$GOOGL · Solving The Money Problem · 2026-11-07
Google 即将到位的大量算力是一项优势：它可以把这些算力投入 Gemini，让 Gemini 缩小与领先基础模型之间的质量差距；因为限制竞争对手获取算力会抬高其成本、损害其用户体验，从而为其他追赶者争取时间。
逐段跳播
我认为这对超大规模云厂商有利，因为它们已经具备相关能力。我认为这对新兴云厂商非常不利，因为它们是租户；这对核心模型也非常不利，因为它们严重缺乏这些能力，而发展又需要这些能力。因此，只要这些领先模型远胜于其他任何人的产品，它们就是需要这些能力的一方。
这让大型公司处于“造王者”的位置，而它们现在面临一个有趣的博弈局面。你能解释一下哪些公司属于大型公司，好让人们理解它们之间的区别吗？可以。Amazon、Meta、Microsoft、Google、SpaceX。
I think this benefits hyperscalers because they already have capabilities in place . I think this is very bad for new clouds because they are tenants, and it's very bad for core models because they are severely lacking in these capabilities and need them to grow.
Therefore, to the extent that these leading models are far superior to anyone else’s products, they are the ones who need these capabilities. This puts large-scale companies in a " kingmaker" position, and they now have an interesting game theory .
Can you explain which companies are large-scale so that people can understand the difference between them? Yes. Amazon, Meta, Microsoft, Google, SpaceX .
你可以决定把全部额外算力用于支持自己的产品。因此，Google 有大量产能将投入使用。它们可以说：“我会把它专门用于 Gemini。”那会发生什么？这实际上会让 Gemini 在质量上追平基础模型。
为什么？因为如果你限制能够产生算力的资产数量，让领先公司可用的空间变少，这些公司就不得不付出更多成本，用户体验会受损，而其他产品就有时间追赶上来。
You can decide to utilize all the additional computing capabilities to support your own products. Therefore, Google has a huge amount of capacity that will enter the service. They could say, "I'll dedicate it to Gemini."
What will happen then? This will actually allow Gemini to catch up in terms of quality compared to the foundational models. Why ? Because if you restrict the amount of assets that can generate computing power, and less space becomes available to leading companies, those companies will have to pay more, the user experience will suffer , and other products will have time to catch up.