#Nvidia深化AI生态合作布局

Nvidia doing the right thing

Nvidia customers just got warned: AI server prices are jumping 15%+ starting early next year, driven by soaring memory chip costs. Systems with Vera Rubin and Grace Blackwell chips included. Big cloud players already passing the word to their customers. Earnings drop Aug 26 https://t.co/dPiesiilUA

美股上周五迎来反弹，但市场真正的压力还在长端利率和AI资本开支 今晚贝森特将公布对伊朗制裁细则，油价和黄金都在等靴子落地；另一边，英伟达服务器被曝明年或涨价超15%，AI基建的成本账越算越贵。 资金也开始明显分化：半导体整体偏弱，特斯拉却大涨超5%，加密概念股同步走强。 本周英伟达财报、杰克逊霍尔年会接连登场，市场又到了重新给AI和利率定价的时候。

NVIDIA built its own coding harness to optimize CUDA GPU kernels and achieved a 100% score on ARC-AGI-3’s 25 public games, solving all 183 levels. With agents, we'll move from a world where it's quite hard to run, optimize, post-train your own AI models and kernels to a world where virtually everybody can do it. 100 million AI builders when?

🔥 LATEST: Nvidia’s $6B deal aims to build an open-weight model challenging DeepSeek, Kimi K3, OpenAI and Anthropic, per WSJ. https://t.co/4TAdJVVgFf

Nvidia could generate more than $1 trillion in operating profit. Not revenue. Profit. Based on current forecasts: → 2027: $260.5B → 2028: $371.1B → 2029: $459.0B That adds up to roughly: $1.09 TRILLION in operating profit across just three forecast years. Think about that for a moment. One company potentially generating more than $1 trillion in operating profit over three years. That’s bigger than the annual economic output of many countries. And this is why I think NVIDIA is more than just another semiconductor company. It sits at one of the most critical choke points in the AI economy: compute. Every major AI company needs enormous amounts of computing power. The more AI adoption accelerates, the more valuable that infrastructure becomes. Of course, forecasts are not guarantees. Competition, custom AI chips, pricing pressure and the enormous capital requirements of AI infrastructure could change the numbers dramatically. But if even a substantial portion of this trajectory materializes, we’re witnessing something extraordinary: AI isn’t just creating trillion-dollar companies. It may be creating companies capable of generating trillion-dollar profits. And NVIDIA could be the first spectacular example. Want to understand how to turn AI from a tool into a personal advantage? 10XMe, AI that actually works for you.10xme.biz⁠

Wall Street's still pretty bullish on$NVDAgoing into the Aug 26 print. Goldman kept their $285 target and are mostly watching how fast Vera Rubin ramps plus whether margins hold above 75%. BofA is even more bullish at $350, their whole thesis is that the new $500B financing deal takes the balance sheet risk off Nvidia and frees up way more cash for buybacks. They're calling the stock "significantly depressed" which is wild to hear about a $5T company lol.

🚨 Michael Burry is shorting Nvidia over its $500B AI financing deal. He says it has "shades of Enron's effort to make wholesale power an investable class." Here's the setup: Nvidia teams up with Apollo, BlackRock, Blackstone, Brookfield, Goldman and KKR to line up over $500B in financing so buyers can afford its chips, and Nvidia backs up to 25% of the residual value. Burry's read: this quietly props up demand for Nvidia's own product. If you're too young to remember Enron: back in 2001 they tried the same move with electricity instead of chips, buried the debt in shell companies, and the whole thing collapsed within months. Burry nailed the subprime call in '08, but he's missed a few since. We'll see if this one's a sequel or a swing and a miss.

From AT&T & GM dominating the 1950s to NVIDIA sitting alone at #1 in 2026. 75 years. One clear story: industrial muscle gave way to software, and software just got turbocharged by AI. The companies that build the intelligence layer now dwarf the ones that once built the physical world. The chart doesn’t lie, the future belongs to the chips.

英伟达正尝试把 GPU 从“快速折旧的硬件”变成可抵押、可产生租金现金流的金融资产 据文中信息，英伟达已与阿波罗、贝莱德、黑石等机构计划搭建规模最高5000亿美元的 AI 基础设施融资平台，部分贷款或以 GPU 与数据中心设施作抵押，并在特定案例提供最高25%的残值支持。 但这套模式的核心分歧仍在残值：H100 等 GPU 租金回升、推理需求延长设备经济寿命；另一面，二手市场价格与技术迭代风险也可能压低抵押品价值。算力资产化能否跑通，取决于未来租金能否覆盖折旧与融资成本。