More on that later. Let's move on though on this and uh focus on those AI stocks amid some warnings from investors at least that AI spending is maybe getting out of hand. Ken Griffin saying for example to Sarah last week that right now there are echoes of the dot bubble and David Einhorn saying quote the numbers that are being thrown around are so extreme that it's really hard to understand them.Feel like I've said something similar to that as well. Joining us now is Byron Deer of Bessemer Venture Partners. ...

Alibaba Group Holding Ltd‘s (NYSE: BABA) (NYSE: BABAF) stealth AI expansion may be Wall Street's blind spot—but it's Nvidia Corp's (NASDAQ: NVDA) margin moat that's directly in the crosshairs. CEO Eddie Wu has pledged to surpass the company's $53 billion AI investment as demand for compute skyrockets, setting up a clash that could echo Beijing's solar strategy a decade ago: flooding the market with cheaper supply, undercutting Western rivals, and rewriting the margin math.Track NVDA stock here.Nvidia's Moat ...

DeepSeek-V3.2-Exp模型发布 - 公司于9月29日正式发布实验性模型DeepSeek-V3.2-Exp 作为迈向新一代架构的中间步骤 [2] - 新模型在V3.1-Terminus基础上引入DeepSeek Sparse Attention稀疏注意力机制 针对长文本训练和推理效率进行探索性优化和验证 [2] - 官方App、网页端、小程序均已同步更新为DeepSeek-V3.2-Exp模型 [2] API价格调整 - API模型版本更新为DeepSeek-V3.2-Exp 访问方式保持不变 [4] - 在新的价格政策下 开发者调用DeepSeek API的成本将降低50%以上 [4] 寒武纪生态适配 - 寒武纪于同日同步实现对DeepSeek-V3.2-Exp模型的适配 并开源大模型推理引擎vLLM-MLU源代码 [4] - 公司高度重视大模型软件生态建设 支持以DeepSeek为代表的所有主流开源大模型 [6] - 通过长期活跃的生态建设和技术积累 实现了对新模型架构的day 0适配和优化 [6] 技术协同与优化 - 寒武纪重视芯片和算法联合创新 以软硬件协同方式优化大模型部署性能并降低部署成本 [6] - 针对新模型架构通过Triton算子开发实现快速适配 利用BangC融合算子开发实现极致性能优化 [6] - 基于计算与通信的并行策略 再次达成业界领先的计算效率水平 [6] 产业链协同意义 - 同步发布适配动作表明中国AI产业链头部公司正在进行深度协同 [7] - 可能早在模型发布之前 寒武纪就在技术方面与DeepSeek进行沟通并启动相关适配研发工作 [7] - DeepSeek Sparse Attention机制叠加寒武纪的极致计算效率 可大幅降低长序列场景下的训推成本 [6]

Bitcoin (BTCUSD) mining companies were expected to quickly go bust as they were unable to cope with Bitcoin's block rewards being halved in the last halving event. Even though the significant increase in Bitcoin's price compensated for that and then some, mining companies were still unable to pull off a notable recovery. This caused many to take a creative approach. These companies use GPUs to mine Bitcoin, and the same hardware can be used for artificial intelligence (AI) and cloud computing purposes. As ...

NEW YORK, Sept. 29, 2025 (GLOBE NEWSWIRE) -- Rezolve Ai (NASDAQ: RZLV), a global leader in AI-powered commerce solutions, today issued the following statement in response to a report published by Fuzzy Panda Research: Rezolve categorically rejects the misleading and inaccurate allegations contained in the report. The publication selectively misrepresents facts, recycles outdated information, and disregards our audited financials, SEC filings, and the reality of our operations. A Report Driven by Self-Intere ...

Okay, welcome back to Morning Trade Live. It's time now for the big picture. Let's welcome in the team from Charles Schwab.Kathy Jones, fixed income strategist at Schwab Center for Financial Research. Michelle Gibli, director of international research over at Charles Schwab. Very good morning to you ladies.Thank you so much for your time this Monday. Kathy, let's start with you. What we're seeing as far as Treasury yields today.What is it telling you. Seeing a bit of a pullback across the board ahead of thi ...

Rezolve AI PLC (RZLV) closed the last trading session at $5.95, gaining 59.1% over the past four weeks, but there could be plenty of upside left in the stock if short-term price targets set by Wall Street analysts are any guide. The mean price target of $7.75 indicates a 30.3% upside potential.The average comprises six short-term price targets ranging from a low of $5.00 to a high of $10.00, with a standard deviation of $2.19. While the lowest estimate indicates a decline of 16% from the current price level ...

汇AI生态，创新质未来 南京河西中央科创区生态建设推进会举行 金融数智创新引领区：借力建邺作为南京金融核心集聚区的深厚底蕴，编织"金融赋能AI、AI重塑金 融"的双向价值网络，立志成为全国金融智能化转型的标杆、数智金融创新的源头活水。 "城市智能体"标杆：打造"全域感知、全时响应、全场景协同"的"城市级AI应用试验场"，让数字生活与 物理世界深度交融，成为探索未来城市发展的典范之作。 这一多层次、立体化的产业生态格局，让河西中央科创区从建设之初便站在时代的高起点。推进会现 场，一批重磅联合创新平台的集中揭牌，如繁星点亮夜空，彰显着生态体系的强大吸引力："大厂+大 学+大所"创新平台揭开神秘面纱，"数据+模型"共振赋能项目惊艳亮相，"VC-PE-并购"全链条基金矩阵 正式发布……每一个瞬间，都见证着这片热土正加速汇聚技术、资本、数据与人才，孕育出富有生命力 的AI创新生态系统。 要素集聚： 9月26日午后，南京国际博览中心内流光溢彩，一场以"汇AI生态，创新质未来"为主题的生态建设推进 会，如磁石般吸引着业界目光。这场盛会，既是河西中央科创区生态建设成果的璀璨亮相，也是对人工 智能产业未来发展路径的深情擘画，为 ...

凤凰网科技讯 9月29日，DeepSeek-V3.2-Exp模型发布，V3.2-Exp在V3.1-Terminus 的基础上引入了 DeepSeek Sparse Attention（一种稀疏注意力机制），针对长文本的训练和推理效率进行了探索性的优化 和验证。 在各领域的公开评测集上，DeepSeek-V3.2-Exp的表现与V3.1-Terminus基本持平。 目前，官方App、网页端、小程序均已同步更新为 DeepSeek-V3.2-Exp，同时API大幅度降价。在新的价 格政策下，开发者调用DeepSeek API的成本将降低50%以上。 ...

新产品发布与性能 - 公司于2025年9月29日发布DeepSeek-V3.2-Exp模型，该模型在V3.1-Terminus基础上引入了创新的DeepSeek Sparse Attention稀疏注意力机制 [2] - 新机制旨在优化长文本的训练和推理效率，且能在几乎不影响模型输出质量的前提下大幅提高长文本处理效率 [2] - 在各大公开评测集上，V3.2-Exp模型表现与V3.1-Terminus几乎持平，例如MMLU-Pro得分均为85.0，GPQA-Diamond得分分别为80.7和79.9 [3] - 部分基准测试显示性能提升，如Codeforces-Div1得分从2046提升至2121，AIME 2025得分从88.4提升至89.3 [3] 成本与可及性 - 公司宣布API价格大幅度下调，在新的价格政策下，开发者调用DeepSeek API的成本将降低50%以上 [2][3] - V3.2-Exp版本已同步更新至官方App、网页端及小程序，提升了产品的可及性 [3] - 公司表示V3.2-Exp模型已在Huggingface与魔搭平台开源，用户可直接访问并进行二次开发 [5]