To The Moon

Broadcom’s AI business is no longer a “future opportunity.”

It is already a $16.7 billion quarterly business.

The harder question now is whether Broadcom can turn that explosive ASIC growth into equally strong profit and free cash flow growth.

That distinction could become increasingly important for AVGO investors.

The AI numbers are getting difficult to ignore

Broadcom’s fiscal Q3 AI semiconductor revenue reached $16.7B, up 221% year over year and 54% sequentially.

Management expects another acceleration in Q4, with AI semiconductor revenue guided to $21.7B, up 236% YoY.

Broadcom is also forecasting roughly $58B of AI semiconductor revenue for fiscal 2026, followed by approximately $115B in fiscal 2027.

On the demand side, the story is clearly powerful.

But rapid revenue growth creates a different question:

How much of that growth becomes incremental profit?

This is where the margin test begins

Broadcom’s Q3 gross margin was 75%, down 210 basis points sequentially.

That doesn’t mean the ASIC business is becoming unattractive.

Broadcom still generated $13.7B of free cash flow, equivalent to 46% of quarterly revenue, while non-GAAP operating margin reached 67.9%.

But the margin movement is worth watching.

Custom AI accelerators can have different economics from Broadcom’s other semiconductor products. As AI chips become a much larger proportion of the business, product mix can affect the company’s overall margin profile.

That creates an important investor question:

Can AI revenue double without sacrificing the profitability that makes Broadcom valuable?

ASICs have a powerful advantage

The reason hyperscalers are increasingly interested in custom accelerators is not simply that they want an alternative to Nvidia.

At massive scale, customers can justify designing hardware around their own workloads.

If a custom ASIC can deliver better performance per dollar or better power efficiency for a specific workload, the economics can become compelling.

Broadcom sits in an interesting position because it doesn’t need to build the entire AI ecosystem itself.

It provides the custom silicon and networking infrastructure while the hyperscalers and AI companies provide the workloads.

That makes Broadcom more of an AI infrastructure supplier than a direct competitor fighting Nvidia for every GPU sale.

But there is a trade-off

The bigger the AI business becomes, the more investors need to understand customer concentration and execution risk.

A handful of extremely large customers can create enormous revenue opportunities.

They can also possess substantial negotiating power.

And custom silicon is not a simple commodity sale. Each generation requires design work, qualification, manufacturing capacity and advanced packaging.

If customers push for lower prices as volumes increase, revenue could grow faster than margins.

That’s why simply extrapolating Broadcom’s AI revenue growth could be dangerous.

The financing question is becoming interesting too

Broadcom’s latest filing shows that it has created an AI infrastructure financing platform designed to support more than 20 gigawatts of XPU-based compute capacity through 2028.

The structure is designed to help customers fund the enormous upfront cost of AI infrastructure.

That could accelerate deployment.

But it also shows how capital-intensive this AI buildout has become.

And recent developments involving Anthropic highlight the scale of the commitments being made around Broadcom’s custom AI infrastructure.

For investors, this means the ASIC opportunity isn’t just about chip demand anymore.

It is increasingly connected to customer financing, infrastructure deployment and long-term commitments.

So is the ASIC trade still attractive?

The answer depends on what happens to the earnings quality behind the revenue growth.

I’d watch four numbers particularly closely:

1. AI revenue growth

Is the rapid growth continuing?

2. Gross margin

Does increasing AI mix put further pressure on margins?

3. Free cash flow

Are higher AI revenues converting into actual cash at a similar rate?

4. Customer commitments

Are today’s enormous AI infrastructure plans turning into sustained shipments and revenue?

If AI revenue continues climbing while margins and free cash flow remain resilient, the ASIC story becomes much more than a revenue-growth narrative.

But if revenue doubles while profitability becomes increasingly diluted, investors may eventually question how much they are paying for that growth.

The bigger takeaway

Broadcom’s AI opportunity is no longer being tested on whether customers want custom chips.

That part increasingly appears established.

The next test is harder:

Can Broadcom scale ASIC revenue at extraordinary speed without giving away too much of the economics?

For AVGO, the next phase of the AI story may therefore be less about “How much revenue can Broadcom generate?”

And much more about:

“How much high-quality earnings and free cash flow does every additional dollar of AI revenue create?”

That is the margin test that could determine whether the ASIC boom becomes a durable earnings engine or simply another case of spectacular revenue growth meeting tougher economics.