为了打破英伟达的算力垄断，谷歌正在扶持云服务商Fluidstack分发自研TPU芯片，目前该公司正洽谈一轮7亿美元的巨额融资。最有意思的是，本轮融资 的潜在领投方，竟是被OpenAI「扫地出门」的天才研究员阿申布伦纳。在这场算力豪赌中，谷歌的野心、前OpenAI核心成员的复仇与资本的狂热正交织 在一起。 在硅谷这场愈演愈烈的算力军备竞赛中，一家名为Fluidstack的云服务商正悄然走到聚光灯下。 据知情人士透露，作为谷歌推广其自研AI芯片的关键盟友，Fluidstack正在洽谈一轮超过7亿美元的新融资。 这笔交易的潜在领投方颇具话题性——由前OpenAI研究员利奥波德·阿申布伦纳（Leopold Aschenbrenner）创立仅一年的新基金「Situational Awareness」。 据悉，阿申布伦纳的这只基金此前已经布局了AI云服务商CoreWeave和AI开发巨头Anthropic，而后者恰好也是Fluidstack的客户。 谷歌与英伟达的暗战前线 Fluidstack正处于谷歌与英伟达芯片大战的风暴眼中。 过去，这家公司主要依靠出租英伟达的GPU建立业务，但最近，它的战略重心发生了一次微妙而关键 ...

Core Insights - Nebius Group has raised approximately $3.75 billion through the sale of convertible bonds and stock, with a significant contract signed with Microsoft related to cloud AI computing resources [1] - The company increased its convertible bond issuance from an initial plan of $2 billion to $2.75 billion, alongside selling about $1 billion in new shares priced at $92.50 each [1] - Nebius's stock price has surged 237% year-to-date, closing at $93.39, with a single-day increase of nearly 50% [2][3] Financial Details - The convertible bonds were issued in two tranches, each raising about $1.375 billion, with maturities in 2030 (1% interest) and 2032 (2.75% interest) [1] - The conversion price for these bonds is set at $138.75 per share, reflecting a 50% premium over the stock sale price, indicating strong market demand for the company's financial instruments [1] Strategic Partnerships - The partnership with Microsoft, valued at nearly $20 billion, aims to secure AI cloud computing capabilities from Nebius, expected to generate $17.4 billion to $19.4 billion in revenue by 2031 [3] - Microsoft’s choice of Nebius is attributed to its rapid delivery capabilities, extensive AI server clusters, and alignment with NVIDIA's high-performance AI GPU roadmap [3][4] Market Positioning - Nebius is positioned as a leader in the "AI dedicated cloud" sector, focusing on faster delivery and optimized training/inference stacks for large model developers and enterprises [4] - The company operates large-scale AI server clusters in both the EU and the US, including significant data centers in Finland and upcoming facilities in New Jersey, enhancing Microsoft's capacity in North America and Western Europe [5][6]

Core Viewpoints - The company has signed expansion contracts with two hyperscale cloud customers in the past eight weeks, with one reflected in Q2 results. The remaining performance contract amount has doubled to $30.1 billion since the beginning of the year, including a $4 billion expansion agreement with OpenAI and new orders from large enterprises to AI startups [1][4][5]. Financial Performance - Q2 revenue reached $1.2 billion, a year-over-year increase of 207%, marking the first time the company achieved over $1 billion in revenue and an adjusted operating profit of $200 million in the same quarter [6][38]. - The remaining performance contract amount (Revenue Backlog) was $30.1 billion at the end of Q2, up 86% year-over-year and increased by $4 billion from Q1 [5][43]. - The company plans to achieve over 900 megawatts of active power delivery capacity by the end of Q4, with a current active power capacity of approximately 470 megawatts and a total contracted capacity of about 2.2 gigawatts [6][9][43]. Capacity Expansion - The company is significantly expanding its capacity to address structural supply shortages, with plans to add over 400 megawatts of power capacity [5][9]. - Capital expenditures totaled $2.9 billion in Q2, reflecting efforts to rapidly expand infrastructure to meet customer demand [10][84]. Customer Demand and Diversification - The customer base is becoming more diversified, with increasing demand for AI applications across various sectors, including media, healthcare, and finance [12][46]. - The company's VFX cloud service product, Conductor, saw usage grow over four times in the first half of 2025, and new contracts were signed with major banks like Morgan Stanley and Goldman Sachs [12][47]. Vertical Integration Strategy - The company is pursuing a vertical integration strategy through acquisitions, including Weights & Biases and a proposed acquisition of CoreScientific, aimed at enhancing infrastructure control and operational efficiency [14][15][63]. - The integration of CoreScientific is expected to provide approximately 1.3 gigawatts of total power capacity and eliminate over $10 billion in future lease liabilities, with an anticipated annual cost saving of $500 million by the end of 2027 [15][71]. Financing and Capital Markets - Since the beginning of 2024, the company has secured over $25 billion in debt and equity financing to build and expand its AI cloud platform [18][85]. - Recent financing transactions have significantly reduced capital costs, with a notable decrease of 900 basis points in borrowing costs [5][18][85]. Market Outlook - The company has raised its full-year revenue guidance for 2025 to a range of $5.15 billion to $5.35 billion, reflecting strong customer demand [8][89]. - The company continues to operate in a structurally supply-constrained environment, with demand far exceeding the supply of its products and services [9][86].

Group 1 - Rumble is considering acquiring Northern Data for approximately $1.17 billion (1 billion euros) to gain control over its GPU-rich Taiga cloud business and large data center division Ardent [1] - The Taiga cloud division has a significant inventory of NVIDIA GPUs, including about 20,480 H100 chips and over 2,000 H200 chips [1] - The proposed transaction values Northern Data at approximately $18.3 per share, representing a discount of about 32% from its recent closing price on the Frankfurt Stock Exchange [1] Group 2 - Northern Data's board is evaluating Rumble's acquisition proposal and is open to further discussions [1] - The final offer is expected to reflect a higher valuation, with Northern Data's major shareholder, Tether, expressing support for the transaction [2] - A condition for the potential deal is that Northern Data must divest its cryptocurrency mining business before closing, with proceeds used to repay existing loans from Tether [2]