Wall Street Morning Report: Nvidia Falls for 7 Consecutive Days, AI Stocks Continue to Bleed Out; U.S. Treasury Buybacks Criticized as Temporary Fix, Inflation and Deficit Are the Real Big Problems

marsbitPublicado em 2026-08-25Última atualização em 2026-08-25

Resumo

Wall Street Morning Report: Tech and Treasury Tensions U.S. stocks were mixed. The Dow gained 0.26%, supported by defensive and consumer staples stocks, while the S&P 500 fell 0.28% and the Nasdaq dropped 0.76% amid a sell-off in AI hardware. The U.S. announced new sanctions targeting Iran's key economic sectors, temporarily easing oil prices (Brent -2.35%). However, concerns over potential disruption to the Strait of Hormuz could pressure European natural gas prices. Meanwhile, the Treasury's upcoming bond buyback plan faced criticism from major banks (Goldman Sachs, Deutsche Bank, Citadel), who argue it doesn't address the root causes of high long-term yields: persistent inflation and the massive fiscal deficit. The 10-year Treasury yield dipped slightly to around 4.70%. Gold rose, with Citi raising its short-term target to $4,800/oz. The AI hardware and semiconductor sector saw intense selling. The Philadelphia Semiconductor Index fell nearly 3%. NVIDIA dropped 2.91%, marking its seventh consecutive daily decline—its longest losing streak since 2022—due to pre-earnings caution despite strong analyst estimates. Memory chip stocks (Micron -6%, others down 5-6%) fell after Samsung's shareholder returns disappointed, raising cycle peak fears. Optical communications was also a big loser, led by Applied Optoelectronics (-14%) on equity dilution concerns. Wall Street is growing wary of off-balance-sheet credit risks in AI infrastructure financing. In other moves, Tesla fell 3...

From Monday to Friday morning, focusing on macro trends, U.S. stocks, AI, precious metals, and crude oil, we use data to review the markets and trends to seize opportunities. Produced by PANews.

Policy Temporarily Caps Oil Prices, But Cannot Disperse Fiscal Clouds

The three major U.S. stock indices diverged: the Dow Jones Industrial Average rose 0.26%, supported by financials, consumer staples, and defensive blue chips; the S&P 500 fell 0.28%; and the Nasdaq Composite fell 0.76%, with the AI hardware sector becoming the center of selling pressure.

Bessent launched the "Economic Orphan Action" on Monday, targeting five "lifelines" of Iran's digital assets, technology, gold, aviation, and shipping, and threatening secondary sanctions on countries that continue dealings with Tehran. Trump is communicating with leaders of various countries, demanding they sever economic interactions. Relevant parties will be given a grace period for rectification, otherwise the Treasury will act unilaterally. Iranian Supreme Leader's advisor Mohber swiftly countered: the response will be more resolute than ever. Sanctions and wars over the past 47 years have not worked; now they are facing internal unity and the Strait of Hormuz deterrent. The parliamentary speaker said trade partners "will not take seriously" the U.S.'s nonsense.

The market quickly "bought the rumor and sold the news": Brent crude fell 2.35% to $92.17 per barrel, with WTI falling in tandem to around $85.01, temporarily easing inflation premiums. Goldman Sachs warned that if the Strait of Hormuz crisis persists into winter, pressure on European natural gas inventories could further intensify, with TTF natural gas prices potentially needing to rise above €100 per megawatt-hour to attract sufficient LNG supplies. Current European gas storage is only about 62%, the lowest for this time of year in nearly two decades.

On the same day, Bessent remained restrained regarding U.S. Treasury operations: "We haven't even bought a single bond yet," with the next buyback not scheduled until September 9. Previous reports suggested the Treasury might use nearly $1 trillion from the TGA (Treasury General Account, balance approx. $935-950 billion) to fund the expanded long-term bond buyback. Morgan Stanley interest rate strategist Martin Tobias estimated the available scale between $80 billion and $200 billion. Goldman Sachs, Deutsche Bank, and Citadel Securities collectively poured cold water on the idea: buybacks cannot fundamentally reset long-term rates; the root causes are the U.S.'s $40 trillion debt, fiscal deficit, and inflation. Citadel Securities bluntly called this move "financial repression," stating pressure would only shift to the dollar, inflation, and other assets. The 10-year Treasury yield fell about 3 basis points to around 4.70%, with the 30-year also falling to 5.24%.

The U.S. Dollar Index rose slightly by about 0.20% to near 99. The dollar's strength mainly stemmed from safe-haven demand due to the Iran sanctions, but the market is also starting to worry that if the Treasury forcibly depresses long-term rates, the dollar could face pressure in the future. Data shows the premium of dollar put options over call options for the next month has risen to its highest level since February.

Gold continued to strengthen, hitting an intraday high of $4,696, reaching a nearly three-month high; New York gold futures were further pushed to $4,755. Citigroup has raised its 0-3 month gold target price to $4,800 per ounce and maintains its 6-12 month target of $5,000 per ounce. JPMorgan cautions that if PCE inflation exceeds expectations, gold may first pull back, but if the data is mild and Jackson Hole doesn't signal a strongly hawkish stance, gold prices could quickly approach $5,000.

Additionally, Trump continues to create pressure on the trade market, criticizing Canada for "taking advantage of the U.S." and warning of more serious consequences if Canada doesn't play by the rules. Market sources suggest the U.S. may raise tariffs on Canadian autos, parts, and steel to 50% starting in 2027, which would affect North American auto supply chain costs.

AI Hardware Sector Bleeds Out Comprehensively; Memory and Optical Communication Hit Hard, Funds Shift to More Stable Tech Platforms

Last night and this morning, U.S. AI hardware and semiconductor stocks faced intense selling pressure. The Philadelphia Semiconductor Index plunged nearly 3%, the memory sector fell about 5.5% overall, with the optical communication sector leading the decline.

Nvidia fell 2.91%, marking its seventh consecutive day of decline, its longest losing streak since 2022. The stock decline shows a clear divergence from fundamental expectations: Nvidia executives stated the Groq 3 LPX racks are now in full production, focusing on low-latency AI inference, and will be deployed alongside Vera CPU and Rubin GPU in Nebius data centers. Moreover, Wall Street earnings estimates for the past three months have actually been revised upward by about 13%, with the vast majority of 82 analysts maintaining buy ratings.

However, traders are more focused on risks ahead of earnings. Nvidia reports earnings this week, and the market worries that even if the results are strong, they might not be strong enough to support current valuations. Goldman Sachs believes Nvidia's demand data in Wednesday's earnings report might be "absurdly strong," but the real issues lie in memory costs and AI infrastructure financing. If hardware costs continue to rise, Nvidia will ultimately have to choose between absorbing costs, raising prices to customers, and reducing memory dependency.

Wall Street is increasingly alert to the credit risks of AI infrastructure. Broadcom's 5-year CDS has risen 28 basis points since August, with the yield on its bonds maturing in 2031 rising about 14 basis points over the same period; its risk premium even exceeds that of Oracle and SpaceX. JPMorgan strategists warn that off-balance-sheet credit support involved in the AI ecosystem—such as lease contracts, purchase commitments, and residual value guarantees—could ultimately reach trillions of dollars, raising deep concerns about "compute loans" and off-balance-sheet leverage risks.

The memory chain was a disaster area. Samsung Electronics' shareholder return plan falling short of expectations triggered a second wave of selling in global memory stocks. Micron, SanDisk, Seagate, Western Digital, and SK Hynix all plunged significantly. The market worries that the previous price increase cycle driven by HBM and AI servers has already been overpriced into stock valuations. The optical communication sector also faced pressure. Applied Optoelectronics fell nearly 14% after announcing its third ATM financing this year, planning to raise $600 million, raising investor concerns about cash flow and equity dilution. This dragged down Ciena, Lumentum, and Coherent.

Specific Project Actions and Stock Price Movements:

  • Nvidia closed down 2.91%, its seventh consecutive day of decline, marking its longest losing streak since 2022, with a cumulative drop of over 7% during the period. Risk-off selling dominated ahead of earnings, despite analysts raising earnings estimates by 13% over the past three months, most maintaining buy ratings, and target prices implying over 50% upside, with forward P/E having fallen to around 18x.

  • Semiconductor sector fell in unison, Broadcom down 2.63%: The yield on its 2031 bonds surged 14 basis points, and its CDS jumped 28 basis points in a month. Wall Street warns of massive off-balance-sheet credit risk accumulating from the residual value guarantees it provides for AI data center leases and purchase contracts. The Philadelphia Semiconductor Index fell 2.7%; AMD fell over 3%, Intel fell over 3%, TSMC fell over 2%.

  • Memory chips collectively sold off, Micron down nearly 6%: Samsung's shareholder returns falling short of expectations exacerbated market concerns about the memory cycle peaking. SanDisk down 6.45%, Seagate down over 6%, Western Digital down over 5%, SK Hynix down nearly 5%.

  • Optical communication was a disaster area, Applied Optoelectronics down nearly 14%: The company plans to raise $600 million via ATM financing, raising market concerns about equity dilution and funding pressure. Ciena down over 6%, Lumentum down over 4%, Coherent down over 4%, Marvell Technology down over 3%.

  • Tesla closed down 3.83%: Mainly impacted by a voluntary recall of about 3 million cars in China due to door handle and driving monitor hazards (a single-brand record). Despite the company scheduling a Cybercab event for September 3, traders predict only a 17% probability of delivering to retail customers this year. Related sectors: XPeng down over 8%, NIO down nearly 6%, Chinese new energy vehicle chains under pressure overall.

  • Meta rose 1.66% against the trend: Meta plans to launch the "Hatch" AI agent platform in the coming weeks, an intelligent shopping tool on Instagram, and plans to release the latest AI model "Watermelon" in October; the premium subscription version may cost up to $199.99 per month. Related sectors: Amazon up 1.33% (AWS launched Glue 6.0 with a 30% price reduction), Microsoft up 0.84%, Google up nearly 1%, Apple up 0.32%. Major platform stocks showed clear resilience.

  • SpaceX down 1.44%: SpaceX AI will adopt Nvidia Vera CPU to advance Agentic AI and plans to extend the optimized Vera Rubin NVL72 to space for Starmind AI satellites. Meanwhile, NASA's Roman Space Telescope is scheduled for launch on August 30 via a Falcon Heavy rocket.

  • Other giants: Bloom Energy up 1.28%, as Pelosi disclosed for the first time buying Bloom Energy stocks and options while adding to Intel, betting on AI power and chips; Boeing down 1.75%, with after-hours news that Boeing secured a U.S. Air Force F-15 contract worth up to $131.2 billion; Visa up 3.07%, Walmart up 2.69%, Disney up 2.62%, providing important support for the Dow's two-day gain against the trend.

What to Watch Next:

August 25 (Tuesday)

  • Jefferies Semiconductor, IT Hardware & Communications Tech Summit, August 25-26: Many well-known semiconductor and tech companies including TSMC, FormFactor, Aehr Test Systems, Arbe Robotics will participate. Industry discussions will focus on AI accelerators, HBM, advanced packaging, data center demand, and automotive semiconductor iteration.

August 26 (Wednesday)

  • Gamescom in Cologne, Germany, August 26-30, Opening at 02:00: Microsoft, Nintendo, Tencent, NetEase, CDPR and others will gather. Market focus on new game releases, AI game tools, cloud gaming, and hardware ecosystem news, potentially affecting sentiment in gaming, graphics cards, consoles, and content platforms.

  • Deutsche Bank California Technology Conference, August 26-27 in Dana Point, California.

  • Shenzhen AGIC General Artificial Intelligence Expo, August 26-28, overlapping with the Digital Expo (August 28-30) in Tongzhou. Dense AI industry conferences, focus on embodied intelligence and AI application progress.

Perguntas relacionadas

QWhat is the main reason for the sharp decline in the AI hardware and semiconductor stock sectors mentioned in the article?

AThe main reasons are broad profit-taking and risk aversion ahead of earnings reports, particularly for Nvidia. Additionally, concerns are growing about the fundamental valuation support for these stocks even if earnings are strong. Specific pressures include heightened market worries about 'compute debt' and off-balance-sheet leverage risks in the AI ecosystem, a disappointing shareholder return plan from Samsung triggering a sell-off in memory stocks, and equity dilution concerns from companies like Applied Optoelectronics raising capital via ATM offerings.

QWhy are financial institutions like Goldman Sachs and Deutsche Bank skeptical about the US Treasury's long-term bond buyback plan?

AThey are skeptical because they believe the buyback plan is a superficial fix that fails to address the root causes of high long-term interest rates. They argue that the core problems are the US's $40 trillion debt, persistent fiscal deficits, and underlying inflation pressures. Castle Securities even described the measure as a form of 'financial repression,' suggesting the pressure would merely shift to the US dollar, inflation, and other assets.

QWhat were the key factors driving the price of gold higher, according to the article?

AGold prices rose due to several factors: geopolitical tensions arising from new US sanctions threats against Iran, which increased safe-haven demand. Additionally, market concerns about potential future US dollar weakness if the Treasury's actions suppress long-term yields contributed. Citigroup raised its short-term gold price target to $4,800 per ounce, citing these supportive conditions.

QWhich major tech platform stocks performed relatively well despite the overall market sell-off, and what were the reasons?

AMeta (up 1.66%), Amazon (up 1.33%), Microsoft (up 0.84%), Google (up nearly 1%), and Apple (up 0.32%) performed relatively well. The article suggests investors shifted funds from volatile AI hardware stocks toward these more stable, large-cap technology platform companies. Specific catalysts mentioned include Meta's upcoming 'Hatch' AI agent platform and new AI model, as well as AWS's launch of Glue 6.0 with a 30% price reduction.

QWhat specific event is causing market concern for European energy markets beyond the immediate oil price reaction to Iran sanctions?

AAnalysts from Goldman Sachs warned that if the geopolitical crisis in the Strait of Hormuz persists into the winter, it could exacerbate pressure on European natural gas storage. With storage levels around 62% (a 20-year low for this time of year), the TTF gas price might need to rise above 100 euros per megawatt-hour to attract sufficient LNG supplies to meet demand.

Leituras Relacionadas

U.S. Aims at Iran's Crypto Sector Due to Oil Payments Exceeding $100 Billion

The U.S. Treasury Department has expanded sanctions on Iran to target its digital asset sector, citing over $100 million in cryptocurrency payments allegedly used to facilitate Iranian oil sales. The Treasury’s Office of Foreign Assets Control (OFAC) announced new sectoral sanctions covering digital assets, technology, gold, aviation, and shipping, alongside sanctions against nearly 60 entities, individuals, and vessels linked to nuclear, missile, cyber, and oil networks. The move allows OFAC to sanction foreign persons and firms operating in or supporting Iran’s crypto sector, which the Treasury stated is increasingly used as a "tool of choice" to circumvent sanctions, including by the Islamic Revolutionary Guard Corps (IRGC). Specific targets include Ukrainian broker Ivan Obukhov and his UAE-registered firm Foscom FZE, accused of processing over $100 million in crypto payments to facilitate oil sales for the IRGC’s Quds Force. This follows recent U.S. actions against Iran-linked crypto exchanges, including sanctions against UK-based Zedcex and Zedxion in January, four Iranian exchanges including Nobitex in June, and exchanges Shelbit and Aban Tether in August. Unlike previous targeted sanctions, the new measures establish a framework to penalize broader involvement in Iran’s crypto sector, significantly expanding OFAC’s authority under Executive Order 13902. Designated persons’ U.S.-linked property will be blocked, and foreign banks facilitating significant transactions for them may face U.S. account restrictions.

cryptonews.ruHá 36m

U.S. Aims at Iran's Crypto Sector Due to Oil Payments Exceeding $100 Billion

cryptonews.ruHá 36m

Conversation with Charles Zhang: Amidst the AI Frenzy, Why Pursue 'Anti-Efficiency' Content?

Dialogue with Charles Zhang: In the Era of AI's Sprint, Why Create "Anti-Efficiency" Content Standing out among China's early internet pioneers, Charles Zhang, founder and CEO of Sohu, maintains a distinct personal identity. He keeps a consistent public presence and embodies traits of high-energy individuals: sleeping only 4-5 hours daily, running 5km or marathons, and hosting frequent offline tech forums and dialogues with Nobel laureates. He has shifted focus from external business expansion to internal self-reconstruction, adhering to his own pace. "People aren't afraid of being busy; the busier you are, the better your health. Having nothing to do can lead to emotional issues and poor health." Marking the 10th anniversary of his live-streamed "Charles Zhang's English Class," he recently took it offline for a real-world classroom session. Following this, he discussed the course's decade-long journey, Sohu Video's content ecosystem, and the value of content in the AI era. **Pen and Mouth Diligence: A Class Against Instant Mastery** His English class began in August 2016 on "Qianfan Live." He spends about 40 minutes daily preparing, reviewing dozens of news headlines, reading five in detail, and selecting three for the lesson. Over ten years, he has completed over 2000 live sessions totaling more than 110,000 minutes, with over 52 million views. Initially, the stream was a way to "talk more" and combat emotional struggles. He started by repeating after TV news anchors, then moved to live-streaming with the launch of Qianfan Live. It was 40% for personal emotional regulation ("like a monk chanting sutras"), 30% to deliver authentic news, and 30% to promote the new platform. This daily "ritual" helped him overcome depression, leading to what he calls his "peak life state." From a product perspective, it was counter-intuitive—launched without prior market research or growth strategies, based on the founder's personal need that resonated with users. Yet, his content creation follows a strict routine. The decade of covering geopolitics, economics, tech, health, and society has built a vast "knowledge compound interest" system. In an age of AI summaries and algorithmic content, Zhang insists on "hard work": meticulous reading, handwritten notes, and manually created teaching materials, refusing AI-generated content. The English class is now a fixed daily rhythm; preparing for his weekly physics class requires a full day of solitary deduction on a blackboard. In his live streams, Zhang positions himself as a "content creator," not a CEO or teacher. He sees physics as extreme brain training and English as a "sensor" to understand the world. This explains why a top entrepreneur persists for a decade in seemingly "unprofessional" pursuits—it's the information input and psychological healing method he found during a low point. It's fundamentally an "anti-efficiency" commitment. He believes the meaning of reading lies in finding community and spiritual support, not just fleeting conversation points. For today's media, he states: "Reading news isn't the goal. As humans with many confusions, we need reference points, to find a group, find kindred spirits, or seek support and spiritual nourishment. Large language models can't satisfy these starting points; they only give you information." This offers solace and a value defense for media professionals impacted by AI summaries. **"Don't Think Earth Stops Turning Because AI Arrived"** A content creator expressed anxiety: after feeding their article to AI, it produced a more comprehensive, profound version, making their work feel valueless. Zhang responded with an analogy: "Carrots from the market, with a bit of soil, not glossy, taste delicious and flavorful. Tomatoes forced by chemical fertilizer have no taste at all." He sees AI giving world content a "plastic feel," but acknowledges its significant role in improving information access efficiency, especially in physics and social science research where he often queries AI for quick professional answers. The key, Zhang stresses, is that the outside world still needs to guide our perspective. LLMs provide "information," while the outside world offers "viewpoints." He cares deeply about *what* that external viewpoint is and *how* someone thinks about an issue. "Don't think the Earth stops turning because AI arrived," he analyzes. AI currently replaces rule-based work, noting significant impact on the software industry with many programmers laid off as even less experienced people can now write decent code. However, AI cannot yet replace arts, social sciences, emotions, and creative content. "Hollywood remains thriving, producing great works. Oscars will continue, Nobel Literature Prizes will still be awarded—all by real people, with works conceived by real people." While wary of AI, Zhang isn't anti-technology and is open to its applications. Sohu itself has launched AI assistants. In Sohu's AI practice, Zhang defines his role as a "large language model corpus provider." He believes LLM training is inseparable from human-original content as its data foundation, so original content retains irreplaceable value. "The era of self-media indeed offers everyone new ways to discover their strengths and achieve growth." He suggests every young person should consider self-media. Most importantly, the bottom line of content creation should be "authenticity." If using AI avatars, they must be clearly labeled; not doing so and misleading viewers wastes their time, especially when people still prefer real human content. Behind AI's sprint lie unavoidable realities. As AI quickly provides seemingly perfect answers, people grow accustomed to accepting "summarized" results, omitting processes of questioning, deduction, and trial-and-error, leading to a decline in deep thinking. Zhang's "anti-efficiency" practice with his two live-streamed classes is his answer to AI's狂奔: In an algorithm-driven era chasing ultimate efficiency, don't expect AI to fill inner voids. Read books, speak in front of a camera, sweat on a running track, and build connections with real people.

marsbitHá 46m

Conversation with Charles Zhang: Amidst the AI Frenzy, Why Pursue 'Anti-Efficiency' Content?

marsbitHá 46m

Trading

Spot
活动图片