你未曾听闻的AI数据中心故事 | Ayșe Coskun

在医生严重短缺的地区，你们如何提供医疗服务？

How do you provide health care in a place with extreme physician shortages?

2011年，南苏丹成为一个独立国家。

In 2011, South Sudan became an independent country.

我们是当今世界上最新的国家。

We are the youngest nation in the world right now.

我们感到被忽视了。

We feel neglected.

这位是约翰内斯·赖希医生，他是他所在社区第一位成为医生的人。

That's doctor Johannes Reich, who's the first in his community to become a doctor.

作为一名医生，我对未来的希望是建立一个健康的社会，这正是我们正在努力的目标。

My hope for the future first as a doctor is a healthy society, and that is what we are working for.

当现在的U.

When the current U.

S.

S.

政府上台后，我们收到了突然的通知。

Government came into power, we got an abrupt notification.

明天，我们将暂停所有活动。

Tomorrow, we suspend all activities.

这立即对我们造成了影响。

So that had immediate effect on us.

成千上万的社区无法获得基本的医疗服务。

Thousands of communities are not able to access basic health care services.

他通过自己的非营利性社区医疗组织YoCare，在极其困难的条件下取得了非凡的成就，帮助南苏丹人民获得必要的医疗服务，并掌握自身健康需求的主动权。

With his nonprofit community health care organization, YoCare, he's made heroic strides against great odds to empower the people of South Sudan to get access to integral medical care and take charge over their own health needs.

了解更多关于约翰内斯及其令人惊叹的故事，请收听我们最新一期的《TED演讲每日秀》中的“同伴星期五”访谈。

Learn more about Johannes and his breathtaking story in our most recent Fellow Fridays interview only on TED Talks Daily.

您正在收听《TED演讲每日秀》，我们每天为您带来新思想和对话，激发您的好奇心。

You're listening to TED Talks Daily where we bring you new ideas and conversations to spark your curiosity every day.

我是您的主持人伊莱斯·胡。

I'm your host, Elise Hu.

人工智能数据中心常被指责加剧了电网压力、大量消耗水资源并推高电力需求。

AI data centers are often blamed for straining power grids, heavy water usage, and driving up electricity demand.

这可能是我们在人工智能领域面临的最大问题之一，但计算机科学家艾莎·乔什昆表示，我们不应与人工智能数据中心争夺宝贵的自然资源，而应提出一个不同的问题。

This may be one of the biggest issues we face when it comes to AI, but computer scientist Aisha Joshkun says instead of competing with AI data centers over invaluable natural resources, we should be asking a different question.

我们如何利用人工智能来真正帮助稳定电网？

How can we leverage AI to actually help stabilize the grid?

在这场演讲中，艾莎分享了如何将被视为能源浪费者的设施转变为灵活的电网支持资产，这一举措有望防止停电、降低成本，甚至加速清洁能源的普及。

In this talk, Ayesha shares how to turn the facilities seen as energy hogs into flexible grid supporting assets, a move that could prevent blackouts, lower costs, and even accelerate the adoption of clean energy.

她阐述了，只要我们敢于尝试，人工智能的可持续未来是存在的。

She shares why a sustainable future for AI exists if we're bold enough to try.

目前，全球正处于一场人工智能竞赛中。

Right now, the world is in an AI race.

企业、政府和大学都在竞相构建更大、更智能的模型，而在幕后，它们也在竞相建设更多数据中心来支撑人工智能。

Companies, governments, universities are all racing to build bigger models, smarter systems, and behind the scenes, they are racing to build more data centers to power AI.

但这里有一个问题。

But there's a problem.

我们正一头撞上基础设施的极限。

We are running head first into the limits of our infrastructures.

电力网络包括所有基础设施、发电厂、输电线路等，用于发电并输送电力到我们的家庭、企业，如今还包括AI数据中心。

The power grid includes all the infrastructures, power plants, transmission lines and all to generate and deliver power to our homes, our businesses, and now to AI data centers.

在美国，电网运营商报告称，新的AI数据中心项目所要求的电力负荷相当于整个城市的需求。

In The United States, the grid operators are reporting that new AI data center projects are requesting power loads equal to entire cities.

在某些地区，公用事业公司根本无法跟上需求。

In some regions, utilities simply can't keep up.

所以当你听到AI数据中心时，你会想到什么？

So when you hear AI data centers, what comes to mind?

对许多人来说，那就是巨大的电能消耗者。

For many, it's one thing energy hogs.

他们并没有错。

And they are not wrong.

AI正在急剧推高数据中心的电力需求。

AI is dramatically accelerating the electricity demand of data centers.

仅训练GPT-4就消耗了大约数千户美国家庭的年用电量。

Just training GPT-four is estimated to have consumed around the annual electricity use of thousands of US homes.

另一个令人震惊的例子是，在爱尔兰，如今数据中心消耗了全国近20%的电力。

In another striking example, in Ireland, nearly 20% of the nation's electricity is drawn by data centers today.

这些不仅仅是统计数据。

And these are not just statistics.

它们也是社区的真实故事。

They are also community stories.

在弗吉尼亚州的数据中心走廊，居民们最近发现电费上涨了。

In the data center alley in Virginia, residents recently saw higher electricity bills.

相比几年前，电费已经上涨了20%，因为公用事业公司正竭力为大规模的新AI设施供电。

20% higher already compared to just a few years ago, as utilities scramble to serve massive new AI facilities.

因此，‘电老虎’这个标签似乎实至名归。

So, energy hog label seems well deserved.

但这只是故事的一半。

But that's only half the story.

这是新的视角。

Here is the new view.

这些设施不仅仅是耗能的智能大脑，它们还可以作为电网的肌肉，按需灵活调节。

These facilities are not just energy hungry brains, they can also be the muscles of the grid flexing on demand.

与我们的家庭或医院不同，AI数据中心运行的任务是可预测、可控制且通常可以延迟的。

Unlike our homes or hospitals, AI data centers run jobs that are predictable, controllable, and often delayable.

这使得它们成为帮助平衡电网供需的理想选择。

That makes them ideal to help balance supply and demand on the grid.

通过使AI数据中心的电力使用更具弹性，我们可以更快地将其接入电网，同时让电力更便宜、更具韧性。

By making AI data centers power flexible, we can connect them much more rapidly to the grid, while at the same time making electricity more affordable and resilient.

更重要的是，AI热潮恰逢可再生能源热潮也在兴起。

What's more, the AI boom is arriving just as the renewable boom is also taking off.

风能和太阳能不遵循我们的作息，但数据中心可以。

Wind and solar don't follow our schedules, but data centers can.

这意味着，如果我们敢于重新思考它们的角色，就能让AI的发展与清洁能源的崛起相协调。

Which means we can align the rise of AI with the rise of clean energy if we are bold enough to rethink their role.

所有这些向电力灵活性的转变并非凭空而来。

All this transformation to power flexibility didn't just come out of thin air.

它建立在数十年关于节能计算、调度、优化及其他相关领域的研究基础之上。

It builds on decades of research on energy efficient computing, scheduling, optimization, and many others.

我自己亲身经历了这一历程。

I've lived this journey myself.

在我职业生涯早期，我提出了一个许多人认为不切实际的问题。

Early in my career, I asked a question that many found unrealistic.

计算机系统能否根据电力网络的需求调整自身行为，同时不违背对用户承诺的性能？

Could computer systems adapt their behavior depending on power grid needs, but without breaking their performance promise to their users.

当时，这听起来很激进，因为谁会故意设计一个会主动降低速度的系统呢？

At the time, this sounded radical because why would we ever design a system that would slow itself down on purpose?

但随后，突破出现了。

But then came the breakthroughs.

首先，我们发现并非所有计算任务都是紧急的。

First, we discovered not all computing tasks are urgent.

有些任务可以等待几分钟甚至几小时，有些任务即使慢一点，用户也根本察觉不到。

Some can wait for minutes or hours and some can be slowed down without anyone really noticing it.

例如，一位研究人员用人工智能分析数百张医学影像时，可能愿意多等一会儿。

For example, a researcher analyzing hundreds of medical images with AI may be okay with waiting just a little longer.

或者，如果你在未来几天内微调你的AI模型，你可能愿意让它暂时慢几个小时。

Or if you are fine tuning your AI model over the course of the next few days, you may be okay with slowing it down for just a few hours.

计算本身具有的这种灵活性，为我们提供了管理电力所需的弹性。

This inherent flexibility in computing gives us the flexibility we need to manage power.

第二，我们重新定义了这个问题。

Second, we reframe the problem.

我们不再问：如何尽可能快地完成计算？

Instead of asking, how do we compute as fast as possible?

而是问：如何让计算机系统在满足电网约束的同时，依然兑现对用户的性能承诺？

We asked, how do we make computer systems meet the constraints of the power grid while at the same time still delivering on user performance agreements.

这种转变催生了新的策略，比如限制功耗、转移工作负载，以及将数据中心作为电网的灵活备用资源。

This shift led to new strategies, capping power, shifting workloads, and provisioning the data center as a flexible reserve to the grid.

这里的一个关键点是我们始终向用户保证性能。

A key aspect here is that we do keep the performance promise to users.

所以这并不是随意的。

So it's not arbitrary.

用户体验仍然是一个核心目标，而且更进一步地说，它变得更加可预测了。

User experience remains as a key target and better yet, it becomes more predictable.

因此，我们在真实的数据中心服务器上构建了原型，结果成功了。

So, we built prototypes on real data center servers, and they worked.

这些系统能够在遵循电力目标的同时，依然提供结果。

Systems that could follow a power target while still delivering results.

但这一整个过程并不顺利。

But all this journey wasn't smooth.

曾遭遇论文被拒、经费被拒，同事也告诉我这根本行不通。

There were paper rejections, funding rejections, colleagues telling me this would never work.

从小到大，我一直被说是个执着的人，有时甚至有点固执。

Well, since I was a kid, I was told I'm a persistent person, perhaps stubborn at times.

而大胆的构想需要坚持，因为变革在变得显而易见之前，几乎总是看起来不可能。

And bold ideas require persistence Because change almost always looks impossible before it looks obvious.

所以你要接受这些反馈，不断重新思考，持续构建，不断证明。

So you take that feedback, you reframe it again and again and you keep building, You keep proving.

因此，十二年前在白板上潦草画下的想法，如今已在真实的AI数据中心运行。

So what began as scribbles on a whiteboard twelve years ago is now running on real AI data centers.

为什么现在这很重要？

Why does this matter now?

因为电力网络的挑战不仅仅是产生更多的电力。

Because the power grids challenge isn't just to generate more power.

而是时机的问题。

It's about timing.

太阳能在正午时提供大量电力，但需求可能在傍晚达到高峰。

Solar gives us a glut of electricity at noon, but demand might peak in the evening.

风能可能某一天充沛，而第二天却稀缺。

Wind might be abundant one day and scarce the next.

核能建设需要数十年和数十亿美元，而且通常难以在城市地区选址。

Nuclear takes decades and billions of dollars to build and is often hard to locate in urban areas.

电池至关重要，但扩大其规模成本高昂、进展缓慢，且往往不够环保。

Batteries are critical, but scaling them is costly, slow and often not environmentally clean.

与此同时，AI数据中心在弗吉尼亚等地连接电网的等待时间长达五到七年。

Meanwhile, AI data centers themselves face five to seven year wait times just to connect to the grid in places like Virginia.

在AI领域，技术每六个月就会发生重大变革，五到七年简直是永恒。

In AI time, where technology is shifting a major way every six months, five to seven years is an eternity.

所以，这里有一个机遇。

So here's the opportunity.

通过恰当的协调，AI数据中心如今就可以具备灵活性。

With the right orchestration, AI data centers can be flexible today.

无需等待。

No waiting.

无需新建大规模电力基础设施。

No new massive power infrastructure construction.

它们可以在下午吸收多余的太阳能，在用电高峰时降低负载，如今就能充当虚拟电池。

They can soak up excess solar in the afternoon, scale down at peak times, and act as virtual batteries today.

而且这关系重大。

And the stakes are real.

以得克萨斯州8月23日为例。

Take Texas, August 23.

在一场酷热天气中，不断上升的电力需求将电网推至极限。

During a brutal heat wave, the rising electricity demand pushed the grid to its limits.

单日下午，批发电价飙升超过800%。

Wholesale electricity prices spiked over 800% in a single afternoon.

因此，如果灵活负载广泛可用，本可以降低电价，并避免向消费者发出紧急警报。

So, flexible loads, if they were widely available, could have reduced the costs and could have prevented the emergency alerts that went to the consumers.

所以我们这里有两个机会。

So we have two opportunities here.

第一，我们可以让现有的数据中心具备灵活性，帮助预防停电并降低电力成本。

One, we can make current data centers flexible and help prevent blackouts and reduce electricity costs.

第二，也是或许更重要的，通过让未来的数据中心具备电力灵活性，我们可以在不等待重大电网升级的情况下更早地接入它们。

Two, and perhaps the more significant, by making future data centers power flexible, we can connect them much earlier without waiting for major power grid upgrades.

如果我们忽视这一机会，我们不仅是在浪费可再生能源，也不仅是在提高电费。

If we ignore this opportunity, we are not just wasting renewable energy and we are not just raising our electricity bills.

我们还在拖慢人工智能的普及，使其延迟、成本更高，并且对社会更难获取。

We are also slowing AI adoption, making it delayed, more expensive, and less accessible to society.

但这里有个问题。

But there's a catch.

要协调这种灵活性并不容易。

Orchestrating this flexibility is not easy.

电价每小时都在变化。

Prices change hourly.

工作负载可能 unpredictably 到来。

Workloads may arrive unpredictably.

电网规则在各州、各国之间各不相同。

Grid rules change across states, across countries.

因此，没有任何人工操作员或单一固定的数据中心管理策略能够跟上这种节奏。

So no human operators and no single fixed data center management policy can keep up.

这时，AI本身又回到了故事中。

This is where AI itself comes back into the story.

推动这种意想不到需求的技术，很可能也是唯一足够智能以驾驭它的工具。

The very technology driving this unforeseen demand is also probably the only thing smart enough to tame it.

AI可以学习模式，预测电网需求，并实时协调跨数据中心、跨公用事业公司甚至跨国界的操作。

AI can learn patterns, anticipate grid needs and coordinate across data centers, across utilities, even nations in real time.

想象一个数据中心，或整个数据中心网络，就像一支拥有数百种乐器同时演奏的交响乐团。

Imagine a data center or a whole network of them as an orchestra with hundreds of instruments all playing at once.

如果任其自行发展，听起来会像一片混乱。

Left on their own, it can sound like chaos.

但一旦有了指挥家，所有的噪音瞬间化为音乐。

But bringing a conductor, suddenly all that noise turns into music.

在这种情况下，指挥家就是AI。

The conductor in this case is AI.

AI可以指挥数据中心的运行，使数据中心能够根据电网需求、可用电力和用户需求精确匹配电力限制。

AI can direct data center operation so that the data centers can precisely match power constraints depending on what the grid needs, what power is available, and what users demand.

结果就是和谐。

The result is harmony.

可靠的电力、高效的计算，以及一个完美协同的系统。

Reliable electricity, efficient computing, and a system that works beautifully together.

而这正是我们所构建的。

And that's exactly what we've built.

我们开发了软件，能够在数据中心内减缓、加速或暂停工作负载，或将工作负载在不同数据中心之间转移。

We built software that slows down, speeds up, or pauses workloads in a data center or shifts workload among data centers.

我们的Conductor平台实时调节性能与功耗，同时始终满足用户和云服务商的性能需求。

Our Conductor platform tunes performance and power at real time, all the while respecting user and cloud provider performance needs.

通过在需要时灵活调整，我们能够更快地将AI数据中心接入电网，更高效地利用电网中的可用电力，并推动AI的更快普及。

In this way, by flexing when needed, we can connect AI data centers much faster to the grid, make better use of the available power in the power grid, and enable faster AI adoption.

我从一个曾经看似不可能的想法开始，经历了实验室中的原型，直到如今在实地运行的系统，全程参与了这一历程。

I've been inside the story from an idea that once seemed impossible to prototypes in a lab to systems now running in the field.

我相信这仅仅是个开始。

And I believe this is just the beginning.

人工智能已经在重塑我们的计算方式，但它也可能重塑我们为世界供电的方式。

AI is already reshaping how we compute, but it could also reshape how we power the world.

所以问题不在于人工智能消耗多少能源。

So the question isn't how much energy AI consumes.

真正的问题是，人工智能能释放出多少灵活性、韧性与清洁能源？

The real question is how much flexibility, resilience, and clean power can AI unlock?

如果我们敢于重新思考人工智能数据中心，那些如今看似负担的机器，或许将成为构建可持续人工智能未来最宝贵的资产。

If we are bold enough to rethink AI data centers, the very machines that now seem like a burden could be our greatest assets in building a sustainable AI future.

谢谢。

Thanks.

这是艾莎·乔斯库恩在2025年于加利福尼亚州旧金山举行的TED AI大会上的演讲。

That was Aisha Joshkoon speaking at TED AI in San Francisco in California in 2025.

如果你想了解TED的选题标准，敬请访问ted.com/curationguidelines。

If you're curious about TED's curation, find out more at ted.com/curationguidelines.

今天就到这里了。

And that's it for today.

TED每日演讲是TED音频合集的一部分。

TED Talks Daily is part of the TED audio collective.

本演讲由TED研究团队进行事实核查，并由我们的团队——玛莎·埃斯特瓦诺斯、奥利弗·弗里德曼、布莱恩·格林、露西·利特尔和坦西卡·苏恩马诺翁——制作和编辑。

This talk was fact checked by the TED research team and produced and edited by our team, Martha Estevanos, Oliver Friedman, Brian Greene, Lucy Little, and Tansika Sungmarnivong.

本集由克里斯托弗·法齐·博甘混音。

This episode was mixed by Christopher Faizy Bogan.

特别感谢艾玛·陶布纳和达尼埃拉·巴雷罗的支持。

Additional support from Emma Taubner and Daniella Balarezo.

我是伊莉丝·胡。

I'm Elise Hu.

明天我会带着一个全新的想法回到你的信息流中。

I'll be back tomorrow with a fresh idea for your feed.

感谢收听。

Thanks for listening.

乔纳森·海特的著作《焦虑的一代》引发了一场运动，旨在警告青少年及其父母注意社交媒体的危害。

Jonathan Haidt's book, The Anxious Generation, sparked a movement to warn kids and their parents about the harms of social media.

是的。

Yes.

我的观点是，这将改变大脑发育，使你在成年后变得能力更弱、更不自信、更不快乐、更不善于社交。

My claim is that will change brain development in ways that will make you less capable, confident, happy, and sociable as an adult.

但年轻人是怎么想的呢？

But what do young people think?

Z世代只会觉得，唉，我们被诅咒了。

Gen z is just gonna think, well, we're cursed.

下一期《TED电台秀》播客，来自NPR，我们将继续探讨这个话题。

That's next time on the TED Radio Hour podcast from NPR.

请在您收听播客的平台订阅或收听《TED电台秀》。

Subscribe or listen to the TED Radio Hour wherever you get your podcasts.