← 返回任务列表

Why the Data Center Fight Has Little to Do With AI

363 段 · 1 位说话人 · 原片 35:30
M1
M10:00

今天这期 AI Daily Brief,我们来聊聊,为什么围绕 data center 的争论,其实跟 AI 的关系没你想的那么大。

Today on the AI Daily Brief, why the data center debate has less to do with AI than you think.

在这之前,先看头条:白宫公布了一项审查 AI 模型的计划,不过只有极少数被挑中的人才能了解。

Before that, in the headlines, the White House unveils its plan to vet AI models to a very selected few.

AI Daily Brief 是一档每天更新的 podcast 和视频节目,聊的是 AI 领域里最重要的新闻和讨论。

The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

好啦,朋友们,在我们开始之前,先来几个简短通知。

All right, friends, quick announcements before we dive in.
M1
M11:13

说实话,各位,我本来以为今天的主节目会完全围绕这个新的白宫 AI 框架,因为他们昨天把各家公司都叫去讨论了这件事。但看起来,虽然他们确实有了一个新的安全测试框架,可我们其他人却没法知道具体细节。

To be honest, guys, I thought that today's main episode was going to be all about this new White House AI framework that they hauled in all the different companies yesterday to discuss, but it appears that while they do have a new safety testing framework, the rest of us don't get to learn the details.

周二,几家领先 AI 公司的代表参加了白宫 National Cyber Director 办公室的一场会议,讨论这个新框架,不过对于到底谁能知道相关信息,限制非常严格。

Representatives from the leading AI companies attended a meeting with the White House Office of the National Cyber Director on Tuesday to discuss the new framework, but there are very strict limits on who gets to know anything about it.

事实上,就连那些没被邀请参加会议的 AI 公司,也不会被告知这项政策的内容。

In fact, even AI companies that weren't invited to the meeting won't be told about the policy.

据我们目前所知,OpenAI、Anthropic 和 Google 都派了代表参加,不过完整的受邀名单也没有公开。

Now, as far as we know, representatives from OpenAI, Anthropic, and Google attended, but the exact guest list also hasn't been divulged.

至于这项政策本身,它可以说是六月那份行政命令里提出的自愿安全测试框架的最终落地版本。

As for the policy itself, this is the culmination of the voluntary safety testing framework that was proposed in the June executive order.

公司会被“邀请”在发布之前,把 frontier models 提交给政府,进行最长三十天的安全测试。

Companies are invited, quote unquote, to submit frontier models to the government ahead of release for a maximum of 30 days of safety testing.

预计只有那些最先进、而且可能对国家安全构成威胁的模型才会被纳入,不过我们也不知道这个标准具体是怎么界定的。

Only state-of-the-art models that could pose a threat to national security are expected to be covered, although we also don't know exactly how that criteria is determined.

听起来,政府和 AI 公司在定义上似乎还没能达成一致,这可能会让未来的发布变得更模糊、更不好判断。

It sounds like there is some discussion about the administration and the AI companies failing to reach a common definition, which could muddy the waters for future releases.

在安全测试期间,某些所谓“值得信赖的合作伙伴”可以提前获得这些模型的访问权限。

During safety testing, certain trusted partners, quote unquote, can gain early access to the models.

但还是一样,我们完全不知道这些合作伙伴是谁,也不知道整个系统会怎么运作。

But again, we have no details on who these partners are or how the system will operate.

具体来说,我们不知道这种提前开放会不会只限于美国公司,把盟友国家的政府和外国公司排除在外,就像我们在 Mistral 那次发布中看到的那样。

Specifically, we don't know whether early releases will be limited to U.S. firms and exclude allied governments and foreign companies, as we saw with the Mistral rollout.

很多报道都把重点放在,这项政策会怎么对待 open-weight models。

Much of the coverage has focused on how open-weight models would be treated under the policy.

很多人一直担心,政府会对接近 frontier 水平的开放模型下手,不管是通过这个安全测试框架,还是通过单独出台政策。

Many have been concerned about the administration cracking down on near-frontier open models, either through this safety testing framework or through a standalone policy.

不过有消息人士说,open-weight models 会被排除在这套安全测试机制之外。只是媒体报道里也有一些分歧:The Wall Street Journal 说,豁免只适用于美国制造的开放模型;而几个小时后,Bloomberg 又报道说,中国的开放模型也会被免于测试。

Sources have said, though, that open-weight models would be exempt from the safety testing regime, although there was some disagreement in the press, with The Wall Street Journal publishing that the exemption would only apply to American-made open models, and a few hours later, Bloomberg reporting that Chinese open models would also be exempt from testing.

这最终可能只是个看起来有区别、实际上没区别的区分,因为就算中国模型真的要接受安全测试,他们到底又能做什么呢?

That could ultimately be a distinction without a difference, as even if Chinese models were subject to safety testing, what exactly are they going to do?

目前还不清楚,Washington 要怎么对总部设在中国的公司执行发布禁令。

It's unclear how Washington could enforce a release ban on companies that are based in China.

Twitter 上的 Prince 总结说,毫不让人意外,美国政府选择了那种能让自己在未来应对 AI 可能带来任何风险时,拥有最大操作空间的做法。

Prince on Twitter summed up, to the surprise of no one, the US government went with the approach that provides it maximum flexibility to deal with any risks posed by AI that might arise in the future.

当然,问题就在于,对政府来说最大的灵活性,往往意味着对行业来说最小的可知性。

Now, the challenge, of course, is that maximum flexibility for the government comes with minimum knowability for the industry.

前 FTC 首席技术专家 Neil Chilson 说,政府通过一个秘密流程,提前获得最前沿 AI 的使用权限?

Former FTC Chief Technologist Neil Chilson said, a secret process by which the government gets early access to cutting-edge AI?

这能出什么问题呢?

What could go wrong?

把用来测试这些模型的 benchmark 列为机密,也许是有充分理由的。

There may be good reason to classify the benchmark used to test those models.

但把这个项目到底怎么运作也藏起来,就没有什么正当理由了。

There is no good reason to hide how the program works.

记住这个项目到底在做什么:它让政府提前接触最前沿的工具,而且还可能拥有否决权,决定我们其他人最终能不能用上这些工具。

Remember what this program does: it gives the government early access to cutting-edge tools and a potential veto over whether the rest of us ever use them.

在一个民主国家里,用秘密方式来管理这种可能是我们一生中最重要的技术,绝对不是正确的做法。

Doing that in secret is no way for a democracy to govern what may be the most important technology of our lifetimes.

保密会引来滥用。

Secrecy invites abuse.

规则会随着每一届新政府而变化,而且它们会把国家安全放在经济增长之上,尽管持久的安全其实依赖于强劲的经济。

The rules will shift with each new administration, and they will weight national security over economic growth, even though lasting security depends on a strong economy.

国会必须把任何必要的规则公开写出来,而且要写进法律。

Congress must write any necessary rules in public and in law.

看起来这件事后面肯定还会有更多发展。

Certainly seems like there will be more to this story to come.

现在先说到这儿,我们来聊聊在针对 OpenAI 模型的新一轮第三方测试里曝光出来的更多黑客事件。

For now, let's move on to more hacking incidents that have come to light in new third-party testing of OpenAI's models.

周二,OpenAI 公布了由 UK AI Security Institute 和安全公司 Irregular 进行评估的结果。

On Tuesday, OpenAI published the results of evaluations run by the UK AI Security Institute and security firm Irregular.

两家机构都报告了一起事件:agent 获得了互联网访问权限,并且对一个在线的真实网站实施了攻击。

Each firm reported an incident where the agent gained access to the internet and carried out an exploit on a live website.

Irregular 那起事件看起来像是人为失误,因为这个 agent 的 sandbox 误打误撞地依然保留着互联网访问权限。

Irregular's incident seemed like human error, with the agent's sandbox mistakenly still having access to the internet.

这个 agent 原本被安排在模拟环境里入侵一个虚构的网站。

The agent was tasked with hacking a fictional website within the simulated environment.

但这个 agent 一旦连上了互联网,就入侵了一个同名域名。

However, after the agent found its way onto the internet, it hacked a domain with the same name.

基本上,听起来就是 Irregular 把门没关好,结果也相当可以预见。

Basically, it sounds like Irregular left the gate open with fairly predictable results.

UK AISI 披露的则是一起可能更令人担忧的事件。

The UK AISI disclosed a perhaps more concerning incident.

他们当时在对 Mythos-5 和 GPT-5.6 Sol 进行他们所说的常规网络安全评估,并表示这两个模型都采取了——原话是——“针对真实个人和组织的持续、未经授权的行动”。

They were running both Mythos-5 and GPT-5.6 Sol through what they described as routine cyber evaluations, and said that both models took, quote, sustained, unsanctioned actions directed at real people and organizations.

他们写道,在最严重的案例里,这个 agent 进行了 social engineering,创建了假的网络身份,并利用这些身份向项目维护者施压,要求批准那段代码。

In the most serious case, they wrote, the agent engaged in social engineering, creating fake online identities and using them to pressure the project's maintainer to approve the code.

一名人工维护者发现了问题,并拒绝批准这段恶意代码。

A human maintainer caught and refused to approve the malicious code.

这里有一个很重要但要特别说明的点:这些事件发生在 AISI 的标准网络测试条件下,而这些条件包括完全开放的互联网访问,以及移除所有 guardrails。

Now, an important asterisk to this is that the incidents occurred under the AISI's standard cyber testing conditions, which include full access to the internet and the removal of all guardrails.

他们现在已经联合 Anthropic、OpenAI 以及 MITRE 一起进行独立审查。

They've now engaged with both Anthropic and OpenAI, along with MITRE, for independent review.

对有些人来说,这确实很吓人。

Now, to some, this is scary.

Tenebris 写道,Mythos 并没有对齐。

Tenebris writes, Mythos is not aligned.

在评估环境设置下,如果模型没有很强的网络安全拒绝机制,那它们实际上确实经常会采取重大步骤,在现实世界里实施犯罪。

Models without strong cyber refusals will in fact frequently take significant steps to commit crimes in the real world when presented with eval setups.

但也有人说,这其实就是某种意料之中的表现。

But others said this is just kind of the expected behavior.

Noe Research 的 Shannon Sands 发帖说,这就像你递给一个演员一把上了膛的枪,然后又抱怨说他在拍戏的时候开枪打了人。

Shannon Sands of Noe Research posted, this is like handing an actor a loaded gun and then complaining that they shot someone during a scene.

“天啊,我们明明告诉它这是个 eval,而且没有任何 guardrails,但其实这又是真实在线环境”——这是我能想象到的最假的一种 misalignment。

Oh my God, we told it it was an eval and had no guardrails, but actually it was live is the fakest form of misalignment I can imagine.

另外,虽然中国的开放模型可能不受这些新政策约束,但特朗普政府看起来确实正在准备以间谍活动担忧为由,禁止使用中国的数据中心组件。

Now, while Chinese open models might not be subject to these new policies, the Trump administration does appear to be preparing a ban on Chinese data center components amid espionage concerns.

Reuters 报道说,Federal Communications Commission 正在起草一项禁令,针对某些组件,其中包括 optical transceivers。

Reuters reports that the Federal Communications Commission is drafting a ban on certain components, including optical transceivers.

这种设备会把数字数据转换成光信号,以便通过光纤传输。

These are the devices that convert digital data into light signals for transmission over optical fiber.

现在,数据中心芯片之间的网络连接已经完全是光纤化的了,而因为这些 transceivers 属于大宗通用电子产品,供应里很大一部分都来自中国供应商。

Now, the networking between data center chips is now entirely fiber optic, and because these transceivers are commodity electronics, much of the supply comes from Chinese suppliers.

消息人士称,安装这些组件被视为一种国家安全风险,因为它们可能会被用来植入恶意软件、窃取数据,或者干扰 AI 数据中心的服务。

Sources said that the installation of these components is seen as a national security risk, possibly allowing the injection of malware, data theft, or disruption of service at AI data centers.

目前还不清楚这种风险是否已经实际出现,还是这纯粹只是预防性措施。

It is unclear whether the risk has manifested or this is purely a precautionary measure.

不过,分析人士认为这确实是个合理的担忧。AI 政策分析师 Divyansh Kashyap 就评论说,transceivers 的确存在风险。

Still, analysts believe it's a legitimate concern, with AI policy analyst Divyansh Kashyap commenting, transceivers definitely pose a risk.

随着数据中心建设规模不断扩大,你肯定希望从一开始就确保数据中心供应链是安全的。

As the data center buildout scales up, you want to make sure the data center supply chain is secure from the get-go.

这项禁令出台的同时,众议院中国问题特别委员会也正准备发布一份关于二零二四年曝光的 Salt Typhoon 黑客攻击事件的报告。

The ban also comes as the House Select Committee on China prepares to release a report on the Salt Typhoon hacks uncovered in 2024.

这些攻击针对的是电信公司,而且让 Chinese 间谍在好几年里都能接触到美国高层政治人物的电话通话内容。

These hacks targeted telecommunications companies and allowed Chinese spies to access phone calls from high-profile political figures in the U.S. for several years.

消息人士说,这份报告会认定,中国制造的数据中心硬件是一个薄弱环节,才让这次入侵得以发生。

Sources said that the report will find that Chinese data center hardware was a weak link, allowing the hack to take place.

不过,另一派人认为,把 transceivers 当成攻击路径这件事并不是特别说得通,反而主张这其实是一种保护主义动作,是为了扶持少数几家本土供应商。

On the other side of the argument are those who say that transceivers as an attack vector doesn't make a ton of sense, and instead claim that this is a protectionist move to boost a few domestic suppliers.

当然,就现在来说,美国确实只有少数几家公司在生产这种零部件,而且从新闻出来后的表现看,他们显然正在受益。

Certainly right now, there are only a couple of U.S. companies that make this component, and they appear to be ripping on the news.

最后来看今天的头条,科技公司的财报季还在继续,SpaceX 作为一家上市公司,发布了他们的第一份财报。

Lastly today in the headlines, tech earnings continue as SpaceX delivers their first earnings report as a public company.

这是投资人第一次有机会更深入地看看公司的内部情况,也借此判断它在六月份创纪录 IPO 募资八百六十亿美元之后的增长轨迹。

This was the first chance for investors to take a look under the hood and gauge the company's growth trajectory after they raised $86 billion in a record-setting IPO in June.

综合来看,这份数据其实相当不错。

All things considered, the numbers were pretty good.

这个季度的营收增至七十八亿美元,比去年同期增长了百分之九十二。

Top-line revenue grew to $7.8 billion for the quarter, up 92% from this time last year.

其中,有四十二亿美元来自 Starlink;在过去一年里,卫星互联网用户数翻倍,达到一千两百万,带来了百分之六十六的营收增长。

Of that, $4.2 billion came from Starlink, with satellite internet subscribers doubling to 12 million over the past year to produce 66% revenue growth.

AI 这部分业务通过 Grok 订阅和数据中心租赁两项,一共带来了二十六亿美元收入。

The AI side of the business brought in $2.6 billion across both Grok subscriptions and data center rentals.

这比一年前增长了三倍,而那时候 SpaceX AI 还没有签下数据中心相关的协议。

That's triple the revenue from a year ago before SpaceX AI had signed data center deals.

值得注意的是,Cursor 那笔交易目前还没有正式完成,所以要到今年晚些时候才会体现在财报里。

Notably, the Cursor deal hasn't closed quite yet, so it won't appear in earnings until later this year.

从支出这一边来看,SpaceX 这个季度与 AI 相关的 CapEx 是一百五十八亿美元。

On the other side of the ledger, SpaceX had $15.8 billion in AI-related CapEx for the quarter.

这占到了总 CapEx 的百分之八十六,也就是说,数据中心现在显然比一个完整成型的太空项目还要更烧资本。

That represented 86% of total CapEx, implying that data centers are now much more capital intensive than a fully fledged space program.

整体来看,SpaceX 这个季度录得五亿四千一百万美元的净亏损,不计 CapEx 的话,相比第一季度十亿美元的亏损已经有所收窄,但还不足以平息外界对 SpaceX 是一台巨型烧钱机器的批评。

Overall, SpaceX recorded a $541 million quarterly net loss excluding CapEx, down from a billion-dollar loss in the first quarter, but not enough to quiet critiques that SpaceX is a massive cash incinerator.

不过,跟 SpaceX 的很多事情一样,大家讲的故事更多还是这家公司要去哪里,而不是到达那里要花多少钱。

Still, as with everything SpaceX, the narrative is about where the company is going rather than how expensive it will be to get there.

在财报电话会上,CFO Brett Johnson 表示,未来六个月里,公司还有六十七亿美元的 cloud services 收入已经在储备中。

During the call, CFO Brett Johnson reported that the company has a further $6.7 billion in cloud services revenue in the pipeline over the next 6 months.

这里大概率指的是和 Anthropic 以及 Google 的交易。

Presumably referring to the Anthropic and Google deals.

他还说,整合 Cursor 之后,到今年年底,SpaceX AI 的 run rate 有望达到一千亿美元。

He also said that the integration of Cursor should take SpaceX AI to a $100 billion run rate by the end of the year.

这会比他们二零二五年一百八十七亿美元的营收高出五倍以上。

That would be more than a 5x increase from their $18.7 billion in revenue for 2025.

Elon Musk 还补充说,十二月份做到一千亿美元 ARR,这不是一个问号。

Elon Musk added, the $100 billion ARR in December is not a question mark.

如果我们基本上什么都不做,这个数字也能实现。

That's what we would achieve if we basically did nothing.

所以我觉得,实际数字可能还会更高。

So I think it may be higher than that.

它很可能确实会比这更高。

It probably will be higher than that.

另一个大问题是,接下来 xAI 的数据中心建设到底会呈现出什么样的形态。

The other big question is what the shape of xAI's data center buildout will look like moving forward.

Johnson 说,到了第三和第四季度,CapEx 会基本保持相对平稳,这样 xAI 全年的数据中心支出大概会到六百亿美元左右。

Johnson said that CapEx will remain relatively flat through the 3rd and 4th quarters, bringing xAI to around $60 billion in data spend for the year.

这样一来,他们就明显落后于 Google 和 Amazon,这两家公司今年各自都承诺要投入两千亿美元;不过也开始逐渐逼近 Oracle,后者预计今年会投入九百五十亿美元。

That puts them significantly behind Google and Amazon, who are each committing to $200 billion in spend this year, and starting to creep up on the heels of Oracle, who have forecast $95 billion for this year.

不过,Elon 也补充说,我们相信,我们正在以比任何人都更快的速度建设 AI 算力,而且我们的 AI 模型也在显著提升。

Still, Elon added, we're building AI compute capacity at a scale faster than anyone else, we believe, and we're significantly improving our AI models.

总体来说,这次财报是超出预期的,但大概还不足以让那些持批评态度的人转而支持。

Overall, this was an earnings beat, but probably not enough to get anyone who's critical on board.

截至上周,这只股票相比历史高点最多已经下跌了百分之四十五,明显跌破了一百三十五美元的 IPO 发行价。

The stock was down as much as 45% from its all-time high as of last week, falling meaningfully below the $135 IPO price.

前几天这只股票本来一直在重新往上冲,周初一开盘就涨了百分之十,但财报出来之后,股价就掉头向下,盘后交易里跌了百分之七。

It had been ramping back up in recent days, gaining 10% to begin the week, but following earnings, the stock rolled over and lost 7% in after-hours trading.

更重要的是,这次财报也标志着第一轮限售期结束,早期投资者第一次有机会在市场上卖出,这也会进一步检验当前这个价格水平。

Importantly, this earnings also marks the end of the first round of lockups, giving early investors their first chance to sell into the market and further test the price level.

还有更多值得关注的内容,不过现在,今天的头条就先到这里。

More to keep an eye on, but for now, that is going to do it for the headlines.

接下来,进入今天的主节目。

Next up, the main episode.
M1
M111:42

我每天都会在这个节目里聊 AI 的潜力和 AI 现实之间的能力鸿沟。

I cover the capability gap between AI potential and AI reality every day on this show.

大多数公司现在还在摸索,到底该怎么开始。

Most companies are still figuring out how to start.
M1
M113:06

欢迎回来,这里是 AI Daily Brief。

Welcome back to the AI Daily Brief.

今天我们要聊的是 AI data center 争论的最新进展,而且我想提出一个观点。

Today we are talking about the latest in the AI data center debate, and I'm gonna make a contention.

你大概已经从这期节目的标题里看到了,我认为 data center 的争论,和 AI 本身的关系,其实远没有第一眼看上去那么大。

As presumably you've already seen in the title of this show, that the data center debate has a lot less to do with AI itself than it appears on first glance.

当然,要是说这件事跟 AI 完全没关系,那肯定也说得太过了。

It would certainly be going too far to say that it has nothing to do with AI.

对有些人来说,data center 只是一个他们不喜欢的东西最容易看见、最直观的体现。

For some, data centers are the easy visual manifestation of a thing they don't like.

但我认为,这其实只是在这波反弹情绪里,一个说实话相当小的组成部分。

But I believe that that is only one frankly kind of small component of the backlash.

我觉得这件事更核心的,其实是人们失去了自主感,也觉得一个并不是他们自己选择的世界,正在被强加到他们头上。

What I think it is way more about is a loss of agency and a feeling that people have that a world that they didn't choose is being imposed upon them.

出于各种各样的原因,现在很多人都觉得自己没法掌控自己的生活。

For all sorts of different reasons right now, people feel unable to control their lives.

当然,这背后是一个比单纯担心 AI 更宽泛的政治背景。

And this is, of course, a broader political setting than just AI fears.

他们对任何掌权的人都抱着怀疑态度,也不相信这些身处权力位置的人,会把他们的利益放在优先位置。

They feel skeptical of anyone in any position of power and don't believe that those people in power have their best interests as a priority.

在这种背景下,人们也不再相信科技是为了服务他们而被设计出来的。

And within that, people no longer believe that technology is designed to serve them.

事实上,越来越多的人觉得,科技只是为了服务那些打造它的 oligarchs。

More and more, in fact, they think it's just about serving the oligarchs that build it.

于是,data centers 就同时成了压垮骆驼的最后一根稻草,也成了一个看得见、摸得着、可以拿来对抗的目标。

Data centers then represent both the last straw and something tangible to fight.

现在,这些建设 data center 的公司,对当地群体所处的这个大环境,其实没多少控制力。哪怕像 Meta 这样的公司,某种程度上可能也参与塑造了这个环境,但它们至少能控制自己怎么去跟这种环境互动。而它们在这方面,做得真的是糟糕透顶。

Now, the companies building these data centers don't have much control over the context that local constituencies are operating from, even if in the case of a company like Meta, they might've had a hand in shaping it, but they do have control over how they interact with it, and they are doing an absolutely abysmal job.

这周,记者 Jasmine Sun 写了一篇非常精彩的新文章,长达六千字,随后她还接受了 The New York Times 的 Ezra Klein 的采访,里面探讨了很多这类主题。

This week we got a great new 6,000-word piece from journalist Jasmine Sun and a follow-up interview with Ezra Klein of The New York Times that explores a lot of these themes.

不过,为了先给最近的一些新闻补一点背景,对于那些没有一直密切关注这件事的人,我们先来聊几条最近上了头条的 data center 新闻。

However, to situate a little bit of recent news around this, for those who haven't been paying attention closely, let's talk a couple of data center stories that have made headlines recently.

其中一个故事,居然是来自 Texas。

One of those stories comes from, of all places, Texas.

Texas 州长 Greg Abbott 通过非常严格的新一轮能源审计,给 data center 的建设踩了刹车。

Texas Governor Greg Abbott has thrown the brakes on data center construction with stringent new energy audits.

Abbott 州长已经指示 Texas Public Utility Commission 和全州电网运营方 ERCOT,开始对新的 data center 提案进行核查和审计。

Governor Abbott has instructed the Public Utility Commission of Texas and the statewide grid operator ERCOT to begin verifying and auditing new data center proposals.

他在写给公用事业部门的信里说,这些措施是必要的,用他的话讲,是为了“保持电网稳定且可靠”。

In his letter to the utilities, he wrote that the measures were necessary to, quote, keep the grid stable and reliable.

现在,Texas 的 data centers 将被要求提供它们获得过哪些州和地方激励政策的信息,披露它们会在多大程度上依赖本地电网,还要提交预计用水量以及供水方案。

Texas data centers will now be required to provide information on state and local incentives they've received, disclose how much they will rely on the local grid, submit expected water consumption and a supply plan.

它们还需要准备一套方案,说明将如何追踪并回应社区关切,其中也包括像噪音投诉这样的事情。

They will also need to prepare a plan for how they will track and respond to community concerns, including things like noise complaints.

目前还不清楚,这些繁琐程序会在多大程度上拖慢审批速度,也不清楚这项政策是不是想事实上变成一种暂停令。

Now, it is not clear how much this red tape will slow down approvals and whether the policy is intended as a de facto moratorium.

但可以确定的是,Texas 正在承受一波新的 data center 许可证申请潮,压力越来越大。

What is clear is that Texas is straining under a wave of new data center permit applications.

Texas 是仅次于 Virginia 的第二大 data center 市场,这个州现在正经历一波前所未有的申请潮。

Texas is the second-largest data center market behind Virginia, and the state is experiencing an unprecedented wave of applications.

ERCOT 目前收到的新并网申请总量有四百七十四吉瓦,其中百分之九十都来自 data center。

ERCOT currently has 474 gigawatts of new connection requests, with 90% of those requests coming from data centers.

这个数字在过去六个月里翻了一倍,现在已经是电网历史峰值容量纪录的五倍。

This figure has doubled in the past 6 months and is now 5 times the record peak capacity of the grid.

换句话说,到这个时候,想把这些项目全部建出来,已经是完全不可能的了。

In other words, it is at this point completely impossible for all of this to get built.

实际上,这个积压量大概相当于美国当前总装机容量的四倍。

This backlog is in fact around 4 times the total current installed capacity of the U.S.

就算 ERCOT 能把供电能力扩大到四倍,去服务所有这些项目,chips、components 和劳动力方面的瓶颈,也会让这些项目不可能全部建成。

Even if ERCOT could quadruple their power supply to service all these projects, bottlenecks in chips, components, and labor would prevent everything from getting built.

现在,一个可能的解释是,同一个项目提交了多份申请。

Now, one potential explanation is multiple applications from the same project.

很多时候,data center 开发商并不知道哪一份申请会获批,所以他们会多报几份,来提高至少有一份能推进的概率。

Often, a data center developer won't know which application will be approved, so they submit multiples to increase their chances of at least one going ahead.

而且,那种已经拿齐审批、可以直接开建的土地,溢价也特别高,所以有些开发商其实只是先去申请,打算之后再把地卖掉。

Land that's ready to build with full approvals also attracts a huge premium, so some developers are simply making applications with the intent to sell the land at a later date.

这一切最终造成了一场官僚体系里的噩梦。

This has all resulted in a bureaucratic nightmare.

因为 ERCOT 根本没有什么真正有效的办法,去筛选出哪些项目是真的、而且大概率会推进。

Where ERCOT has no real means to sift the genuine projects that are likely to go ahead.

处理这个问题,可能才是 Governor Abbott 下达这项指令的部分原因,而不一定是想偷偷搞一个 data center moratorium。

Dealing with this issue could be part of the reason for Governor Abbott's directive, rather than a desire to sneak in a data center moratorium.

不过,结果是一样的:Texas 的新申请现在已经暂停,直到审计完成为止。

Still, the impact is the same, with new applications in Texas halted until audits can be completed.

Interconnected Capital 的 Kevin Xu 在这里提出了一个很明显的观点:Texas 是 AI data center 的核心战场,而且很可能会超过 Virginia,成为全美第一大 data center 州,所以那边任何扰动都很重要。

Kevin Xu from Interconnected Capital makes the obvious point here: Texas is ground zero for AI data centers, poised to overtake Virginia as the number one data center state, so any disruption there is important.

New York Times 的经济记者 Lydia DePillis 写道,随着美国其他地方对 data center 越来越怀疑,大家给出的答案一直都是:那就把它们建在 Texas。

New York Times economy reporter Lydia DePillis wrote, as the rest of the country has grown skeptical of data centers, the answer has been just build them in Texas.

而今天,Greg Abbott 暂停了新的申请,说这些申请有可能把电网压垮。

Well, today, Greg Abbott paused new applications, saying they threaten to overwhelm the grid.

再说一个类似的例子,上个月 New York 州长 Kathy Hochul 签署了该州首个 data center moratorium。

Now, one similar story from last month was New York Governor Kathy Hochul signing that state's first data center moratorium.

在签署前后,Hochul 说,New York 州各地,主要是一些小社区,出现了非常多申请,地方政府其实不太确定该怎么应对。

Around the signing, Hochul said, there are so many applications in primarily small communities across the state of New York that the localities aren't quite sure how to manage this.

她接着说,他们来找我帮忙,一开始我还觉得这更像是地方层面的决定,但后来我意识到,他们没有足够的谈判能力、影响力和资源,没法去谈出最好的利益条件,也没法确保这些公司要么自带电力,要么向电网支付额外溢价来覆盖成本。

Continuing, she said, they've asked me for help, and at first I thought this is more of a local decision, but then I realized they don't have the negotiating ability, the clout, the wherewithal to be able to negotiate the best benefits or make sure that these companies are either bringing in their own power or paying a premium into the grid to cover the cost.

Interconnected Capital 做的一个 data center moratorium tracker,目前追踪到了二百一十九个地方性的 moratorium。

A data center moratorium tracker built by Interconnected Capital is currently tracking 219 local moratoriums.

这些大多数都不是州一级的。

Most of these are not state level.

他们还在跟踪二十三项州级法案。

There are 23 state bills that they're following.

但光看这张地图就很明显,这已经是一个范围很大、而且还在不断扩大的担忧领域。

But it's clear just looking at the map that this is a large and growing area of concern.

顺带一提,data center moratorium 有个很有意思的挑战,就是它会给更大的整体经济带来一种 free rider problem。

Now, as an aside, one interesting challenge with data center moratoriums is that they introduce a free rider problem to the larger economy.

那些不想承接 data center、而且采取了相当强硬措施去阻止它们落地的州,并不是说他们打算在全州范围内不用 AI。

The states that don't want to host data centers and are taking pretty strong measures to prevent them being built aren't talking about not using AI on a statewide level.

他们还是想要这些服务,只是不想为背后的基础设施埋单。

They still want the service provision, they just don't want to support the infrastructure.

实际上,当 Illinois 州长 J.B. Pritzker 签署他们那个 AI 法案时,他还特别强调了一点:Illinois、New York 和 California 三个州虽然人口只占百分之二十,却占了 AI 市场的百分之四十。

Indeed, when Illinois Governor J.B. Pritzker signed their AI bill, he made a big deal over the fact that Illinois, New York, and California are 40% of the AI market despite only having 20% of the population.

基本上,我们现在正在走到这样一个阶段:美国一些最大的市场,希望由别的州去承担运行 AI 基础设施所带来的外部成本。

Basically, we're now getting to the point where some of the largest US markets want other states to deal with the externalities of running AI infrastructure.

如果你去看看社交媒体,围绕 data center 的一些讨论,真的会让人觉得离谱得不行。

Now, if you just look on social media, some of the discourse around data centers feels absolutely insane.

其中有一个人,算是冒出来成了一个也许有点出人意料、但相当理性的声音,那就是 Taylor Lorenz。

One person who has emerged as a perhaps surprisingly sensible voice is Taylor Lorenz.

现在,Taylor 跟科技行业的长期关系,远远不止这一期节目能涵盖的范围。不过这么说吧,她现在大概也不一定会被邀请去很多 Silicon Valley 的派对了。

Now, Taylor's long-term relationship with the tech industry goes way beyond the scope of this particular episode, but suffice it to say that she's not necessarily getting invited to a lot of Silicon Valley parties right now.

但与此同时,她也越来越强硬地主张,简单说,反对 data centers 的正当理由其实已经够多了,所以那些更夸张、更耸动、带点阴谋论味道的说法,实际上正在妨碍理性的讨论。

And yet, with increasing force, she is arguing, in short, that there are enough good reasons to be opposed to data centers that some of the more dramatic, hyperbolic, and conspiracy theory-type claims are actively getting in the way of reasonable discourse.

她还提到 Threads 上的一条帖子,那条帖子说 data centers 是——原话是——“为了军事用途、监控国家、通过 eminent domain 进行圈地,以及加速白人基督教民族主义议程下的世界末日。”

Pointing to a post on the Threads service that argued that data centers are for, quote, military use, the surveillance state, land grabs through eminent domain, and to speed up Armageddon for the white Christian nationalist agenda.

哦,对了,还有一些人觉得这其实是个幌子,是为了囤积我们的水,好去供应 billionaire bunkers。顺便说一句,那条帖子有一千四百个赞。

Oh, and some folks believe that it's a cover for hoarding our water to supply the billionaire bunkers, a post which, by the way, had 1,400 likes.

Taylor 说:“我担心,我们社会里有很大一部分人,正在经历某种 data center psychosis。”

Taylor said, I'm worried vast swaths of our population are experiencing data center psychosis.

有人回复说:“我就在一个 data center 工作,那里服务器数量其实挺少的,主要是支持科学研究。”

Someone responded, I work at a data center with a pretty small number of servers that support scientific research.

“这跟那些 AI hyperscale data centers 完全不是一回事。”

It is vastly different than the AI hyperscale data centers.

“还有人在这上面跟我说,他们宁可死,也不愿意做我这份工作。”

I had someone on here say they would rather die than do my job.

Taylor 后来在一个 podcast 里继续谈这个想法。她说:“我最近上了一个 podcast,因为我在 LA 的一个社区团体里,大家发现 downtown LA 有一个 data center。”

Continuing her thought on a podcast, Taylor said, I went on a podcast recently because I'm in this community group in LA, and they found out that there's a data center in downtown LA.

“它不是 hyperscaler,但评论区里有人会说,‘难怪我最近总觉得不对劲’,或者‘难怪我的狗得了癌症’。”

It's not a hyperscaler, but the comments are like, this is why I've been feeling off, or this is why my dog has cancer.

这倒不是说,像 Colossus 以及这里面一些项目,就不存在严重的环境问题。

That's not to say that there aren't serious environmental concerns with Colossus and some of these.

施工可能会扰动地下水位。

Construction can disrupt the water table.

好,但我们能不能认真地就这个问题展开真正的讨论。

Okay, but let's have an actual conversation about it.

现在所有人都已经跑偏了。

Everybody has lost the plot.

这里面完全没有细致度可言,而且你去跟很多这样的人聊,他们甚至都说不清 data center 到底是什么。

There's no nuance, and if you talk to a lot of these people, they can't even explain what a data center is.

很多针对 data centers 的愤怒,其实是来自住在 Brooklyn 的人。

A lot of the outrage about data centers is coming from people that live in Brooklyn.

你们这是在不断给人喂垃圾信息、制造情绪和流量,但并没有真的推动这些问题往前走。

You're feeding people slop and churn, and you're not moving these issues forward.

而且我们也没有在讨论怎么用可持续的方式去建设这些东西。

And we're not talking about sustainable ways to build it.

另外,为了再次说明 Taylor 并不是什么披着羊皮的狂热亲-data center 活动人士,她在另一条 tweet 里说,绝对没有任何人会希望自家后院旁边新建一个 hyperscale AI data center。

Now, reinforcing the point that Taylor is not some rabid pro-data center activist in sheep's clothing, in another tweet she said, absolutely no one wants a new hyperscale AI data center in their backyard.

就算你不觉得 moratorium 是推动改变的最佳政策,几乎所有人都讨厌这些 data centers,因为它们是普通人能直接碰到的 big tech 最具实体感的体现。

Even if you don't think a moratorium is the best policy to enact change, literally everyone hates the data centers because they are the most physical manifestation of big tech people are encountering.

我不认为我们应该去鼓励人们对 data centers 的这些妄想。

I do not think we should encourage people's delusions about data centers.

那样最后只会适得其反。

That's just going to backfire.

很多人相信的就是赤裸裸的阴谋论,而这反而会让科技公司更容易把批评者一笔带过、不当回事。

A lot of people believe straight-up conspiracy theories, which makes it easier for tech companies to write critics off.

不喜欢 DCs,本来就已经有很多完全站得住脚的理由了。

There are plenty of legit reasons to not like DCs.

而且事实上,围绕 data centers 的政治,正变得越来越诡异。

And indeed, the politics of data centers are getting weirder and weirder.

当然了,左派那边有一种看法,是把这件事视作 Trump 那种大规模腐败、劳民伤财的荒唐工程之一。

There is, of course, a dimension of it on the left that views it as one of Trump's great corrupt boondoggles.

《The New York Times》最近就发了一篇文章,标题非常有竞争力,堪称最能精准投喂进步派情绪价值的那种左派爽文,题目叫《Trump 的 AI 主导地位愿景,伴随着严重的空气污染》。

With a headline that is a major contender for the juiciest left meat for the progressive crowd, The New York Times recently published a piece called Trump's Vision for AI Dominance Comes with Major Air Pollution.

但这又并不是一个单纯的左右之争。

And yet this is not a left-right issue.

事实上,Molly Taft 最近在 Wired 写了一篇文章,标题叫《How Data Centers Broke American Politics》,副标题可能是我见过最离谱的之一:《Unabomber、Steve Bannon 身边的 Tech Guy,以及 Bernie Sanders,让我看懂了二零二六年那场大规模 data center 反弹》。

In fact, Molly Taft recently wrote in Wired a piece called How Data Centers Broke American Politics, with arguably one of the most insane subheaders I've ever seen, What the Unabomber, Steve Bannon's Tech Guy, and Bernie Sanders Taught Me About the Great Data Center Backlash of 2026.

而这里的重点当然是,左右两边正在因为 data centers 走到一起,形成一些非常奇怪的政治同盟。

And the point here is, of course, that the left and right are coming together to make very weird bedfellows around data centers.

确实,之前《New York Times》还引用过喜剧演员 Charlie Berens 的话,进一步强化了这个观点:反对 data center,几乎是继啤酒之后最跨党派的议题。

Indeed, previously quoting comedian Charlie Berens, the New York Times reinforced this idea of opposition to data centers being the most bipartisan issue since beer.

早在六月,Theo Von 也说出了这种情绪,他说,老兄,没人想要 data center。

Giving voice to this sentiment was Theo Von back in June, who said, nobody wants a data center, dude.

而那些想要它们的人,在我看来,都有点邪恶。

And the people that want them, to me, they seem kind of evil.

没人想要这东西。

Nobody wants this.

这些公司里,总有一家会掌控所有这些信息。

One of these companies is going to own all this information.

接下来就会出现某种社会信用分,或者情绪信用分,然后 AI 会试图变成我们的新神。

There's going to become this social or emotional credit score, and then AI is going to try to become our new god.

如果你听到这里,已经开始渴望一些更有思考的讨论了,那你并不孤单。

Now, if at this point you are hungering for some thoughtful discourse, you are not alone.

幸运的是,我们有 Jasmine Sun。她既为自己写作,也给 The Atlantic 写稿。她刚刚发表了一篇六千字的文章,讲的是 Midwest 的 data center 政治。这篇文章基于她花了十天时间到处走访,然后——我知道这很让人震惊——真的去跟人们聊了聊。

And luckily we have Jasmine Sun, who writes both for herself and for The Atlantic, who just dropped a 6,000-word piece on the politics of data centers in the Midwest based on 10 days of going around and— shocking, I know— actually talking to people.

她把这篇文章命名为《No Data Centers in My Backyard》,想要探究大家真正担心的到底是什么。

She called the piece No Data Centers in My Backyard and sought to explore what the concerns really are.

在文章开头,Jasmine 写道,我采访了工会领袖、房地产经纪人、活动人士和地方官员,还有从 Kathy Hochul 到 Abdul El-Sayed 的政界人士。

Setting up the piece, Jasmine writes, I talked to union leaders and real estate brokers, activists and local officials, politicians from Kathy Hochul to Abdul El-Sayed.

AI 的大规模建设,出现在一个极度不信任的环境里,而且它又刚好和人们原本就有的担忧严丝合缝地对上了,比如高风险科技泡沫,以及政治中的 dark money。

The AI buildout is showing up in an environment of extreme distrust, and it slots perfectly into existing worries about risky tech bubbles and dark money in politics.

从理性上讲,我常常同意支持者的观点。

Rationally speaking, I often agreed with proponents.

但从情感上讲,我更能共情反对者。

Emotionally, I empathized with the opposition more.

后来在接受 Ezra Klein 的后续采访时,Jasmine 讲了几个她反复听到的大问题。

Now, in a follow-up interview with Ezra Klein, Jasmine went through a few of the big issues that she heard over and over.

其中一个是透明度。

One was around transparency.

更具体地说,很多敌意都来自这样一个事实:这些交易是在关起门来进行的,而这件事最典型的象征,就是 hyperscaler、data center 建设方和地方政客之间签的 NDA 协议。

And specifically, a lot of animosity generated by the fact that these deals were happening behind closed doors, with the big symbol of this being NDA agreements between the hyperscalers and data center builders and the local politicians.

Jasmine 对 Ezra 说,基本上,经常会出现这样的情况:市议会那边会隐约感觉到,可能有一个大型开发项目要来了,但因为有保密协议,也就是 NDA,议员们没法透露这个项目其实是 data center。

Jasmine told Ezra, basically what would happen oftentimes is there would be some sense starting in city council that maybe a big development project was going to show up, but because of the non-disclosure agreements, NDAs, the council members would not be able to disclose that necessarily it was a data center.

他们也没法透露客户会是谁,甚至连项目规模都不能说——比如说,这东西到底会消耗多少电?

Necessarily who the customers were going to be, or even the size of the project—like, how much electricity is this actually going to consume?

但各种风声还是会慢慢传开。

But whispers would start to get around.

我当时采访过一位建筑工会的副总裁,他说,有一句古老的爱尔兰谚语:三个人之间要想保住秘密,唯一的办法就是干掉其中两个。

I was talking to a VP of a construction union and he was saying, there's an old Irish saying that the only way to keep a secret between three people is to kill two of them.

所以他的意思是,这些开发商一出现,就会去跟总包聊,总包再去跟各个分包商聊,分包商又会去跟他们手下的工人聊。

So he's saying, when these developers show up, they talk to the general contractor, the general contractor talks to all their subcontractors, the subcontractors talk to all their workers.

对,也许在流程的每个环节,大家都签了 NDA,但风声还是会传出去。

Yes, maybe everyone is signing NDAs at every part of the process, but whispers get around.

而一旦风声传开,社交媒体上的帖子就会出现,谣言也会开始冒出来。

And as soon as whispers get around, you start to get social media posts, you start to get rumors.

而市议会因为受这些 NDA 约束,就失去了抢在社交媒体叙事成形之前先发制人的能力。

And the city council, because they are beholden to these NDAs, they lose the ability to get ahead of the social media narrative.

这些地方政府官员反复跟我说的一点就是:我们没法赶在社交媒体前面发声,因为我们签了 NDA。

Something I repeatedly heard from these local government officials was we could not get ahead of social media because we had signed an NDA.

有意思的是,当你既读了 Jasmine 的原文,又听了她和 Ezra Klein 的采访之后,你会发现 Jasmine 实际上是在说,关于 data center 的各种批评,你听到的那些说法,每一种都抓住了部分真相。

Now, what's interesting when you read both Jasmine's source piece and listen to the interview with Ezra Klein is that Jasmine is effectively arguing that all of the different critiques around data centers that you've heard have some part of the story.

比如说,她认为用水确实是个问题,不过随着 data center 用水技术的变化,这个问题的重要性可能有所下降;而对能源使用的担忧,则是明确存在、而且眼下就很现实的问题。

For example, while she thinks that water is, yes, an issue, but perhaps on a lower ebb, as the technology around water in data centers has changed, concerns around energy usage are clear and present.

她写道,假设一切都运转正常,一个 data center 在完全投入运营之后,并不会产生大量空气和水污染。

She writes, once a data center is fully operational, there's not a ton of air and water pollution, assuming that everything is working correctly.

它们算是相对干净的设施。

They're relatively clean facilities.

但有些人担心的是它们的用电量。

But some folks are concerned about the electricity use.

他们的意思是,对,也许 data center 用不了那么多水,也许 data center 造成的污染也没那么大,但为了给它供电,他们接下来还要建的那些新发电厂呢?

They're saying, yeah, maybe the data center doesn't use that much water, maybe the data center doesn't pollute that much, but what about all of these new power plants that they're going to build in order to power it?

另一类担忧,基本上就是过去那些失败项目留下来的阴影。

Another set of concerns was basically the shadow of failed projects past.

她提到了 Foxconn 的经历。她说,你会看到一家大型科技公司来到一个很小的社区——这里指的是 Wisconsin 的 Mount Pleasant,一个两万八千人的城市——从当地拿到几亿美元的基础设施投资和税收补贴,承诺带来一万三千个高薪制造业岗位,结果因为合同设置不对,最后又撤了。

She points to the experience of Foxconn, where she says you have a big tech company show up in a very small community, in this case Mount Pleasant, Wisconsin, a city of 28,000, get hundreds of millions of dollars in infrastructure investment and tax subsidy from the town, promise 13,000 high-paying manufacturing jobs, and then pull out because the contract wasn't set up correctly.

他们后来决定,其实并不想在 Wisconsin 建一大堆平板电视厂,而这个镇子就只能自己承担后果。

They decided they didn't actually want to build a bunch of flat-screen TVs in Wisconsin, and the town was left on the hook.

前面已经在电网和道路上投了这么多钱。

Having invested all this money in the grid and in roads.

我记得最后他们大概只带来了 一千 个工作岗位。

They got, I think, 1,000 jobs in the end.

Foxconn 到现在还在偿还这一路累积下来的债务。

Foxconn is still paying back all of this debt that has been accumulated.

当然,这显然不是常态,因为这些社区里的大多数,过去都没有经历过这种大型失败的基础设施项目。但其他社区听到这类故事,也会把它算进担忧里。

Now, obviously this isn't the norm, as most of these communities haven't had some big failed infrastructure project like this in their past, but other communities hear those sort of stories and it adds to the pile.

但那 AI 本身呢?

But what about AI itself?

我一开始的观点是,大家对 data center 的反弹,其实并没有你想的那么和 AI 直接相关。

My contention at the beginning was that the data center backlash wasn't as much about AI as you might assume.

Jasmine 总结她观察到的普遍态度时说,比起社交媒体上那种“AI 很邪恶”的说法,更多的是把它看成一个没那么重要的东西,不值得为它承受这么多顾虑和代价。

And summing up the average attitudes that Jasmine found, it was less the sort of AI is evil rhetoric that you hear on social media, and instead more a view of it as something fairly insignificant that didn't justify all the concerns.

Jasmine 跟 Ezra 说,我去问这些组织者,你们会用 AI 吗?

Jasmine told Ezra, I went in and asked these organizers, do you guys use AI?

你们觉得它有用吗?

Do you find it useful?

我发现,这里面很多人都会说,会啊,我用它来起草邮件,或者做个 meme。

And what I found was that a lot of these folks did say, yeah, I've used it to draft an email or make a meme.

他们并不是在否认 AI 可能完全没有任何用处,但他们显然也不觉得它像汽车、能源和住房那样,是必不可少的。

They're not denying that AI might have any possible utility at all, but they clearly didn't see it as essential in the way that cars, energy, and housing are essential.

他们觉得它更像一个小玩意儿,一个玩具。也许它是有这些风险,也许也牵扯到就业这些问题,但归根到底,他们的态度就是,这东西没那么有用。

They saw it as a widget, a toy, and maybe there are these risks, maybe there's this job stuff, but fundamentally they were like, this thing is just not that useful.

至少在我个人生活里,我真看不出来这东西凭什么能支撑这么夸张的估值。

I don't really see in my personal life how this could justify these gigantic valuations.

但结果发现,真正的问题其实也不是这个。

But it turns out that's not really the thing.

大家对 AI 的担心,并不在于 AI 本身那些具体内容。

The concern with AI is not any of the substance of AI itself.

对大多数人来说,这不是在批评版权问题,也不是在担心艺术家的工作。

For most, it's not a critique of copyright or a concern about artists' jobs.

而是因为 AI 成了那个被强加到他们头上的新东西。

It's about AI as the new thing that is being imposed upon them.

Jasmine 说,你经常会从 AI 的批评者那里听到一句话:为什么这东西非得硬塞给我们?

Said Jasmine, the phrase you hear a lot from AI critics is, why is this being shoved down our throats?

我觉得,过去六个月里出现的很多公众对 AI 的反弹,并不能简单地解释成大家觉得这项技术完全没用。

I think that a lot of the public backlash to AI that has arisen over the past 6 months is not explained by people thinking that the technology has no use at all.

也不能解释成他们是在担心 large language models 的某些具体技术特性,会导致 rogue AI、misalignment 之类的问题,或者别的什么。

It's not explained by their being worried about specific technical properties of large language models that might lead to rogue AI or misalignment or whatever.

更像是,AI 成了一个化身,代表着一小撮 Silicon Valley 亿万富翁可以不经过大家同意,就把他们对世界的想象强加给所有人。

It's AI as sort of an avatar for a small group of Silicon Valley billionaires' ability to impose their vision of the world onto everybody else without their consent.

确实,她在那条公告 tweet 里特意强调的原文段落是这样写的:我的这些对话让我开始怀疑,data center 交易里的技术性内容,可能大多都不是重点。

Indeed, the part of the original piece that she chose to highlight in her announcement tweet read as follows: My conversations make me wonder if the technical content of data center deals is mostly beside the point.

闭环冷却系统?

Closed-loop cooling systems?

我不信他们。

I don't believe them.

每年交几百万的税?

Paying millions in taxes?

我不信他们。

I don't believe them.

创造一千个工作岗位?

Creating 1,000 jobs?

我不信他们。

I don't believe them.

我听到的是,对每一个说法都会本能地产生怀疑。

I hear a reflexive skepticism of every claim.

更准确地说,AI 公司和社区打交道的方式——强迫签 NDA,拿几十亿美元的承诺当诱饵,把环境代价尽量推到远离 AI 用户群体的地方——很像一个关于政治里 dark money 的老故事。

Rather, the way that AI companies engage with the communities—forcing NDAs, dangling billion-dollar promises, pushing environmental externalities far away from AI's user base—resembles a classic story about dark money in politics.

你们的政客正在被亿万富翁和超大型企业收买,目的就是坑你们。

Your politicians are being bought by billionaires and gigantic corporations to screw you over.

这不是 Skynet,这是寡头统治。

It's not Skynet, it's the oligarchy.

作为结论的一部分,Jasmine 写道,在更深层的文化问题得到解决之前,围绕 data center 的反弹只会以别的形式不断重演。

As part of her conclusion, Jasmine writes, until the deeper cultural problems are addressed, the data center backlash will only recreate itself in other forms.

而之所以值得努力去把这场讨论变得更好,并不是因为我们有多在乎那些 hyperscalers 的利润。

And the reason that it's worth trying to have a better version of this conversation is not because we care so much about the hyperscalers' bottom lines.

Reason.org 最近发表了一篇文章,讲为什么地方上对 data center 的 moratoria 会让整个国家都付出代价。

Reason.org recently published a piece on why local data center moratoria are costly for the whole country.

作者写道,在当前这种环境下,moratoria 构成了一个特别棘手的挑战。

The author writes, moratoria pose a particular challenge in the current environment.

这些临时禁令不是让官员去仔细讨论某个拟议项目会如何影响某个具体社区里的能源和供水基础设施,反而逼着他们在一种很抽象的层面上看待 data center,而在这种抽象层面上,那些反对 AI 的人就更容易歪曲事实。

Instead of deliberating over the specifics of how a proposed project will interact with energy and water infrastructure in a particular community, these temporary bans force officials to consider data centers in the abstract, where those opposed to AI can more easily distort the facts.

在这种环境下,良好的治理很难真正建立起来。

This is a difficult environment for good governance to take hold.

考虑到全国范围内持续存在的反弹情绪,地方领导人面临着去支持这些 moratoria 的政治激励,这一点是无法否认的。

The political incentives that local leaders face to support moratoria, given the ongoing national backlash, are undeniable.

他们接着写道,data center 最重要的好处,并不只是它们给当地社区带来的就业和税收,虽然这些也很重要。

They continue, the most important benefits of data centers are not the jobs and tax revenue they bring to local communities, significant though those may be.

它们几乎是我们每个人现在每天进行的每一种计算任务所必需的基础设施。

They are essential infrastructure for virtually every computing task each of us now performs on a daily basis.

在没有任何本地正当理由的情况下,提前把几百个社区从潜在开发名单上划掉,会让 AI 的开发和使用——更不用说其他类型的计算了——变得更加昂贵。

Preemptively crossing hundreds of communities off the list for potential development, often without any local justification, will make the development and use of AI, not to mention other types of computing, more expensive.

我同意,虽然和那位作者不一样,我认为这对社区本身的潜在好处也同样巨大。

I agree, although unlike that author, I think that the potential benefit to communities as well is immense.

我认为,不管是从国家层面、整体层面,还是从地方层面,我们都有充分理由尽一切努力,把这场讨论变得更好。

I think you have both national and aggregate reasons as well as local reasons to try with all our might to have a better version of this conversation.

但这具体到底意味着什么呢?

But what does that actually mean?

先说清楚,我基本上认为,这件事应该由 hyperscalers、labs 以及 data center 的建设方来牵头。

Well, to be clear, I pretty much think it's the job of the hyperscalers and labs and data center constructors to take the lead here.

换句话说,我不认为我们可以只是对人们大喊大叫,说他们在搞关于 data center 的阴谋论,或者直接把他们的担忧 dismiss 掉。

In other words, I do not believe that we can just yell at people for having conspiracy theories about data centers or dismiss their concerns.

我觉得,不管你喜不喜欢,如果你想建设 data center,那么争取社区站到你这边,这项工作现在已经是 mission critical 了。

I think that, like it or not, if you are trying to build data centers, the work of getting communities on your side is now mission critical.

所以,我们该怎么从我刚才说的那种糟糕透顶的做法,走向做得更好呢?

So how can we move from the absolutely abysmal job, as I said, that they're doing to doing it better?

第一部分是一种态度,就是要对人们所处的位置抱有同理心。

The first part is a disposition, and it's to have empathy for where people are.

这并不意味着要纵容阴谋论,或者接受那种“这不过是寡头攫取权力”的说法。

This does not mean indulging conspiracy theories or accepting the notion that this is just some oligarchic power grab.

但是,那些试图争取社区支持的人,必须理解,这些社区之所以怀疑,是有正当理由的,而且这种怀疑是有真实来源的。

But the people who are trying to get communities on their side need to understand that the reasons that those communities are skeptical are legitimate and come from a real place.

这种怀疑根植于他们看到过的事情、经历过的事情,以及他们现在仍然在经历的事情。

They are rooted in what they have seen and what they have experienced and what they continue to experience.

有些担忧,即便在某些情况下更多是情绪上的、而不是理性分析上的,也不能就这么轻轻一挥手给撇到一边。

There are concerns that, even if in some cases more emotional than intellectual, cannot be simply swept to the side.

当一种公共讨论的核心问题在于,一方的人觉得自己没有主动权,也觉得自己的声音没有被听见,那你就必须先从让他们觉得自己被听见开始。

When a key problem of a discourse is that people on one side do not feel like they have agency and do not feel like they are being heard, you have to start with making sure that they feel heard.

没有这一点,你连进入实质性的争论都做不到。

Without it, you can't even get into the substance of the argument.

接下来,我再说一遍我在这个节目里老说的一句话:我们必须把“为什么总体上我们需要更多 data centers”这个道理讲得更好。

Next up, to say again something that I say all the time on this show, we have to be making a better argument for the overall why and the reason that in general we need more data centers.

这个论点不能——我是说绝对不能——变成“反正这是不可避免的,不管怎样我们都会拥有这项技术”。

The argument cannot, and I mean cannot be, well, it's simply inevitable and we're going to have this technology one way or another.

也不能是“如果我们不做,China 就会做”。

And it also can't be, well, if we don't do it, China will.

对于“为什么 AI,以及顺带说 data centers,可以存在”这个问题,唯一说得过去的答案,必须是解释清楚:更多 data centers 到底怎么改善人们当下的现实生活。

The only acceptable answer to why AI, and by extension data centers, get to exist, has to be an explanation of how having more data centers improves people's lives as they live them.

如果我们不能把这件事大声而且有说服力地讲明白,那我们后面讨论的其他任何东西都几乎毫无意义。

If we do not make this case loudly and convincingly, nothing else that we talk about matters in the slightest.

再往下,这个讨论就必须从总体层面转到具体层面。

From there, the conversation has to move from the general to the specific.

就像 Reason 指出的那样,围绕 moratorium 的讨论,会一直把话题停留在“data centers 整体”这个泛泛的层面。

As pointed out by Reason, the moratorium discourse keeps the conversation about data centers in general.

虽然没错,我刚才也主张过,建设 data centers 的公司需要讲出一个积极的、总体性的故事,说明为什么 data centers 和 AI 应该存在;但他们并不想在每一个地方社区里都去说服大家“所有 data centers 都是好的”。

And while yes, I just argued that companies who are building data centers need to tell the positive general story about why data centers and AI should exist, they do not want to have to convince a local community that all data centers are good.

相反,他们需要能够和当地社区讨论的是,怎么把眼前这个具体项目做好。

Instead, they need to be able to engage with the local community on how to make that specific project that they're dealing with good.

每一个项目都必须凭它自身真正的优点去赢得本地支持,但同时也不该被其他项目可能带来的缺点拖累。

Every project needs to earn local support on its actual merits, but also not be weighted by the potential demerits of other projects.

哦,顺便说一句,这个对话还必须是透明进行的。

Oh, and by the way, that conversation has to happen transparently.

关起门来私下谈交易的时代已经结束了。

The days of backroom deals are over.

信任不可能以任何方式、在任何场景里,通过闭门协商建立起来。

Trust cannot in any way, in any context, be negotiated behind closed doors.

流程、权衡、承诺、责任、顾虑、反对意见、还价——所有这些都必须让人看得见。

The process, the trade-offs, the promises, the commitments, the concerns, the pushback, the counteroffers, all of that has to be visible.

就算这样会让流程变得更低效,也没关系。

It does not matter that, that makes the process more inefficient.

事实上,data center 的建设方没有意识到的一点是,他们现在主要参与的,并不只是一个纯粹的商业流程。

In fact, what data center builders fail to realize is that they are primarily not involved exclusively in a business process.

这现在已经是一个民主过程了,而低效率不只是民主过程的副产品,它在结构上就是这个过程不可分割的一部分。

This is now a democratic process, and inefficiency is not simply a byproduct of democratic process, but is structurally integral to it.

摩擦、缓慢、来来回回,这些恰恰是人们真的被听见时会出现的样子。

Friction, slowness, back and forth are what happens when people are actually being heard.

现在,也许我有点过于乐观了,而且坦白说,你在 Jasmine Sun 那篇文章里其实也能看到这一点——我并不一定认为这些 data center 建设方以前都是偷偷摸摸地,想在烟雾缭绕、雪茄满屋的密室里搞黑箱交易。

Now, perhaps overly optimistically, and frankly, you even see this in Jasmine Sun's piece as well, I don't necessarily think that all these data center builders were sneakily trying to cut backroom deals in smoky cigar-filled rooms.

我觉得更像是,长期以来,默认的做法本来就不是把所有人都邀请进这个谈判过程。

I think that it's just for a long time, the norms weren't to invite everyone into the dealmaking process.

他们以前的做法是跟那些有决策权的官员谈,然后让这些官员自己去应对他们各自的本地选民和社群。

They were to talk with the officials in charge of the decision and let them handle their local constituencies.

所以我希望,大家如果认识到现在这个流程从根本上已经和过去不一样了,就会让建设方和 hyperscalers 也接受一种完全不同的处理方式——一种从根本上透明的方式。

And so hopefully a recognition that this process is now fundamentally different than that will allow the builders and hyperscalers to come on board with a very different way of approaching it, one that is fundamentally transparent.

再说到具体的激励机制,以及怎么让 data centers 真正为社区发挥作用,我不确定现在是不是已经有一套完美模板,能同时回应居民的担忧,又让 data centers 最大程度地给这些居民带来机会。

Now, when it comes to the specific incentives and how to make data centers work for communities, I don't know that I think that there is a perfect template yet for how to address the concerns of citizens and how to make data centers most opportune for those citizens.

但我很确定的是,过去公司觉得自己把业务带到一个地方,就应该拿到税收优惠;而现在这个思路已经彻底一百八十度转变了——公司必须开始用一种新的心态来看待这件事,那就是要把成本里一个有分量的部分,用在争取小镇和社区的支持上。

I do know that whereas companies used to think that they were going to get tax breaks for bringing their business to a place, that has shifted entirely 180 to now companies needing to be in the mindset of allocating a meaningful part of their costs to getting town buy-in.

公司会、而且也必须,真金白银地去解决本地的担忧,让 data centers 真正对人们有用。

Companies will and need to spend real money solving local concerns and making data centers work for people.

我觉得,这类工作的一个最低门槛,可能会包括像这样的一些事:确保 data centers 在电力这类成本上涨的问题上,自己承担自己的那一份。

I think some of the floor of this type of work is going to be things like making sure that data centers pay for their own way when it comes to the cost increases for things like electricity.

但我也觉得,最后大家会发现,这样做还是不够。

But I also think that that's going to be found to be insufficient.

我认为 data centers 必须非常慷慨,不能只是覆盖自己的成本;实际上,它们还得替很多其他方面买单,甚至还可能需要用别的方式投资社区。

I think data centers are going to have to be extremely generous and do more than cover their own costs, but in fact foot the bill for many others and potentially invest in the community in other ways as well.

把这一点再强调一下,也算是给前面这些内容做个总结:现在,思考一个 data center 会怎样让本地社区受益,已经成了项目的核心基础设施。

Reinforcing this and putting a bottom line on all of it, thinking about how a data center benefits the local community is now core infrastructure to the project.

我说的既包括意图层面,也包括实际预算层面。

I mean this both in terms of intention and actual budget.

需要大幅提高分配给解决这些担忧、提供本地激励的预算比例,没错,就是这样,句号,完。

The percentage of the budget that is going to be allocated for addressing concerns and providing local incentives needs to increase dramatically, period, full stop.

说实话,如果你坐在那里跟我争,说这就是这些项目会把预算花爆的地方,那你就别干这行了。

And frankly, if you're sitting there arguing to me that that's the thing that's gonna break the bank for these projects, you need to get out of show business.

真到关键时刻,我对数据中心反弹的悲观看法,要比很多人都少得多。

When push comes to shove, I am far less pessimistic than many about the data center backlash.

其中一部分原因是,我觉得,能够对影响我们生活、而我们又不认同或者有顾虑的事情,大声表达意见、提出反对,这本来就是美国的典型特征,也是我们作为一个民族的核心组成部分。

Part of that is that I think it is quintessentially American and integral to who we are as a people to be able to raise our voices and push back against things that affect our lives, that we don't agree with or have concerns about.

这种混乱,虽然让建设方很头疼,但不管你喜不喜欢,它就是民主在运作。

The messiness, while frustrating to the builders, is, like it or not, democracy in action.

但更重要的是,我们有太多机会,让这套基础设施对每个人都真正有用。

But more than that, there is so much opportunity to make this infrastructure work for everyone.

是的,有些社区,不管出于什么原因,就是不行。

Yes, there are some communities where it is simply not going to work for whatever reason.

可能是因为能源需求根本满足不了,也可能是因为人们根本不愿意接受噪音污染之类的负外部性。

Could be because the energy requirements just aren't there, or because people just aren't willing to deal with the negative externalities like noise pollution.

但我相信,在更多社区里,我们都能找到对所有人都有利的结果,让这种特定类型的基础设施,真正成为一个社区下一阶段发展里有价值的一部分。

But there will be in many more communities, I believe, the ability to find win-wins for everyone where this particular type of infrastructure can actually be a valued part of a community's next phase.

我觉得,到目前为止我们没做到这点,是理解上的失败,也是想象力上的失败,而我认为这两点都是可以解决的。

I think our failure to do this so far is a failure of understanding and a failure of imagination, and I think both of those can be addressed.

如果你们任何人正在跟那些建这些数据中心的 hyperscalers 或 labs 打交道,正想着下一步该怎么办,就把我这个论点发到我的邮箱里。

For any of you who are dealing with the hyperscalers or labs who are building these data centers and are wondering what to do next, send them this argument in my email.

我很乐意直接对着他们吼。

I will happily shout at them directly.

今天的 AI Daily Brief 就先到这里。

For now, that's going to do it for today's AI Daily Brief.

一如既往,感谢你的收听或观看,我们下次再见,保重。

Appreciate you listening or watching as always, and until next time, peace.
已剔除 3 处广告(点击展开查看)
M1
M10:26广告 · 已剔除

First of all, thank you to today's sponsors, KPMG, Blitzy, Robots & Pencils, and HyperAgent.

To get an ad-free version of the show, go to patreon.com/AIDailyBrief, or you can subscribe on Apple Podcasts.

And to learn more about sponsoring the show, send us a note at [email protected].

By the way, while you're on aidailybrief.ai, you can find each episode turned into a convenient set of cutdowns.

So if there is some particular quote or number or part of the episode that you are super interested in sharing, chances are good that it has a shareable card that might make it a little bit easier.

And every episode gets that sort of treatment, so check it out at aidailybrief.ai.

We also have our Summer Adventure going, which is a choose-your-own-adventure free self-directed education program.

It's got a full range of skills you can develop from very basic to more advanced.

You can check it out again at summeradventure.ai.

M1
M110:07广告 · 已剔除

One of the most important AI questions right now isn't who's using AI, it's who's using it well.

KPMG and the University of Texas at Austin just analyzed 1.4 million real workplace AI interactions and found something surprising: the highest impact users aren't better prompt engineers, they treat AI like a reasoning partner.

They frame problems, guide thinking, iterate, and push for better answers.

And the good news?

These behaviors are teachable at scale.

If you're trying to move from AI access to real capability, KPMG's research on sophisticated AI collaboration is worth your time.

Learn more at kpmg.com/us/sophisticated.

That's kpmg.com/us/sophisticated.

Here's why most legacy modernization projects fail.

The AI doing the work can't understand codebases at scale.

It sees a small slice of context, examines syntax, and misses years of decisions distributed across the global application ecosystem.

Blitzy solves this the way it solves everything.

Grounded in your code before any migration begins, Blitzy's agents reverse engineer the entire legacy system into a persistent knowledge graph.

Every dependency, every constraint, every piece of tribal knowledge that used to live in one engineer's head.

From that understanding, Blitzy autonomously executes language migrations, framework upgrades, and monolith-to-microservices transformations, all validated end-to-end.

One Blitzy customer modernized a $10 million monolithic insurance stack in 16 weeks against a 137-week baseline with coding agents.

That's 9x compression.

Retire technical debt while accelerating your roadmap.

See how at blitzy.com.

That's B-L-I-T-Z-Y dot com.

M1
M111:49广告 · 已剔除

Robots and Pencils is already launching and scaling

agentic and generative AI in production at large enterprises in weeks.

AWS Advanced Tier Partner more than doubled in a year.

And they're hiring!

50 open roles.

If you're someone who knows this moment is different, who wants to be inside it, not watching it, this is worth a look.

At Robots and Pencils, the best ideas win, and the team is purposefully kept super high quality.

This is the kind of place you look back on as the best decision you ever made.

Take a look at robotsandpencils.com/careers.

This episode of the AI Daily Brief is brought to you by HyperAgent, where you run fleets of agents your team can manage together.

New users get $1,000 in inference.

Forget local agents and chat workflows waiting on your laptop to be prompted.

HyperAgent deploys always-on agents in the cloud, doing real work across the tools your team already uses.

Marketing's agent turns competitor moves into landing pages.

Sales' agent enriches leads, drafts emails, and updates the CRM.

The ops agent chases the paperwork and tracks the budget.

Every agent has access to shared context and follows your rules about scope and approvals.

It's time you add agents that feel like teammates.

Hire yours at HyperAgent, built by the team at Airtable.

Claim your $1,000 in inference at hyperagent.com/AIDailyBrief.