资本市场的一次考验：为人工智能融资

欢迎收听《市场观点》。

Welcome to Thoughts on the Market.

我是摩根士丹利TMT信贷研究分析师林赛·泰勒。

I'm Lindsay Tyler, TMT credit research analyst at Morgan Stanley.

我是摩根士丹利全球债务资本市场主管阿尼什·沙阿。

And I'm Anish Shah, global head of debt capital markets at Morgan Stanley.

今天，我们来聊聊发行人和投资者如何应对人工智能融资这个快速变化的领域。

Today, how issuers and investors are approaching the rapidly evolving world of AI Financing.

现在是纽约时间7月16日，星期四上午10点。

It's Thursday, July 16 at 10AM in New York.

随着人工智能需求快速增长，信贷市场正被要求为基础设施提供融资，规模之大以往只出现在公用事业、电信或能源行业。

As AI demand accelerates, credit markets are being asked to finance infrastructure on a scale that used to be associated with utilities, telecom, or energy.

这对发行人和投资者来说，提出了一个核心问题：人工智能生态系统究竟能承受多少债务？成本又是多少？

That raises a central question for issuers and investors: how much debt can the AI ecosystem absorb and at what price?

阿尼什，你能为我们的听众梳理一下你负责的主要产品吗？

Anish, can you walk our listeners through the key products in your purview?

在我近二十年的摩根士丹利生涯里，这可能是我见过的信贷市场最令人惊叹的时期。

Certainly in my nearly twenty years at Morgan Stanley, this is probably the most incredible time period I've ever seen in the credit markets.

我有幸在资本市场和贷款领域的多个岗位上工作过。

I've had the privilege of working across a number of different roles in capital markets and lending.

几年前，我们把投资级和高收益债券的承销业务整合在一起，因为大家越来越清楚，整个信贷生态已经高度互联。

And a couple of years ago, we integrated the debt underwriting business across both investment grade and leveraged finance franchises in recognition of how interconnected the whole credit ecosystem has become.

除了帮助客户为战略重点筹措资金这一核心工作，我们还特别关注两个方向：一是挖掘私人信贷市场的潜力，二是为AI支出的惊人增长提供顶尖的融资解决方案。

In addition to our core activities helping clients raise capital for their strategic priorities, two of the big focus areas that we've had have been finding ways to harness the power of the private credit universe, and also delivering best in class capabilities in funding this incredible growth in AI spend.

AI融资确实也是我们研究团队一直关注的重点。

AI financing has certainly been a theme we've also been focused on in research.

我们的股票研究团队预测，几家关键公司将在两年内新增超过30吉瓦的算力，带动约2万亿美元的资本开支。

Our equity research colleagues project that a handful of key players could add more than 30 gigawatts of capacity over a two year time frame driving around $2,000,000,000,000 of aggregate cash CapEx in that period.

作为参考，一个吉瓦的数据中心基础设施，仅建筑外壳就要大约120亿美元，而芯片和机架的投入往往还要再翻一倍以上。

And to put that into context, a single gigawatt of data center capacity can require roughly $12,000,000,000 for the shell and often more than double that for chips and racks.

从你的角度看，我们现在处于哪个阶段？是什么让你相信信贷市场还能持续为这种规模的机遇提供资金？

So from your vantage point, what inning are we in and what gives you confidence that credit markets can continue funding this opportunity at scale?

这些数字确实惊人，正如你提到的。

I mean, the numbers certainly are staggering, as you note.

如果你看看超大规模云服务商和整个AI基础设施的资本支出预测，我们现在还处在非常早期的阶段。

And if you just observe the CapEx estimates for the hyperscalers and broadly for AI infrastructure, we're certainly in the early innings.

像你了解的那样，大型科技公司过去很少举债。

The largest tech companies have historically, as you know, raised very little debt.

事实上，很多公司连信贷额度都不需要。

In fact, many of these companies have not even needed a credit facility.

去年下半年，随着资本支出预测大幅上调，我们开始看到大规模融资的启动。

As CapEx projections were materially increased in the second half of last year, we saw the beginning of scaled capital raises.

超大规模云服务商的发债规模，已经从占投资级市场的不到1%，迅速攀升到超过10%。

Hyperscalers issuance has quickly gone from less than 1% of the investment grade market to more than 10% of the market.

展望未来，根据我们实地看到的情况，我们认为，无论是数据中心建设还是算力融资，与AI相关的融资总额可能占到所有信贷产品总发行量的15%以上。

You know, as I look ahead, based on what we're seeing on the ground, we think that AI related funding, whether it's for data center development or financing compute capacity, could top 15% of the total issuance across all credit products.

这对资本市场来说是一次前所未有的考验，无论是在资金容量的深度，还是在产品种类的广度上。

This has been an unprecedented test for the capital markets, both in terms of the depth of capacity and the breadth of product.

这些团队一直走在深度投资者沟通和产品创新的最前沿。

The teams have been on the forefront of deep investor dialogue and product innovation.

覆盖了企业投资级债券、开创性的融资方式，以及高收益债和杠杆贷款市场，还有资产支持融资的新模式。

This spans corporate investment grade, first of their kind financings and high yield and leveraged loan markets, and new takes on asset backed financing.

这些领域在公开和私人市场上都出现了显著的发行量。

And each of these areas has seen material issuance both in public and private markets.

背景确实不错。

Great backdrop.

我们先来看看企业投资级债券，你以前领导投资级团队时对这个领域非常熟悉。

Let's dig first into investment grade corporate debt, an area you know well from your time previously leading the investment grade team.

你能帮我们梳理一下，这部分融资的规模和重要性如何？AI相关的债务在其中又增长到了什么程度？

Can you help frame the scale and the significance of this financing bucket and how AI related debt is scaling within it?

你知道，投资级债券市场，尤其是美元市场，是全球最深、流动性最强的资本池。

Well, you know, as you know, the investment grade bond market, specifically in dollars, is the deepest, most liquid pool of capital in the world.

过去几年发行量大幅增长，今年很可能突破两万亿美元。

Volumes have grown materially over the last few years, and are likely to eclipse $2,000,000,000,000 in issuance this year.

超大规模云服务商是全球信用资质最好的企业之一，它们能快速进出市场，几乎不需要提前营销，而且发行规模通常很大。

Hyperscalers are among the very best credits in the world, and they have the ability to come in and out of markets with relatively quick twitch, little to no pre marketing, and in fairly large size.

以前，200亿美元以上的债券发行在投资级市场里非常罕见，现在每个季度都会发生好几次。

You know, $20,000,000,000 plus deals used to be rare in the investment grade market, now happen multiple times a quarter.

这就是为什么AI驱动的融资主要集中在投资级市场的原因。

This is why we've seen the predominance of AI driven capital raising take place in the investment grade market.

总体来看，投资者对这种供应量吸收得相当好。

For the most part, investors have digested that supply very well.

虽然我们看到一些超大规模云服务商和其他科技发行人的信用利差略有扩大，但相比它们预期的投资回报率来说，这点影响几乎可以忽略。

While we've seen some modest widening credit spreads for hyperscalers and some of the other tech issuers, I'd say it's de minimis relative to their expected ROI.

林赛，我刚才讲了很多关于供给和发行的情况。

Lindsay, I've talked a lot about supply dynamics and issuance.

在评估投资级科技债券的公允价值时，你和投资者还会考虑哪些其他因素？

What other factors are you and investors considering when assessing fair value for investment grade rated technology bonds?

当然。

Sure.

稳妥的做法是综合考虑技术面、基本面和相对价值。

It's prudent to really weigh a mix of technicals, fundamentals, and relative value.

正如你提到的技术面，结合我跟债券和股票投资者的交流，我一直关注的是建设规模、市场承载能力、不同货币市场的消化情况、收益率曲线上的位置、股权融资的影响，以及AI融资是否会挤占TMT领域其他信用板块的空间。

You know, as you discussed on the technical side and related to my discussions with debt and equity investors, I've been focused on scale of build outs, market capacity, digestibility across currencies, positioning along the curve, implications of equity issuance, and whether AI financing could crowd out other areas of TMT credit.

再往基本面看，你提到了投资回报率。

But moving more to the fundamental side of things, you mentioned ROI.

对于那些在扩大计算能力的公司来说，有几个关键问题反复出现：如何实现盈利和回报。

And for the players that are scaling compute capacity, there are a handful of key monetization and return questions that keep coming up.

这些公司一旦设备上线，能多快实现满负荷运行？

How quickly can these companies bring new capacity online once it's live?

这种能力如何转化为持久的收入和现金流？

How does it translate into durable revenue and cash flow?

这些算力是用于自家产品、专有模型、更广泛的云服务，还是租给第三方？

Is that capacity supporting internal products, proprietary models, broader cloud offerings, or compute leased to third parties?

如果需求或回报发生变化，这些算力在不同用途之间能有多大的灵活性？

And then how fungible is the capacity across those use cases if demand or returns shift?

再从基本面来看，我们做了一些独特的研究，聚焦于长期承诺的增长。

Further on the fundamental side, we've done some differentiated work around growing long term commitments.

我们发现，顶级云服务商和几家半导体公司正通过租赁、担保和其他义务，为人工智能生态提供核心支撑。

We've seen that high quality hyperscalers and a few of the semis companies are anchoring the AI ecosystem through leases, guarantees, other obligations.

这些承诺的范围远不止普通的债券发行那么简单。

These commitments really extend beyond vanilla bond issuance.

所以我建议投资者别只盯着已发行的债务，更要深入理解这些承诺在会计和信用评级上的影响。

So I encourage investors to look beyond the funded debt and really understand the accounting and the ratings implications here of some of those commitments.

这正好引出了我想谈的下一个话题——项目融资债务。

And this ties nicely into the next topic that I wanted to raise which is project finance debt.

我注意到，很多投资级企业做出的承诺，实际上为另一层融资提供了支持。

I've noticed that, you know, a lot of the commitments that we're seeing from IG players support another layer of financing.

租赁承诺可以作为项目融资债务的基础，这个领域近年来发行规模庞大，创新也很多。

Lease commitments can underpin project finance debt, an area of sizable issuance and innovation.

自2025年秋季以来，公开的高收益市场已成为数据中心建设的新融资渠道，15笔交易累计融资额已超过300亿美元。

The public high yield market has emerged as a new funding source in this way for data center construction, with more than $30,000,000,000 priced across 15 deals since fall twenty twenty five.

安ish，你能给我们讲讲这些结构背后的创新吗？这些高收益债券和别的项目融资方式有什么不同？

Can you walk us through, Anish, the innovation behind these structures and how are these high yield deals different than other ways to kind of raise project finance debt?

是的，这确实非常有意思。

Yeah, it's incredibly interesting.

正如我提到的，大部分发行集中在投资级市场，但真正的创新主要来自非投资级市场。

I mean, the bulk of the issuance, as I noted, has come in the investment grade market, but I would say the bulk of the innovation has come in the sub investment grade market.

过去，在能源、电力或房地产这类资本密集型行业，项目贷款一直是最初融资最高效的方式。

You know, historically, for very capital intensive sectors like energy and power or real estate, the project loan market was the most efficient source of initial funding.

开发商通常会找银行安排结构复杂的建设贷款。

The developer would tap banks to underwrite a highly structured construction loan.

等项目投入运营后，就可以依靠稳定的现金流，把贷款置换为更机构化的融资，比如投资级债券、B级定期贷款或证券化市场。

Once the project is up and running, you could then refinance that loan with the predictable cash flows into a more institutional financing, like the investment grade bond market or the term loan B or securitization markets.

这种模式在很多行业依然很有效，但我们很早就意识到，银行提供的建设贷款无法满足人工智能投资周期所需的规模。

That product may still be very viable in many sectors, but we felt early on that bank provided construction loans would not meet the capacity needs of the AI investment cycle.

市场迫切需要一种绕过建设贷款的机构信用产品。

The market really needed an institutional credit product that bypassed the need for construction loans.

关键的创新在于发行了史上首笔高收益债券，用于资助一个新数据中心综合体的建设。

The key innovation came in the form of first of its kind high yield bonds that funded the development of a new data center complex.

考虑到建设周期相对较短，而且有全球顶尖信用主体提供的稳定承购承诺，我们认为这种融资结构在高收益市场会大受欢迎。

Given the relatively short construction period and the quote unquote off take supported by some of the highest quality credits in the world, we felt like this financing structure would be incredibly well received in the high yield market.

这种模式的优势在于，开发商能获得固定利率的长期资金，同时保留提前赎回债券、以更低成本再融资的灵活性。

The win here is that the developer accesses fixed rate, long term capital, and maintains flexibility to call the bonds and refinance at a lower cost.

从这些融资的市场反应来看，投资者的热情非常高涨。

Judging by how these financings have gone, there's a strong level of investor enthusiasm.

我觉得这还只是冰山一角，未来这类交易会越来越多，甚至可能扩展到杠杆融资市场的其他产品，毕竟投资者需求实在太旺盛了。

I think that they've only scratched the surface, and I would expect that we see much more of this, and potentially even expanded to other products in the leveraged finance markets given the tremendous level of investor demand.

是的。

Yeah.

确实，跟进这些交易让人很兴奋。

It's certainly been exciting to follow many of those deals.

除了公开市场，我们还看到私人信贷和资产支持融资领域正在兴起一波创新浪潮。

Beyond the public space, we're also seeing a wave of innovation in private credit and asset backed finance.

阿尼什，公司是怎么决定该在公开市场还是私募市场融资的呢？

Anish, how do companies decide whether capital is best raised in the public or the private markets?

你提到私募市场这个话题真好，这可能是近几年信贷市场最重要的变化。

Well, I'm glad you raised the whole avenue of private markets because it may be the most significant change in the credit markets over the last few years.

私募信贷的范围已经从直接参与杠杆收购，扩展到现在为大型投资级项目提供资金。

Broadening the scope of private credit from directly lending into leveraged buyouts to now financing large investment grade projects.

比如在GPU和TPU融资方面，我们通过资产和现金流作为担保来安排贷款，数据中心开发也是类似的例子。

They're great examples in the world of GPU and TPU financing, where we structure loans secured by the asset and the cash flows, or in data center development.

林赛，从你的角度看，当这些私募结构与公共信用挂钩时，投资者最关注什么？

Lindsay, from your perspective, what are investors focused on when these private structures intersect with public credits?

确实如此。

Sure.

很多这类资产支持的私募融资，促使投资者更仔细地审视相关上市公司，无论是作为发行人、租户、客户还是支持方。

Many of these asset backed private financings have prompted investors to look more closely at any of the public companies involved, whether as issuers, tenants, customers, or support providers.

这其实又回到了我之前提到的那个观点。

This ties back to the point I raised earlier.

风险到底在哪儿？最终谁来承担责任？

Where does the risk reside and who ultimately is on the hook?

这些融资方式还引发了更广泛的讨论，比如循环依赖、供应商融资，以及即使有摊销结构也依然存在的技术过时风险。

These financings have also sparked broader discussions around circularity, vendor financing, and technology obsolescence risk, even when amortizing structures are in place.

我觉得这些担忧确实值得认真考虑，它们也反映出AI融资趋势演变之快，以及信用评估工作量之大。

I do think those are fair concerns to weigh and they really speak to how quickly the AI financing trend is evolving and how much credit work there is to do.

所以，Anish，考虑到需求强劲、创新迅速，但同时也存在一些实实在在的信用问题，我们来做一个快速问答环节吧。

So, Anish, with that balance in mind, relatively strong demand, rapid innovation, but also some real credit questions, let's end with a quick lightning round.

Lindsay，来吧。

Lindsay, let's do it.

第一个问题：什么是最有可能动摇投资者对AI相关债务兴趣的最大风险？

First, what is the biggest risk that could test investor appetite for AI related debt?

我觉得投资者最关注的是建设延期问题。

I would say investors are acutely focused on construction delays.

别小看当前市场上各方所投入的尽职调查力度。

Don't underestimate the level of diligence being done by the breadth of capacity you're seeing in the markets.

投资者都在认真做功课，我们也花了很多时间通过结构化保障来缓解他们的顾虑。

Investors are doing their homework, and we're spending a lot of time trying to mitigate any of their concerns with structural protections.

明白了。

Got it.

第二，除了数据中心外壳和芯片，下一个潜在的AI融资机会是什么？

Second, beyond data center shells and chips, what is the next potential AI financing opportunity?

绝对是能源和电力领域。

It most certainly is energy and power.

我们会看到大量资金涌入公用事业行业。

We're going to see a ton of capital being raised in utilities.

这和超大规模云服务商的做法会有些不同。

It's going to be a little different than what the hyperscalers are doing.

由于他们的资产负债表特点，我们会看到更多次级资本参与。

Just given the nature of their balance sheets, you're going to see more junior capital.

公用事业公司已经掀起了一波次级附属债务的发行潮。

We've seen a wave of junior subordinated debt issuance out of the utilities.

我们项目融资和税务股权团队最近也特别忙，毕竟现在所有焦点都在能源基础设施上。

We're also seeing a lot of activity from our project finance and tax equity team, just given all things energy infrastructure.

很好。

Great.

第三，如果一年后我们坐在这里回看，你觉得那时候最值得关注的AI融资话题会是什么？

And third, if we're sitting here a year from now, what do you think could be the biggest AI financing story we're talking about?

我们确实没料到过去一年融资活动的规模会这么大。

Well, we certainly underestimated the level of financing activity that we saw in the past year.

我觉得一年后回头看，我们会发现那些AI实验室早就准备好独立融资了，不再依赖外部支持。

I think when we look back a year from now, we will probably see that the AI labs were much more ready to finance on their own on a standalone basis.

这会缓解市场的一些压力，但也会带来全新的结构性创新挑战。

That's going to alleviate some of the pressures in the market, but I think it's going to create a whole new set of considerations in structural innovation.

亲眼看着这个融资趋势一步步成形，确实令人惊叹，而下一章可能更值得期待。

Well, it's certainly been remarkable to watch this financing theme take shape in real time, and the next chapter sounds like it could be even more interesting to follow.

Anish，感谢你加入我们，分享这些独到的见解。

Anish, thanks for joining us and sharing your insights.

很高兴能参加，林赛。

Great to join, Lindsay.

谢谢。

Thanks.

也感谢您的收听。

And thank you for listening.

如果您喜欢我们对市场的分析，欢迎在您收听的平台给我们留下评价，并今天分享给朋友或同事。

If you enjoy thoughts on the market, please leave us a review wherever you listen and share the podcast with a friend or colleague today.

以上内容仅供参考，基于发布时可获得的信息。

The preceding content is informational only and based on information available when created.

这不是任何要约或招揽，也不构成税务或法律建议。

It is not an offer or solicitation, nor is it tax or legal advice.

内容未考虑您的财务状况和目标，可能并不适合您。

It does not consider your financial circumstances and objectives and may not be suitable for you.