3 min read Jun 25, 2025
3分钟阅读2025年6月25日
Fluid compute exists for a new class of workloads. I/O bound backends like AI inference, agents, MCP servers, and anything that needs to scale instantly, but often remains idle between operations. These workloads do not follow traditional, quick request-response patterns. They’re long-running, unpredictable, and use cloud resources in new ways.
流体计算存在用于新的工作负载。I/O绑定的后端如AI推理，代理，MCP服务器以及任何需要立即扩展但经常在操作之间闲置的东西。这些工作负载不遵循传统的快速请求响应模式。它们是长期的，不可预测的，并以新的方式使用云资源。
Fluid quickly became the default compute model on Vercel, helping teams cut costs by up to 85% through optimizations like in-function concurrency.
流体迅速成为Vercel上默认的计算模型，通过优化诸如功能并发之类的优化，团队最多将成本降低了85％。
Today, we’re taking the efficiency and cost savings further with a new pricing model: you pay CPU rates only when your code is actively using CPU.
今天，我们通过新的定价模型进一步采取效率和成本节省：仅当您的代码积极使用CPU时，您才能支付CPU费率。
In the early days of cloud computing, teams ran long-lived servers. You had to manage provisioning, handle scaling manually, and decide what happens during traffic spikes. Over-provisioning cloud resources was common, and idle time meant wasted money.
在云计算的早期，团队运行了长期寿命的服务器。您必须管理配置，手动处理缩放并确定在流量峰值期间发生的情况。过度提供的云资源很普遍，空闲时间意味着浪费的钱。
Serverless changed that. It abstracted away infrastructure configuration and introduced automatic scaling. Each request triggered its own isolated instance.
无服务器改变了这一点。它抽象了基础架构配置并引入了自动缩放。每个请求触发了自己的孤立实例。
But this came with trade-offs. These instances were single-purpose and short-lived. Powerful, but ephemeral. That led to cold starts, duplicated overhead, and underutilized compute. Once again, teams were paying for resources they didn't fully use.
但这是由权衡取舍的。这些实例是单一用途和短暂的。强大但短暂的。这导致了寒冷的开始，重复的头顶和未充分利用的计算。团队再次为他们没有完全使用的资源付费。
Fluid compute breaks away from the traditional one-to-one serverless model. Instead of spinning up a separate instance for each invocation, Fluid compute intelligently orchestrates compute across invocations. Multiple concurrent requests can share the same underlying resources, eliminating cold starts and reusing idle time. This allows I/O bound workloads like AI to run more efficiently.
流体计算从传统的一对一无服务器模型中脱颖而出。流体没有为每个调用旋转一个单独的实例，而是智能地在调用中策划计算。多个并发请求可以共享相同的基础资源，消除冷启动并重复使用空闲时间。这允许I/O绑定的工作负载（例如AI）可以更有效地运行。
The impact was visible immediately. Fluid became the default for AI on Vercel, powering over one trillion invocations. Teams saw up to 90% cost savings by sharing compute across workloads intelligently.
影响立即可见。流体成为Vercel上AI的默认值，为1万亿个调用提供了动力。团队通过智能地共享计算来节省90％的成本。
Fluid improved performance and cost, but there was still room to optimize. Even with high concurrency, there could still be moments where all invocations are waiting on external responses and no code is actively running. During these idle periods, functions stay in memory, doing no work, yet still incur CPU cost.
流体提高了性能和成本，但仍然有优化的空间。即使有很高的并发，仍然可能会有一些瞬间在等待外部响应，并且没有代码积极运行。在这些闲置时期，功能保持在记忆中，没有工作，但仍会产生CPU成本。
Active CPU pricing solves this. It's a new pricing model that charges for CPU only when your code is actively using the CPU.
主动CPU定价解决了这一点。这是一个新的定价模型，仅当您的代码积极使用CPU时，才能为CPU收费。
This aligns pricing with actual usage. Compute costs scale with real work, not just with the time a function is alive.
这将定价与实际用法保持一致。通过实际工作计算成本量表，而不仅仅是函数还活着的时间。
Fluid compute now charges based on three key metrics, each designed to reflect actual resource usage:
流体计算现在基于三个关键指标的费用，每个指标旨在反映实际资源使用情况：
Active CPU reflects the compute time your code is actively executing on a virtual CPU (vCPU). It’s measured in milliseconds, calculated as the number of vCPUs allocated multiplied by the time they’re actively used. Pricing starting at at $0.128 per hour Provisioned Memory covers the memory required to keep a function alive while it's running. It’s measured in GB-hours and billed at a much lower rate (less than 10% of Active CPU), thanks to Fluid’s ability to reuse memory across multiple concurrent invocations. Pricing starting at at $0.0106 per GB-Hour Invocations are counted per function call (just like in traditional serverless) and remain part of the overall pricing
Active CPU反映了您的代码在虚拟CPU（VCPU）上积极执行的计算时间。它以毫秒为单位进行测量，计算为分配的VCPU数量乘以它们的积极使用的时间。定价起价为每小时0.128美元，配置内存涵盖了在运行时保持函数还活力所需的内存。它以GB小时为单位进行衡量，并以低得多的速度（占活动性CPU的10％）计算，这要归功于Fluid在多个并发调用中重复记忆的能力。定价起价为每GB时0.0106美元，每个功能呼叫的计数（就像在传统无服务器中一样），并且仍然是整体定价的一部分
This pricing model in action:
这种定价模型正在行动：
A function running on a Standard machine size at 100% active CPU would now cost ~$0.149 per hour (1 Active CPU GB-Hour + 2 GB of provisioned memory). Previously, this would have cost $0.31842 per hour (1.7 GB Memory × $0.18).
现在以100％活动CPU的标准机器大小运行的功能现在每小时成本约为$ 0.149（1个活动CPU GB小时 + 2 GB的配置内存）。以前，这本来可以花费$ 0.31842每小时（1.7 GB内存×0.18美元）。
Fluid is our proprietary compute platform, built for modern workloads.
Fluid是我们专有的计算平台，专为现代工作量而建。
It simplifies cloud infrastructure while preserving flexibility and performance. Developers can use standard runtimes such as Node.js, Python, and Go, making it easy to run existing code without changes.
它简化了云基础架构，同时保持灵活性和性能。开发人员可以使用标准运行时间，例如node.js，python和go，从而轻松运行现有代码而无需更改。
Fluid powers core product experiences across our platform, including Functions, and recently announced Sandbox. All of them run on the same compute engine, optimized for concurrency, reuse, and efficiency, with unified billing across the stack.
Fluid Power powers遍布我们平台的核心产品体验，包括功能，以及最近宣布的Sandbox。所有这些都在同一计算引擎上运行，并在整个堆栈上进行了统一的计费，以进行并发，重复使用和效率进行了优化。
Active CPU pricing is now enabled by default for all Hobby, Pro, and new Enterprise teams. For existing Enterprise customers, availability depends on your current plan configuration. Most teams will have access right away. Reach out to your Vercel account representative to learn more and enable the new pricing model.
现在，默认情况下，所有业余爱好，专业和新企业团队都可以启用Active CPU定价。对于现有的企业客户，可用性取决于您当前的计划配置。大多数团队将立即访问。与您的Vercel帐户代表联系以了解更多信息并启用新的定价模型。
With Active CPU, Fluid compute automatically optimizes your costs to match actual usage. It reduces waste, scales with real usage, and reflects the way modern apps actually run.
使用Active CPU，Fluid Compute会自动优化您的成本以匹配实际使用情况。它减少了浪费，使用真实用法的缩放，并反映了现代应用程序实际运行的方式。
