026-08-04 19:08:30 UTC · 引用原帖：Japan's "comms" around its fiscal policy needs to be improved. This is what the IMF said this spring: "The primary deficit in 2025 is estimated to have been smaller than it was in 2019 before the pandemic, and is among the smallest in the G7. The improvement, also marked by overperformance relative to staff’s projections" 1/2↗ @Brad_Setser · 请求账号 @Brad_Setser · 2026-08-04 18:35:32 UTC · 原创原帖：Setting the month to month fluctuations aside, the US trade deficit is clearly heading back up -- largely on demand for imported computers and such for the data center buildup. Oil exports helped limit the rise in the headline deficit in June, but that probably won't last 1/↗ @Brad_Setser · 请求账号 @Brad_Setser · 2026-08-04 17:50:10 UTC · 引用原帖：A couple of points from me -- tis true that Xi hasn't wanted the renminbi to be weaker than 7.3, and defended that level in the trade wars ... But that in trade terms is still a very weak yuan! 1/2↗ @sama · 请求账号 @sama · 2026-08-04 15:32:05 UTC · 引用原帖：i would rather be an optimist and work hard than a pessimist posting about why things won't work. it's much more difficult and the most likely path is failure, but society fails if people don't try. no amount of "it will never work" essays will drive society forward.↗ @michaelxpettis · 请求账号 @michaelxpettis · 2026-08-04 15:25:55 UTC · 原创原帖：1/2 The New Yorker has a very good article by Evan Osnos on the potential emergence of China's technological leadership. Osnos wonders if the future of technology might be Chinese, but he also notes that this future is very opaque.↗ @nft_hu · 请求账号 @nft_hu · 2026-08-04 14:10:14 UTC · 原创原帖：幸运的看到了富士山，账户也在回血中。希望等回到台湾，韩国抛售也结束了，然后筑底反弹，迎接新一轮牛市！↗

YouTube 更新

今日最值得看

DeepLearningAI：3rd Place Winner: Voice AI Prevents Data Loss. Coding Agent Calls Developer Before Deleting Records

本片由 DeepLearningAI 发布，围绕标题所述议题展开。简介提示：In this finalist demo for the 7-Day Voice AI Builder Challenge with @VocalBridgeAI, developer Sapna shows how voice escalatio。标题与节目主张不等于独立核验结论，建议结合原始资料复查。

Two Minute Papers：The Billion Dollar AI Race Just Broke

本片由 Two Minute Papers 发布，围绕标题所述议题展开。简介提示：❤️ Check out Lambda here and sign up for their GPU Cloud: 📝 Qwen 3.8 Max: Sources: 🙏 We would like to thank our generous Patr。标题与节目主张不等于独立核验结论，建议结合原始资料复查。

Practical Engineering：The Grid That Doubles the Strength of the Ground

本片由 Practical Engineering 发布，围绕标题所述议题展开。简介提示：I love geocells ⚙️ Check out SendCutSend for custom parts fabricated from your CAD: Sometimes it’s not all about more massive。标题与节目主张不等于独立核验结论，建议结合原始资料复查。

Money & Macro：Why do 90% of economists oppose rent control?

本片由 Money & Macro 发布，围绕标题所述议题展开。简介提示：Check out our advertising sponsor, The Economist, and get access to global coverage at an exclusive 35% discount at Recommend。标题与节目主张不等于独立核验结论，建议结合原始资料复查。

Patrick Boyle：Why Wall Street is Ignoring Big Tech's Debt

本片由 Patrick Boyle 发布，围绕标题所述议题展开。简介提示：Start your own store with #printify: The first 100 people who use the code PATRICK will get 15% off their first order! Nikkei。标题与节目主张不等于独立核验结论，建议结合原始资料复查。

其他频道更新

Greg IsenbergThese AI Marketing Agents Get You Customers
本片由 Greg Isenberg 发布，围绕标题所述议题展开。简介提示：I bring Cody Schneider back on the show to build two marketing agents end to end, live. The first one monitors LinkedIn posts。标题与节目主张不等于独立核验结论，建议结合原始资料复查。↗ FireshipThe safest way to store Bitcoin was just hacked...
本片由 Fireship 发布，围绕标题所述议题展开。简介提示：Build, test, and ship real apps in a day with Lovable: Last week the most secure way to store Bitcoin quickly became the leas。标题与节目主张不等于独立核验结论，建议结合原始资料复查。↗ DeepLearningAI2nd Place Winner: Coding Agent Calls Developer to Pitch Launch Strategy
本片由 DeepLearningAI 发布，围绕标题所述议题展开。简介提示：In this finalist demo for the 7-Day Voice AI Builder Challenge (in collaboration with @VocalBridgeAI) developer Eugenia sets 。标题与节目主张不等于独立核验结论，建议结合原始资料复查。↗ DeepLearningAI1st Place Winner: Coding Agent Calls Developer to Resolve Code Block
本片由 DeepLearningAI 发布，围绕标题所述议题展开。简介提示：In this 1st place demo for the 7-Day Voice AI Builder Challenge with @VocalBridgeAI developer Nick sets an AI agent to fix a 。标题与节目主张不等于独立核验结论，建议结合原始资料复查。↗ Bravos ResearchA Once in a Lifetime Economic Reset Is Arriving.
Bravos Research 对宏观“重置”的判断属于频道观点，不构成已核验预测或投资建议。应把片中数据、推断和交易结论拆开，并回到独立资料核验。↗ Wendover ProductionsWhy Texas Wins
本片由 Wendover Productions 发布，围绕标题所述议题展开。简介提示：Compare news coverage. Spot media bias. Avoid algorithms. Try Ground News today and get 40% off your subscription by going to。标题与节目主张不等于独立核验结论，建议结合原始资料复查。↗