Meta Is Having Its AI Moment — But the Bond Market Is Sending a Warning

For almost two years, Mark Zuckerberg has been asking investors to trust him.

Spend more.

Build bigger.

Buy more chips.

Construct enormous data centers.

Hire expensive AI researchers.

And eventually, Meta would have something valuable enough to justify the bill.

Wall Street was not always convinced.

Now it may be getting its first real answer.

Meta shares have surged roughly 36% during September, putting the company on pace for its best month since 2013 and bringing its market capitalization close to the extraordinary $2 trillion threshold.

The catalyst has a name:

Muse.

Meta's new personal AI assistant quickly climbed to the top of Apple's U.S. App Store and has given investors something they have wanted to see for months:

Not another promise about AI.

An actual product people appear to want.

This is why the stock exploded

Muse represents something potentially much larger than another chatbot.

Meta already controls Facebook, Instagram, WhatsApp and Messenger—platforms used by billions of people.

If Meta can place an advanced AI agent directly inside that ecosystem, it doesn't necessarily need to convince consumers to develop an entirely new technological habit.

The distribution already exists.

Muse has also been designed to do more than answer questions. It can retrieve information, fill out forms and complete online tasks, creating potential subscription, commerce and business opportunities beyond Meta's traditional advertising machine.

That changed the conversation.

Jefferies recently increased its Meta price target from $710 to $875, while JPMorgan upgraded the company and raised its target to $820, citing the potential for AI agents to become another major growth engine.

For the bulls who spent months arguing that Meta's enormous AI investments would eventually produce something tangible, the rally feels like vindication.

But then comes another question.

Why are people who lend Meta money becoming more nervous at exactly the same time?

The credit market sees a different Meta

Stocks and bonds look at companies differently.

A shareholder asks:

How much could this company become worth?

A creditor asks:

How much money is this company committing, how much is it borrowing, and will all of these projects generate enough cash to justify the obligations?

Those questions can produce very different answers.

A recent Seeking Alpha analysis noted that Meta's credit-default-swap spreads have widened even as the stock has exploded upward. Credit-default swaps—or CDS—are essentially insurance against a company failing to repay its debt. Rising prices for that insurance generally indicate that investors perceive greater credit risk.

This does not mean Wall Street believes Meta is about to default.

Meta remains one of the world's most profitable companies.

But the trend is worth understanding.

Barron's reported last month that Meta's five-year CDS had reached roughly 95 basis points, part of a broader rise in credit protection costs among companies financing the AI infrastructure boom.

Why?

Look at the spending.

Meta is making an extraordinary bet

Meta began 2026 expecting capital expenditures between $115 billion and $135 billion.

By the first quarter, management had already increased that projection to $125 billion to $145 billion, citing higher component costs and additional data-center capacity.

Pause on that number.

Potentially $145 billion in one year.

That is not normal corporate expansion.

That is infrastructure being built at almost sovereign scale.

And Meta is increasingly using financial engineering and outside capital to accelerate the buildout.

In July, Meta announced a partnership with BlackRock for a massive El Paso data-center campus. The project is expected to cost approximately $14 billion, with BlackRock-managed funds owning 80% and Meta retaining 20%. The structure includes about $12.5 billion in debt financing.

Meta also issued $25 billion of senior notes this year across maturities stretching as far as 2066.

This is where the stock market and credit market begin looking at the same AI revolution through different windows.

What happens if AI delivers?

Imagine Meta is correct.

Muse becomes deeply integrated into daily life.

Businesses begin paying Meta for AI services.

Consumers subscribe.

AI dramatically improves advertising.

Commerce begins moving through Meta's agents.

AI increases engagement across Instagram, Facebook and WhatsApp.

Suddenly, tens of billions spent on infrastructure look less like reckless spending and more like the cost of constructing the next major computing platform.

Under that scenario, today's stock rally could make sense.

And Meta already has an enormous cash-generating machine supporting the transition. During the first quarter, the company reported $32.23 billion in operating cash flow and held more than $81 billion in cash, equivalents and marketable securities.

This is not a speculative startup borrowing money while waiting for revenue.

Meta has revenue.

A lot of it.

But what happens if everyone builds too much?

Now consider the opposite scenario.

Meta spends $140 billion.

Amazon spends enormous amounts.

Google spends enormous amounts.

Microsoft does the same.

OpenAI, Anthropic, Oracle and other players sign gigantic compute agreements.

Hundreds of billions of dollars pour into data centers.

Then AI prices fall.

Models become cheaper.

Competition compresses margins.

Consumers refuse to pay substantial subscription fees.

Or the infrastructure becomes outdated faster than companies anticipated.

Suddenly those enormous investments still exist.

And so does the financing.

The Bank for International Settlements has already warned that AI hyperscalers are increasingly financing capital expenditures through both traditional borrowing and more complicated off-balance-sheet structures. CDS spreads across the sector have risen as investors question how much economic return the infrastructure ultimately produces.

Capital Group reached a similar conclusion this week: investment-grade markets have absorbed the extraordinary issuance so far, but credit spreads for hyperscalers and other AI-linked borrowers have widened even while broader investment-grade spreads remained relatively stable.

In other words:

The bond market is not saying AI will fail.

It is saying:

You are spending an enormous amount of money. Prove the economics work.

And that may be the real Meta story

The bullish article and the bearish article do not necessarily contradict each other.

They may simply be measuring different risks.

The equity market sees Muse exploding in popularity and thinks:

Meta might have found the product that makes its AI investment worthwhile.

The credit market sees $125 billion to $145 billion of annual capital spending, tens of billions in new financing and an industry racing to construct infrastructure faster than anyone can precisely calculate future demand—and thinks:

This is becoming expensive.

Both observations can be correct.

That is what makes Meta fascinating right now.

Zuckerberg finally has to prove the second half

Building AI is no longer the challenge.

Meta has demonstrated that it can spend.

It can recruit.

It can build data centers.

It can create frontier models.

And with Muse, it may now have demonstrated something even more important:

It can create consumer excitement.

The next test is monetization.

How many Muse users remain after the novelty fades?

How many pay?

How much commerce flows through it?

How much does AI improve Meta's existing advertising business?

And eventually:

Does the cash coming back justify the mountain of money going out?

That is the battle investors should watch.

Meta's stock market celebration says Zuckerberg's bet is beginning to work.

The credit market is simply reminding everyone that the bet has become enormous.

And perhaps that is the healthiest way to interpret what is happening.

The bulls finally have evidence.

The creditors still want proof.