8:10 Krypton: Tencent stock surges, Nvidia and Microsoft collaborate on unified tech stack, plus childcare subsidies news 8点1氪丨腾讯股价暴涨，创2021年后单日最高涨幅；预计今年安排约1100亿育儿补贴；英伟达与微软合作推出统一技术栈

ByteDance's Doubao is finally ready to unveil its paid menu. This long-awaited move is less an exploration of business models than a meticulously calculated market positioning. The $9.9 standard version and $73 professional version set a pricing spectrum that fits precisely between OpenAI’s offerings and the budgets of domestic users. ByteDance’s strategy is clear: wrapping the rationale for payment around "complex tasks" and "productivity scenarios," it aims to capture a cohort of users willing to pay for PPT creation, data analysis, and video production during AI’s transition from toy to tool. This is pragmatic, but it also carries a sense of urgency—when Tencent integrates Meituan through Yuanbao and Alibaba bundles its services with Tongyi Qianwen, relying solely on the "Doubao is free" card is no longer enough.

The soaring stock price of Tencent, on the other hand, signals another kind of release. What the market has been waiting for is perhaps not just a financial report, but a clear move to complete the AI loop within a "super app." A casually mentioned remark by Meituan CEO Wang Xiang on an earnings call—that the integration of "Xiaomei" with Tencent’s Yuanbao is about to launch—directly became the trigger for boosting market value. This means that a billion users will soon experience a "seamless leap" within WeChat, from chatting to ordering takeout. It is not merely the integration of features; it is a frontal assault on the heartland of Alipay’s lifestyle services. Tencent’s market value surged by nearly HK$500 billion overnight, buying future imaginative potential: when an AI assistant can mobilize an entire ecosystem of services, the path to monetizing traffic will be completely rewritten. The eight-month slump in the stock price prior to this was, to some extent, the market’s skepticism over whether Tencent still has a growth story to tell; now, that story has a fresh AI cover.

Perhaps the most intriguing is Jensen Huang’s statement in Taipei. The godfather who ascended to the pedestal as a GPU devotee has unexpectedly declared that the company’s Vera CPU "will be more popular than GPUs" and become "the new main growth driver." This sounds like a technological self-revolution. For a long time, NVIDIA’s narrative has been built on the mantra that "AI needs more GPUs." But at this moment, Jensen hints that the real bottleneck and future value may shift toward the CPU, which has long played a supporting role. The logic behind this might be: as Agentic AI requires autonomous planning, decision-making, and operation of complex systems, general-purpose computing power and low-latency response become critically important. He acknowledges supply constraints while remaining steadfast in Vera’s future—this seemingly contradictory statement actually reveals the quiet shift in NVIDIA’s strategic focus. From selling "shovels" to providing complete "mining systems," NVIDIA’s ambition clearly extends beyond just chips.

On the other side, Microsoft has launched its own reasoning model, Mai-Thinking-1, while Tencent Cloud has slashed the price of DeepSeek-V4 by up to 97.5%. One is building a moat upward, while the other is leveling the competitive playing field downward. Microsoft’s entry adds a heavyweight player to the large model "arms race," and Tencent Cloud’s "price-cutting sword" will undoubtedly reshape the cost structure of domestic AI application development, further catalyzing an explosion at the application layer. This is no longer a war between large model companies, but an ecosystem battle where platform ecosystems "co-opt" developers through cost control.

From Häagen-Dazs stores being taken over by the tea brand Ningji, to Tiger Brokers pausing new domestic account openings, these seemingly AI-unrelated fragments piece together a broader picture: the consumer market is seeking new stories amid saturation, while financial regulators are drawing new boundaries through standardization. Meanwhile, AI is permeating the capillaries of commerce at an unprecedented speed, reshaping the rules of collaboration, and even challenging our fundamental understanding of computing power. None of the day’s news items are trivial. Together, they outline the contours of an era in which technology, capital, and regulation are accelerating their alignment. Here, there are no eternal kings—only constantly rewritten rules of the game.

字节跳动的豆包终于要亮出付费菜单了。这个姗姗来迟的动作，与其说是商业模式的探索，不如说是一次精心计算的市场卡位。9.9美元的标准版，73美元的专业版，这个定价区间像一把标尺，精确地卡在了OpenAI和国内用户的钱包之间。字节的算盘很清楚：用“复杂任务”和“生产力场景”来包装付费理由，试图在AI从玩具到工具的蜕变期，抢先圈定一批愿意为PPT、数据分析和视频制作掏钱的用户。这很务实，但也透着一股急迫——当腾讯用元宝接美团、阿里用通义千问整合全家桶时，光靠“豆包免费”这张牌，已经不够听了。

腾讯股价的暴涨，则是另一种信号的释放。市场苦等的，或许不是一份财报，而是一个“超级应用”完成AI闭环的明确动作。美团CEO王兴在电话会上轻描淡写的一句话——“小美”与腾讯元宝合作即将上线——直接成了引爆市值的导火索。这意味着，十亿用户即将在微信里，体验从聊天到下单外卖的“无缝跃迁”。这不仅仅是功能的接入，这是对支付宝生活服务腹地的正面突袭。腾讯的市值一夜飙升近5000亿港元，买的是未来想象空间：当AI助手能调用全生态服务时，流量货币化的路径将彻底被改写。股价之前长达八个月的低迷，某种程度上是市场对“腾讯还有没有增长故事”的怀疑；如今，故事换上了AI的新封面。

最耐人寻味的，是黄仁勋在台北的表态。这位以GPU信仰者身份登上神坛的教父，居然公开宣称，公司的Vera CPU“将比GPU更受欢迎”，并成为“新的主要增长动力”。这听起来像是一次技术的自我革命。长期以来，英伟达的故事建立在“AI需要更多GPU”这句咒语上。但老黄此刻却暗示，真正的瓶颈和未来价值，可能转向了那个长期扮演配角的CPU。这背后逻辑或许是：当AI代理（Agentic AI）需要自主规划、决策和操作复杂系统时，通用计算能力和低延迟响应就变得至关重要。他承认供应限制，却同时笃定Vera的未来，这种看似矛盾的表述，恰恰揭示了英伟达战略重心的悄然转移。从卖“铲子”到提供完整的“挖矿系统”，英伟达的胃口，显然不止于芯片。

另一边，微软推出了自己的推理模型Mai-Thinking-1，而腾讯云则将DeepSeek-V4的价格砍到了脚踝，最高降幅97.5%。一个在向上构建护城河，一个在向下拉平竞争起跑线。微软的入局让大模型“军备竞赛”多了一个重量级玩家，而腾讯云的这把“价格屠刀”，无疑会让国内AI应用开发的成本结构发生剧变，进一步催化应用层的爆发。这不再是大模型公司之间的战争，而是平台生态通过成本控制来“招安”开发者的生态战争。

从哈根达斯门店被奶茶品牌柠季接手，到老虎国际暂停境内新开仓，这些看似与AI无关的碎片，拼凑出一幅更广阔的图景：消费市场在存量中寻找新故事，金融监管在规范中划定新边界。而AI，正以前所未有的速度，渗入商业的毛细血管，重塑合作的规则，甚至挑战我们对算力的根本认知。这一天的新闻，没有一件是小事。它们共同勾勒出的，是一个技术、资本和监管在加速磨合的时代轮廓。在这里，没有永恒的王者，只有不断重写的游戏规则。