AI Semiconductors, VMware Driving Broadcom To New Heights: CEO Hock Tan

‘The whole premise behind going towards custom accelerators continues. It's not a matter of cost alone. It is that, as custom accelerators get used and get developed on a roadmap with any particular hyperscaler, that's a learning curve on how they could optimize the way the algorithms on their large language models get written and tied to silicon. And that ability to do so, it's a huge value-add [in] increasing algorithms that can drive the LLMs to higher and higher performance,’ says Broadcom President and CEO Hock Tan.

A dual boost in sales from semiconductor and VMware sales, driven by the steady growth in AI, helped Broadcom beat expectations during its fiscal 2025 second quarter and is pushing further growth going forward.

Hock Tan, president and CEO of semiconductor and infrastructure software giant Broadcom, Thursday told analysts on his company’s second fiscal quarter 2025 financial analyst conference call that the quarter was characterized by completely organic growth, given that it was the first full quarter that includes year-over-year comparisons since Broadcom acquired VMware.

“Revenue was driven by continued strength in AI semiconductors and the momentum we have achieved in VMware,” Tan said during his prepared remarks.

Second quarter semiconductor growth of 17 percent, year over year, was led by the 46 percent growth Broadcom saw in AI semiconductor revenue, which reached over $4.4 billion during the quarter, marking the ninth consecutive quarter of growth, Tan said.

That was also driven by double-digit custom AI accelerator growth as well as 170-plus percent AI networking growth, he said.

“AI networking, which is based on Ethernet, was robust and represented 40 percent of our AI revenue,” he said. “As a standards-based open protocol, Ethernet enables one single fabric for both scale-out and scale-up and remains the preferred choice by our hyperscale customers.”

Broadcom’s networking portfolio of Tomahawk switches, Jerico routers, and network interface cards is driving success within hyperscaler AI clusters, Tan said.

“And the momentum continues with our breakthrough Tomahawk 6 switch, just announced this week,” he said. “This represents the next-generation 102.4 terabytes per second switch capacity. Tomahawk 6 enables clusters of more than 100,000 AI accelerators to be deployed in just two tiers instead of three. This flattening of the AI cluster is huge because it enables much better performance in training next-generation frontier models through lower latency, higher bandwidth, and lower power,” he said.

Broadcom is also seeing progress in its XPU custom ASIC for AI, particularly in terms of helping three customers and four prospects deploy custom AI accelerators as discussed over six months ago, Tan said.

“We eventually expect at least three customers to each deploy 1 million AI-accelerated clusters in 2027 largely for training their frontier models, and we forecast – and continue to do so – a significant percentage of these deployments to be custom XPUs,” he said. “These partners are still unwavering in their plan to invest despite the uncertain economic environment. In fact, what we've seen recently is that they are doubling down on inference in order to monetize their platforms. And reflecting this, we may actually see an acceleration of XPU demand into the back half of 2026 to meet urgent demand for inference, on top of the demand we have indicated from training.”

Broadcom expects fiscal 2025 third quarter AI semiconductor revenue to be $5.1 billion, up 60 percent year-over-year, Tan said.

When asked by an analyst whether Broadcom’s expected third quarter 60 percent growth in AI semiconductor revenue represents a fair way of thinking about the business’ trajectory for this year and next, Tan said that's exactly what the company is trying to do.

“As we come into the second half of 2025 and we've improved visibility and updates we're seeing in the way our hyperscale partners are deploying data center AI clusters, we are providing you more of some level of guidance, visibility, what we are seeing, how the trajectory of ‘26 might look like,” he said. “I'm not giving any update on ’27. …. But what we're doing now is giving you more visibility into where we're seeing ‘26 headed.”

The AI Opportunity

When asked about the growth of AI networking, Tan said that part of the business grows hand-in-hand with the deployment of AI accelerated clusters.

“It doesn't deploy on the [same] timetable that accelerators get deployed, whether they are XPUs or GPUs,” he said. “It does happen. And they're deployed a lot in scale-out where Ethernet, of course, is the choice of protocol. But it's also increasingly moving into the space of what we all call scale-up within those data centers where you have much higher – more than we originally thought – consumption or density of switches than you have in the scale-out scenario. In fact, the increased density in scale up is five to 10 times more than in scale-out. And that's the part that kind of pleasantly surprised us, and which is why this past quarter, Q2, the AI networking portion continues at about 40 percent [growth] from when we reported a quarter ago for Q1. At that time, I said I expected it to drop. It hasn't.”

When asked about how Broadcom views the AI opportunity from the inference side of the business, Tan said he believes there’s no differentiation between training and inference when using merchant accelerators versus custom accelerators.

“The whole premise behind going towards custom accelerators continues,” he said. “It's not a matter of cost alone. It is that, as custom accelerators get used and get developed on a roadmap with any particular hyperscaler, that's a learning curve on how they could optimize the way the algorithms on their large language models get written and tied to silicon. And that ability to do so – it's a huge value-add [in] increasing algorithms that can drive the LLMs to higher and higher performance.”

This is not happening by segregating the approach between hardware and software, Tan said.

“You literally combine end-to-end hardware and software,” he said. “Take that journey. And it's a journey. They don't learn that in one year. Do it a few cycles, get better and better at it. And there lies the value, the fundamental value, in creating your own hardware versus using a third-party merchant silicon. You are able to optimize your software to the hardware and eventually achieve way higher performance than you otherwise could. And we see that happening.”

On the infrastructure software segment, Broadcom reported revenue of $6.6 billion, up 25 percent over last year and above the company’s outlook of $6.5 billion, Tan said.

“As we have said before, this growth reflects our success in converting our enterprise customers from perpetual vSphere to the full VCF [VMware Cloud Foundation] software stack subscription,” he said. “Customers are increasingly turning to VCF to create a modernized private cloud on-prem, which will enable them to repatriate workloads from public clouds while being able to run modern container-based applications and AI applications. Of our 10,000 largest customers, over 87 percent have now adopted VCF.”

Broadcom By The Numbers

For its second fiscal quarter 2025, which ended May 4, Broadcom reported total revenue of $15.00 billion, up about 20 percent from the $12.49 billion the company reported for its fiscal 2024 second quarter.

Revenue beat analyst expectations by $30 million, according to Seeking Alpha.

Total revenue included semiconductor solutions revenue of $8.71 billion, up 17 percent over last year, and infrastructure software revenue of $6.60 billion, up 25 percent.

Broadcom also reported GAAP net income of $5.0 billion or $1.03 per share, up significantly from last year’s $2.12 billion or 44 cents per share. On a non-GAAP basis, the company reported net income of $7.79 billion, up from last year’s $5.39 billion.

Looking forward, Broadcom is guiding fiscal 2025 third quarter revenue of about $15.8 billion.