微软将其首款搭载英伟达芯片的Surface笔记本定价为2599.99美元，并在发布会上声称，在本地AI提示词处理方面，其性能是MacBook Pro M5的两倍，但并未公布这两项数据的具体测量方法。该设备内部搭载的芯片基于Arm架构，而Arm正是剑桥公司，英伟达曾在2022年试图收购但未能成功。
Microsoft priced its first Nvidia-powered Surface at $2,599.99 and claimed on stage that it doubles a MacBook Pro M5 on local AI prompts, without publishing how either figure was measured. The chip inside is built on Arm’s architecture, from the Cambridge company Nvidia failed to buy in 2022.
微软已将其首款搭载英伟达芯片的笔记本定价为2599.99美元，并开启预购。该公司表示，Surface Laptop Ultra在运行本地AI模型时，返回提示词的速度是苹果MacBook Pro M5的两倍，视频生成速度快六倍。对于这两项声明，微软均未公布测试方法。
Microsoft has priced its first Nvidia-powered laptop at $2,599.99 and opened pre-orders. It says the Surface Laptop Ultra returns prompts from locally run AI models twice as fast as Apple’s MacBook Pro M5, and generates video six times quicker. It has published no test methodology for either claim.
据彭博社报道，高管们于周三在旧金山的舞台上进行了这一对比。微软官方产品页面列出了价格、最高20个英伟达Grace核心处理器、最高128GB统一内存以及最高2TB存储空间。
Executives made the comparison on stage in San Francisco on Wednesday, Bloomberg reported. Microsoft’s own product page carries the price, a processor of up to 20 Nvidia Grace cores, up to 128GB of unified memory and up to 2TB of storage.
其脚注中唯一提及的测量指标是屏幕。该公司称，在HDR内容下，通过10%窗口测得的峰值亮度为2000尼特，对比对象为2026年10月上市的笔记本。对于苹果的数据，则没有相应的对比说明。我们于周三上午询问了价格问题。
The only measurement it footnotes is the screen. Peak brightness of 2,000 nits, it says, measured over a 10% window with HDR content, against laptops available in October 2026. The Apple numbers get no equivalent. We asked the price question on Wednesday morning.
我们的发布预览指出，分析师认为成本是悬而未决的问题，因为内存短缺推高了硬件价格，而英伟达将其DGX Spark的价格提高了约75%，达到6950美元。以2599.99美元的价格，Surface比16英寸MacBook Pro便宜约400美元。芯片是更有趣的部分。
Our launch preview noted that analysts thought cost was the open question, because a memory shortage had pushed hardware up and Nvidia had raised its DGX Spark by about 75%, to $6,950. At $2,599.99 the Surface undercuts a 16-inch MacBook Pro by roughly $400. The chip is the more interesting part.
RTX Spark将Blackwell图形处理器与英伟达Grace CPU相结合，后者由联发科共同设计，英伟达声称其AI算力最高可达1拍字节（petaflop）。戴尔、惠普、联想、华硕和微星正在出货基于该平台的设备。这使得英伟达进入了英特尔和AMD共享数十年的笔记本市场。它通过他人的指令集实现了这一目标。
RTX Spark pairs a Blackwell graphics processor with Nvidia’s Grace CPU, co-designed with MediaTek, and Nvidia claims up to one petaflop of AI compute. Dell, HP, Lenovo, Asus and MSI are shipping machines built on it. That puts Nvidia into a laptop market Intel and AMD have shared for decades. It gets there on someone else’s instruction set.
Grace基于Arm的架构，该架构来自这家总部位于剑桥的公司——英伟达曾同意以400亿美元收购该公司，但在监管机构反对且软银于2022年2月确认交易告吹后，未能如愿。如今，每台RTX Spark笔记本电脑都要向Arm支付专利费。买家才是问题所在。
Grace is built on Arm’s architecture, licensed from the Cambridge company Nvidia agreed to buy for $40B and could not, after regulators objected and SoftBank confirmed the deal dead in February 2022. Every RTX Spark laptop now pays Arm a royalty instead. The buyers are the problem.
由于内存成本推高了系统价格，且企业延长了设备更新周期，欧洲笔记本电脑出货量预计将在本季度下降6.4%，第四季度下降20%，台式机出货量也将下降近30%。高德纳咨询公司（Gartner）预计，AI PC要到2028年才会占到出货量的一半以上。微软的宣传有着不寻常的受众。
European laptop shipments are forecast to fall 6.4% this quarter and 20% in the fourth, with desktops down almost 30%, as memory costs raise system prices and companies stretch their refresh cycles. Gartner does not expect AI PCs to be a majority of shipments until 2028. Microsoft’s pitch has an unusual audience.
其高管辩称，在眼前的机器上运行人工智能，比向云服务商付费更便宜、更安全。荷兰税务局本周就其电子邮件得出了这一结论，并正将邮件从微软的云端转移到自己的服务器上。
Its executives argued that running AI on the machine in front of you is cheaper and more secure than paying a cloud provider. The Dutch tax office reached that conclusion about its email this week, and is moving it off Microsoft’s cloud onto its own servers.