There are very few stocks where both the bull and bear cases sound equally compelling.

CoreWeave is one of them.

The company sits at the center of the largest infrastructure buildout since cloud computing, yet it also represents one of the most capital-intensive and highly valued businesses in the market.

Whether CoreWeave becomes the "AWS of AI" or one of history's most expensive infrastructure mistakes depends on a handful of variables that investors can monitor today.

The biggest mistake bears make is treating CoreWeave like an equipment leasing company.

It isn't.

It's building AI infrastructure.

The distinction matters.

Amazon wasn't renting servers.

It was building AWS.

Digital Realty wasn't leasing warehouses.

It was building internet infrastructure.

CoreWeave isn't renting GPUs.

It's constructing AI factories.

Every AI application ultimately requires:

It doesn't matter whether the winner is:

They all need compute.

CoreWeave sits directly in front of that demand.

Unlike most technology companies...

CoreWeave isn't hoping customers appear.

Customers are asking for capacity faster than the company can build it.

Management has disclosed tens of billions of dollars of contracted backlog and continues signing large multi-year agreements with hyperscalers and AI-native companies.

If AI adoption continues, the constraint may not be customers.

It may be available power.

Traditional investors focus on:

Negative Free Cash Flow.

The bull focuses on:

Incremental Return on Invested Capital.

If every new AI campus earns:

15–20% returns...

then raising billions becomes rational.

Every additional dollar invested increases intrinsic value.

The larger the balance sheet becomes...

the larger earnings eventually become.

Depreciation exceeds $33 billion in the 2030 consensus.

Yet much of CoreWeave's assets consist of:

These assets may remain productive for decades.

If accounting depreciation materially exceeds true economic depreciation...

reported earnings understate intrinsic earning power.

Management has already demonstrated improving financing costs.

Better customers →

Lower financing costs →

Higher project returns →

More projects →

More cash flow →

Lower financing costs.

Infrastructure businesses become extraordinarily valuable when this flywheel begins.

Most investors focus on LLMs.

The next decade may belong to:

Every one of those workloads requires massive infrastructure.

CoreWeave benefits regardless of which AI architecture wins.

Today investors think CoreWeave sells compute.

Tomorrow it could sell:

AWS didn't stop with servers.

Neither may CoreWeave.

History is filled with businesses that looked unstoppable.

Railroads.

Fiber optics.

Solar.

Telecom.

Shipping.

The internet.

Demand proved real.

Returns on capital did not.

Today investors aren't paying for today's business.

They're paying for perfection.

Current expectations assume:

When expectations become this high...

even excellent execution may disappoint shareholders.

Amazon built AWS.

Microsoft built Azure.

Google built Google Cloud.

Meta increasingly builds internally.

Every dollar a hyperscaler spends on its own infrastructure...

is one less dollar available to third-party providers.

The outsourcing window may narrow over time.

Today's infrastructure is optimized for current AI models.

But AI changes rapidly.

What if:

Demand for expensive GPU clusters may not grow as quickly as investors expect.

Consensus expects debt to increase substantially over the coming years.

Leverage works wonderfully...

until it doesn't.

Higher interest rates.

Slower deployments.

Customer delays.

Lower utilization.

Any of these could materially reduce equity value.

Analysts expect well over $200 billion of cumulative capital expenditures through 2030.

That creates one enormous question.

Will those investments actually earn superior returns?

If they earn:

20%

Shareholders win.

If they earn:

8%

Shareholders lose.

The difference between those outcomes determines almost all long-term value.

Bulls argue depreciation overstates economic cost.

Bears argue the opposite.

GPUs become obsolete rapidly.

Networking evolves.

Cooling improves.

Power density changes.

If infrastructure requires continual replacement...

then today's depreciation may actually understate future capital needs.

Many industries experience:

Demand ↑

Revenue ↑

Capital spending ↑

Returns ↓

The industry wins.

Investors don't.

Railroads.

Telecom.

Solar manufacturing.

Fiber.

Shipping.

The bear argues AI infrastructure eventually follows the same path.

Almost every debate surrounding CoreWeave reduces to one variable.

What return will the company earn on each new dollar invested?

Imagine two worlds.

Every new AI factory earns 20% after-tax returns.

Management continues reinvesting for a decade.

Book value compounds rapidly.

Intrinsic value explodes.

Today's valuation ultimately looks inexpensive.

Every new AI factory earns 8%.

Competition increases.

Returns normalize.

Capital markets tighten.

The premium valuation collapses.

Revenue continues growing...

while shareholder returns disappoint.

Both sides are making rational arguments.

The bulls are betting that AI infrastructure resembles cloud computing—a once-in-a-generation buildout where early leaders compound capital at extraordinary rates for many years.

The bears are betting that AI infrastructure resembles railroads, telecom, fiber optics, and shipping—industries where demand proved enormous but excess capital ultimately drove returns toward the cost of capital.

The answer will not be found in next quarter's earnings.

It will be found in five metrics:

Those five variables—not AI headlines—will determine whether CoreWeave becomes one of the greatest infrastructure investments of the decade or one of the market's most expensive lessons in capital allocation.