[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"content-doc-dciof6a525e1":3},{"user":4,"document":8,"mainDocument":28,"columnUrl":30,"subscription":31,"footer":43,"text":78},{"isAuthenticated":5,"isAdmin":5,"displayName":6,"avatarUrl":6,"nid":6,"groupLevel":7},false,"",-10,{"id":9,"fullTitle":10,"subTitle":6,"url":11,"columnId":12,"columnName":13,"columnUrl":14,"summary":6,"contentHtml":15,"mainContentHtml":6,"posterUrl":16,"createDate":17,"displayDate":18,"displayDateSlash":19,"pageviews":20,"tags":21,"hidden":5,"isSubContent":5,"replyDocOrTargetId":6,"contentType":24,"videoId":6,"liveVideoUrl":6,"duration":25,"price":25,"priceText":26,"priceBadgeText":26,"priceBadgeClass":27,"freeForMinGroupLevel":25,"redirectUrl":6,"readyToStream":5},"dciof6a525e1","DeepSeek让英伟达GPU重新变得供不应求","\u002Fdoc\u002Fdeepseek让英伟达gpu重新变得供不应求_1529734387","col18178739ee","美股资讯","\u002Fcol\u002Fcol18178739ee","\u003Cp data-start=\"69\" data-end=\"191\">上周，一位英伟达代理商向媒体透露，春节后首周，两家中国科技企业为支撑开源大模型\u003Cstrong data-start=\"115\" data-end=\"127\">DeepSeek\u003C\u002Fstrong>的推理服务，紧急向其公司采购\u003Cstrong data-start=\"141\" data-end=\"152\">L20 GPU\u003C\u002Fstrong>设备。“去年，客户下单前会反复比价，如今连库存里的禁售型号都被一抢而空。”\u003C\u002Fp>\n\u003Cp data-start=\"193\" data-end=\"314\">DeepSeek的崛起，颠覆了市场对AI赛道高性能GPU垄断地位的传统认知，使英伟达一度遭遇市场看空。然而，产业观察人士发现，近期无论是高性能H系列GPU，还是\u003Cstrong data-start=\"273\" data-end=\"283\">RTX 40\u003C\u002Fstrong>系列高端显卡，\u003Cstrong data-start=\"290\" data-end=\"311\">英伟达GPU再次陷入供不应求的状态\u003C\u002Fstrong>。\u003C\u002Fp>\n\u003Ch2 data-start=\"316\" data-end=\"340\">\u003Cstrong data-start=\"319\" data-end=\"338\">DeepSeek引爆算力抢购潮\u003C\u002Fstrong>\u003C\u002Fh2>\n\u003Cp data-start=\"342\" data-end=\"429\">上述英伟达代理商表示，目前正加紧调配英伟达产品，以满足客户激增的需求。他指出，英伟达芯片在中国市场\u003Cstrong data-start=\"391\" data-end=\"401\">几乎全线紧缺\u003C\u002Fstrong>，部分\u003Cstrong data-start=\"404\" data-end=\"416\">禁售型号H800\u003C\u002Fstrong>的存货已“一卡难求”。\u003C\u002Fp>\n\u003Cp data-start=\"431\" data-end=\"514\">与此同时，接入\u003Cstrong data-start=\"438\" data-end=\"450\">DeepSeek\u003C\u002Fstrong>的企业已达数百家，涵盖\u003Cstrong data-start=\"461\" data-end=\"487\">云计算、互联网、手机、芯片、金融、医疗、汽车\u003C\u002Fstrong>等多个行业。这一趋势为相关企业带来了巨大商业机遇：\u003C\u002Fp>\n\u003Cul data-start=\"516\" data-end=\"601\">\n\u003Cli data-start=\"516\" data-end=\"536\">\u003Cstrong data-start=\"518\" data-end=\"527\">云服务厂商\u003C\u002Fstrong> 借机兜售算力\u003C\u002Fli>\n\u003Cli data-start=\"537\" data-end=\"558\">\u003Cstrong data-start=\"539\" data-end=\"547\">芯片企业\u003C\u002Fstrong> 借势推广自家产品\u003C\u002Fli>\n\u003Cli data-start=\"559\" data-end=\"581\">\u003Cstrong data-start=\"561\" data-end=\"567\">车企\u003C\u002Fstrong> 以更低成本升级智能软件\u003C\u002Fli>\n\u003Cli data-start=\"582\" data-end=\"601\">\u003Cstrong data-start=\"584\" data-end=\"592\">上市公司\u003C\u002Fstrong> 借此推升市值\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp data-start=\"603\" data-end=\"697\">一家IT公司向媒体透露，其调用\u003Cstrong data-start=\"618\" data-end=\"630\">DeepSeek\u003C\u002Fstrong>的平台在\u003Cstrong data-start=\"634\" data-end=\"658\">上线首日新增上万用户，导致服务器几近崩溃\u003C\u002Fstrong>，现有算力远远跟不上用户需求。团队不得不\u003Cstrong data-start=\"678\" data-end=\"694\">紧急采购更多英伟达GPU\u003C\u002Fstrong>。\u003C\u002Fp>\n\u003Cp data-start=\"699\" data-end=\"769\">多家上线DeepSeek的平台都出现“\u003Cstrong data-start=\"718\" data-end=\"727\">服务器繁忙\u003C\u002Fstrong>”提示，实际上并非技术故障，而是\u003Cstrong data-start=\"743\" data-end=\"766\">提供的算力远远无法匹配用户的爆发式增长\u003C\u002Fstrong>。\u003C\u002Fp>\n\u003Cp data-start=\"771\" data-end=\"881\">\u003Cstrong data-start=\"771\" data-end=\"801\">超算云服务商并行科技（839493.BJ）董事长陈健\u003C\u002Fstrong>指出，AI模型的生命周期通常包括\u003Cstrong data-start=\"817\" data-end=\"829\">训练、部署、推理\u003C\u002Fstrong>三个阶段，目前\u003Cstrong data-start=\"836\" data-end=\"861\">DeepSeek的推理需求正处于井喷式爆发\u003C\u002Fstrong>，带动市场对英伟达GPU的抢购热潮。\u003C\u002Fp>\n\u003Cp data-start=\"883\" data-end=\"907\">目前，\u003Cstrong data-start=\"886\" data-end=\"902\">英伟达在华合法销售的型号\u003C\u002Fstrong>包括：\u003C\u002Fp>\n\u003Cul data-start=\"908\" data-end=\"1007\">\n\u003Cli data-start=\"908\" data-end=\"931\">\u003Cstrong data-start=\"910\" data-end=\"924\">H20、L20、T4\u003C\u002Fstrong>（企业级）\u003C\u002Fli>\n\u003Cli data-start=\"932\" data-end=\"960\">\u003Cstrong data-start=\"934\" data-end=\"946\">RTX 40系列\u003C\u002Fstrong>（除RTX 4090外）\u003C\u002Fli>\n\u003Cli data-start=\"961\" data-end=\"982\">\u003Cstrong data-start=\"963\" data-end=\"980\">RTX 30系列及更早型号\u003C\u002Fstrong>\u003C\u002Fli>\n\u003Cli data-start=\"983\" data-end=\"1007\">\u003Cstrong data-start=\"985\" data-end=\"1005\">最新RTX 5090D（限制版）\u003C\u002Fstrong>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp data-start=\"1009\" data-end=\"1077\">虽然市面上已有十余款国产GPU品牌，但在\u003Cstrong data-start=\"1029\" data-end=\"1045\">性能、稳定性、生态兼容性\u003C\u002Fstrong>等方面仍难以匹敌英伟达，导致企业仍将其视为\u003Cstrong data-start=\"1066\" data-end=\"1074\">最优选择\u003C\u002Fstrong>。\u003C\u002Fp>\n\u003Ch2 data-start=\"1079\" data-end=\"1111\">\u003Cstrong data-start=\"1082\" data-end=\"1109\">英伟达股价大起大落，DeepSeek影响几何？\u003C\u002Fstrong>\u003C\u002Fh2>\n\u003Cp data-start=\"1113\" data-end=\"1206\">DeepSeek的\u003Cstrong data-start=\"1122\" data-end=\"1133\">R1模型发布后\u003C\u002Fstrong>，英伟达股价一度遭遇\u003Cstrong data-start=\"1143\" data-end=\"1161\">一年内最大单日跌幅（17%）\u003C\u002Fstrong>，引发市场震动。然而，自2月3日起，其股价已回升13%，\u003Cstrong data-start=\"1189\" data-end=\"1203\">显示市场情绪正在修复\u003C\u002Fstrong>。\u003C\u002Fp>\n\u003Cp data-start=\"1208\" data-end=\"1330\">一位国内模型厂商高管向美股投资网表示，\u003Cstrong data-start=\"1224\" data-end=\"1252\">DeepSeek的出现让行业开始反思“算力泡沫”\u003C\u002Fstrong>，是否未来AI训练能摆脱\u003Cstrong data-start=\"1264\" data-end=\"1278\">大规模高端GPU集群\u003C\u002Fstrong>的依赖？但现实情况是，\u003Cstrong data-start=\"1289\" data-end=\"1309\">模型推理的算力需求仍在爆炸式增长\u003C\u002Fstrong>，最终仍然推高了对英伟达GPU的需求。\u003C\u002Fp>\n\u003Cp data-start=\"1332\" data-end=\"1375\">\u003Cstrong data-start=\"1332\" data-end=\"1345\">并行科技董事长陈健\u003C\u002Fstrong>指出，DeepSeek给行业带来的\u003Cstrong data-start=\"1362\" data-end=\"1370\">最大冲击\u003C\u002Fstrong>在于：\u003C\u002Fp>\n\u003Cul data-start=\"1376\" data-end=\"1443\">\n\u003Cli data-start=\"1376\" data-end=\"1405\">\u003Cstrong data-start=\"1378\" data-end=\"1403\">大幅降低AIGC大模型的训练和推理算力消耗\u003C\u002Fstrong>\u003C\u002Fli>\n\u003Cli data-start=\"1406\" data-end=\"1443\">\u003Cstrong data-start=\"1408\" data-end=\"1426\">短期内，推理需求远超训练需求\u003C\u002Fstrong>，但训练的长期需求仍将稳步增长\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2 data-start=\"1445\" data-end=\"1472\">\u003Cstrong data-start=\"1448\" data-end=\"1470\">DeepSeek算力需求是否可持续？\u003C\u002Fstrong>\u003C\u002Fh2>\n\u003Cp data-start=\"1474\" data-end=\"1589\">尽管DeepSeek展现出强大推理能力，但\u003Cstrong data-start=\"1495\" data-end=\"1508\">其算力消耗成本极高\u003C\u002Fstrong>。上述国内模型厂商人士透露，\u003Cstrong data-start=\"1522\" data-end=\"1558\">DeepSeek的“满血版”至少需要两张H800 GPU才能运行\u003C\u002Fstrong>，而国内许多基础模型仅需\u003Cstrong data-start=\"1570\" data-end=\"1584\">一张RTX 4090\u003C\u002Fstrong>即可。\u003C\u002Fp>\n\u003Cp data-start=\"1591\" data-end=\"1616\">AI模型的能力通常取决于\u003Cstrong data-start=\"1603\" data-end=\"1613\">参数量和精度\u003C\u002Fstrong>：\u003C\u002Fp>\n\u003Cul data-start=\"1617\" data-end=\"1656\">\n\u003Cli data-start=\"1617\" data-end=\"1636\">\u003Cstrong data-start=\"1619\" data-end=\"1634\">参数量决定AI知识储备\u003C\u002Fstrong>\u003C\u002Fli>\n\u003Cli data-start=\"1637\" data-end=\"1656\">\u003Cstrong data-start=\"1639\" data-end=\"1654\">精度影响AI的推理能力\u003C\u002Fstrong>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp data-start=\"1658\" data-end=\"1722\">但由于许多企业上线DeepSeek时\u003Cstrong data-start=\"1676\" data-end=\"1687\">配备的算力不足\u003C\u002Fstrong>，导致模型在推理精度上\u003Cstrong data-start=\"1698\" data-end=\"1710\">无法发挥最大潜力\u003C\u002Fstrong>，进而影响用户体验。\u003C\u002Fp>\n\u003Cp data-start=\"1724\" data-end=\"1759\">不过，\u003Cstrong data-start=\"1727\" data-end=\"1756\">DeepSeek当前的算力需求或许不会长期维持高位\u003C\u002Fstrong>。\u003C\u002Fp>\n\u003Cul data-start=\"1760\" data-end=\"1903\">\n\u003Cli data-start=\"1760\" data-end=\"1809\">\u003Cstrong data-start=\"1762\" data-end=\"1806\">开发者通常会将超600B参数的大模型进行蒸馏，压缩至更低参数规模，以便商业化部署\u003C\u002Fstrong>。\u003C\u002Fli>\n\u003Cli data-start=\"1810\" data-end=\"1844\">由于DeepSeek发展过快，团队\u003Cstrong data-start=\"1829\" data-end=\"1841\">尚未完成模型蒸馏\u003C\u002Fstrong>。\u003C\u002Fli>\n\u003Cli data-start=\"1845\" data-end=\"1903\">预计\u003Cstrong data-start=\"1849\" data-end=\"1858\">未来几个月\u003C\u002Fstrong>，经过优化后的DeepSeek版本将降低推理算力需求，使企业减少对英伟达GPU的依赖。\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2 data-start=\"1905\" data-end=\"1916\">\u003Cstrong data-start=\"1908\" data-end=\"1914\">总结\u003C\u002Fstrong>\u003C\u002Fh2>\n\u003Cp data-start=\"1918\" data-end=\"1946\">DeepSeek的爆发式增长，正在重塑AI算力市场：\u003C\u002Fp>\n\u003Col data-start=\"1947\" data-end=\"2101\">\n\u003Cli data-start=\"1947\" data-end=\"1980\">\u003Cstrong data-start=\"1950\" data-end=\"1969\">英伟达GPU在中国市场全线告急\u003C\u002Fstrong>，供不应求状况加剧\u003C\u002Fli>\n\u003Cli data-start=\"1981\" data-end=\"2018\">\u003Cstrong data-start=\"1984\" data-end=\"2004\">DeepSeek推理需求远超预期\u003C\u002Fstrong>，企业纷纷加码GPU采购\u003C\u002Fli>\n\u003Cli data-start=\"2019\" data-end=\"2049\">\u003Cstrong data-start=\"2022\" data-end=\"2037\">英伟达股价受冲击后回暖\u003C\u002Fstrong>，市场情绪正逐步恢复\u003C\u002Fli>\n\u003Cli data-start=\"2050\" data-end=\"2101\">\u003Cstrong data-start=\"2053\" data-end=\"2077\">长期来看，模型蒸馏技术或将降低GPU需求\u003C\u002Fstrong>，但短期内DeepSeek仍将维持高算力消耗\u003C\u002Fli>\n\u003C\u002Fol>\n\u003Cp data-start=\"2103\" data-end=\"2160\" data-is-last-node=\"\">DeepSeek的兴起，或许是AI算力市场的一次革命，但在“去GPU化”真正实现之前，\u003Cstrong data-start=\"2146\" data-end=\"2159\">英伟达仍是最大赢家\u003C\u002Fstrong>。\u003C\u002Fp>","https:\u002F\u002Fstockwewebfiles.blob.core.windows.net\u002Fweb-202502-stk\u002FUploaderspbn3so3qkl_2025_02_17_10_48_54.jpg","2025-02-17T18:46:02.407","2025.02.17","2025\u002F02\u002F17",49545,[22,23],"NVDA","DEEPSEEK","Article",0,"免费","success",{"id":9,"fullTitle":10,"subTitle":6,"url":11,"columnId":12,"columnName":13,"columnUrl":14,"summary":6,"contentHtml":15,"mainContentHtml":6,"posterUrl":16,"createDate":17,"displayDate":18,"displayDateSlash":19,"pageviews":20,"tags":29,"hidden":5,"isSubContent":5,"replyDocOrTargetId":6,"contentType":24,"videoId":6,"liveVideoUrl":6,"duration":25,"price":25,"priceText":26,"priceBadgeText":26,"priceBadgeClass":27,"freeForMinGroupLevel":25,"redirectUrl":6,"readyToStream":5},[22,23],"\u002Fcol\u002Fstocknews",{"visible":5,"marketingHtml":32,"services":33,"recentDocuments":42},"\u003Cfigure class=\"image\">\u003Ca href=\"https:\u002F\u002Fstockwe.com\u002Fdoc\u002Fdcio537efad5\" target=\"_blank\" rel=\"noopener noreferrer\">\u003Cimg style=\"display:block;margin-left:auto;margin-right:auto;\" src=\"https:\u002F\u002Fstockwewebfiles.blob.core.windows.net\u002Fweb-202408-stk\u002F1586109431mceclip0.jpg\">\u003C\u002Fa>\u003C\u002Ffigure>\u003Cdiv class=\"text-center\">\u003Ch2 class=\"card-title mx-auto\">\u003Cbr>\u003Ca target=\"_blank\" rel=\"noopener noreferrer\" href=\"https:\u002F\u002Fstockwe.com\u002Fdoc\u002Fdcio537efad5\">案例介绍：英伟达深度研究报告\u003C\u002Fa>\u003C\u002Fh2>\u003C\u002Fdiv>",[34,38],{"productId":35,"serviceName":36,"priceText":37},"prod_PPxdDdK87QaiLv","月付","$12.95美元",{"productId":39,"serviceName":40,"priceText":41},"prod_PPxeMs3bix1da5","年付","$149.00美元",[],{"links":44,"images":69,"summaryHtml":74,"aboutTitle":75,"aboutHtml":76,"copyrightHtml":77},[45,48,51,54,57,60,63,66],{"label":46,"url":47},"深度报告","\u002Fcol\u002FdepthReport",{"label":49,"url":50},"VIP会员","\u002Fplan",{"label":52,"url":53},"期权推荐","\u002FOption",{"label":55,"url":56},"低价暴涨股","\u002FPenny",{"label":58,"url":59},"常见问题","https:\u002F\u002Fstockwe.com\u002FFAQ",{"label":61,"url":62},"美股课程","\u002Fcol\u002Fvideos",{"label":64,"url":65},"免责声明","\u002Fdisclaimer",{"label":67,"url":68},"联系我们","\u002FContactUs",[70,71,72,73],"https:\u002F\u002Fstockwebsiteblob.blob.core.windows.net\u002Fweb-202509-stk\u002FUploaderzic2tuwsol2_2025_09_11_18_21_07.gif","https:\u002F\u002Fstockwebsiteblob.blob.core.windows.net\u002Fweb-202509-stk\u002FUploadercakzdvydksw_2025_09_03_09_00_56.png","https:\u002F\u002Fstockwebsiteblob.blob.core.windows.net\u002Fweb-202509-stk\u002FUploadergtjyagwvoyk_2025_09_14_08_32_05.png","https:\u002F\u002Fstockwebsiteblob.blob.core.windows.net\u002Fweb-202509-stk\u002FUploader3u0tt4jhlqh_2025_09_23_22_30_48.png","邮箱: buy@TradesMax.com 美国电话 626-378-3637","公司介绍","\u003Cp class=\"MsoNormal\">美股大数据 StockWe.com 是一个美国领先的金融和美股信息大数据提供商，紧盯华尔街金融市场和行情，2008年成立于美国硅谷，创始人是前纽约证券交易所资深分析师Ken，联合多位摩根斯坦利分析师，谷歌 Meta工程师利用AI和大数据，配合十多年美股实战经验和业内量化交易模型，每天处理千万级股票数据：挖掘潜力大牛股，捕捉期权异动大单，实时主力资金流向、机构持仓变化、川普突发新闻，精准买卖信号第一时间发到您手机APP。\u003C\u002Fp>","专业美股投资者都在这里",{"loading":79,"search":80,"searchPlaceholder":80,"hotContent":81,"draft":82,"noData":83,"searchNoData":84,"edit":85,"editVideo":86,"courseContent":87,"more":88,"buyNow":89,"subscribeNow":90,"encoding":91,"paidContent":92},"Loading...","搜索","热门内容","草稿","目前没有任何内容公布","当前检索内容没有数据","编辑","编辑视频","课程内容","更多","立即购买后观看","- 立即订阅 -","视频编码中...","付费内容"]