每兆瓦收入：AI模型工厂的单位经济模型
Revenue per Megawatt & The AI Model Factory
给AI创业者一个可用的单位经济框架：用每兆瓦收入衡量模型公司的盈利性，并理解效率（如GLM-5.3-Flash）如何提升毛利。
An AI model company buys wholesale electricity by the megawatt & resells it as cognitive work.
一家人工智能模型公司按兆瓦购买批发电，并将其转售为认知工作。
“The base cost of compute tends to be around 10 or 13 or $15 million per megawatt. In the case of Anthropic, the revenue has gone as high as $50 million per megawatt. And what that now enables them to do is, hey, if I spend 10 bucks on inference capacity, I actually generate 50 bucks of revenue. And then I can turn around and incrementally spend all of that profit on training.”
“计算的基础成本往往在每兆瓦1000万、1300万或1500万美元左右。以Anthropic为例，其收入已高达每兆瓦5000万美元。这现在使他们能够做到的是，嘿，如果我在推理能力上花费10美元，我实际上能产生50美元的收入。然后我可以转身，将所有这些利润逐步投入到训练中。”
— Dylan Patel, SemiAnalysis, on the Dwarkesh Podcast1
— Dylan Patel，SemiAnalysis，在Dwarkesh播客1上
Gross profit measured as a function of electricity proves model companies can be profitable on a contribution basis.
以电力为函数的毛利润证明，模型公司可以在贡献基础上实现盈利。
Anthropic’s gross margin was −94% in 2024 : $1.94 of compute for every $1 of revenue.2
Anthropic在2024年的毛利率为-94%：每1美元收入对应1.94美元的计算成本。2
By 2025 the corner turned, & Anthropic swung from −94% to a 40-50% gross margin.
到2025年，情况出现转折，Anthropic从-94%的毛利率扭转为40-50%的毛利率。
In 2026 revenue passed cost : $50m per megawatt against a $10-15m cost. Anthropic booked its first profitable quarter, $10.9b of revenue & $559m of operating profit.
2026年，收入超过了成本：每兆瓦5000万美元的收入对1000-1500万美元的成本。Anthropic迎来了首个盈利季度，收入109亿美元，营业利润5.59亿美元。
That 5% operating margin sits well below the 70 to 80% gross margin the megawatt math implies. Training runs & headcount consume the difference.
这5%的营业利润率远低于兆瓦数学所暗示的70%至80%的毛利率。训练运行和员工人数消耗了其中的差额。
Gross profit per megawatt is not just about intelligence, but also efficiency. A model that serves the same intelligence at a fraction of the compute generates more profit per megawatt, even at a lower price.
每兆瓦的毛利润不仅关乎智能，还关乎效率。一个以更少计算提供相同智能的模型，即使在更低价格下，也能产生更高的每兆瓦利润。
GLM-5.3-Flash scores 57 on the Artificial Analysis Intelligence Index : identical to Claude Opus 4.8.3 But it does it on 18 billion active parameters, at a 90 to 97% reduction in cost. Fewer active parameters & less attention compute means more tokens per megawatt.
GLM-5.3-Flash在人工分析智能指数上得分为57：与Claude Opus 4.8.3相同。3但它仅使用180亿个活跃参数，成本降低了90%至97%。更少的活跃参数和更少的注意力计算意味着每兆瓦能产生更多令牌。
The frontier itself keeps climbing in waves. Three jumps of 3 points or more carry half of the 25-point gain from 37 to 63. The efficient models chase a target that resets every quarter.
前沿本身也在波浪式攀升。三次或以上的3点跳跃占据了从37到63的25点增长的一半。高效模型追逐的目标每个季度都会重置。
Which is what the profit per megawatt buys. Patel’s last clause is the hinge : “turn around and incrementally spend all of that profit on training.” The margin funds the factory.
这正是每兆瓦利润所购买的。Patel的最后一句是关键：“转身，将所有这些利润逐步投入到训练中。”利润率资助了工厂。
Nvidia paid $6b to acquire Poolside & invested another $1b on that premise. Model building becomes an industrial process : thousands of experiments across a search space, not artisanal hand-tuning.
Nvidia支付了60亿美元收购Poolside，并在此基础上又投资了10亿美元。模型构建成为一种工业过程：在搜索空间中进行数千次实验，而非手工精细调整。
“The model is the output. The ability to keep building better models, faster & more efficiently each time is the actual innovation. The Model Factory is the compounding asset.”
“模型是输出。持续构建更好模型的能力，每次更快、更高效，才是真正的创新。模型工厂是复利资产。”
— Jason Warner, Poolside4
— Jason Warner，Poolside4
Laguna S 2.1 went from kickoff to release in 52 days.
Laguna S 2.1从启动到发布仅用了52天。
Inference margin funds the factory that makes the next model cheaper to build & more efficient to run.
推理利润为制造下一代模型的工厂提供资金，使其构建成本更低、运行效率更高。
- Dylan Patel on the Dwarkesh Podcast, “Anthropic & OpenAI will have most of the world’s compute by 2028” (Aug 2026). Apple Podcasts ↩︎
- The Information, “Anthropic’s Gross Margin Flags Long-Term AI Profit Questions.” The Information ↩︎
- Artificial Analysis Intelligence Index. Artificial Analysis ↩︎
- Jason Warner, Poolside. X ↩︎
- Dylan Patel 在 Dwarkesh 播客中，“Anthropic 和 OpenAI 到 2028 年将拥有全球大部分计算资源”（2026 年 8 月）。Apple Podcasts ↩︎
- The Information，“Anthropic 的毛利率引发长期 AI 盈利问题。”The Information ↩︎
- Artificial Analysis 智能指数。Artificial Analysis ↩︎
- Jason Warner，Poolside。X ↩︎
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力