Broadcom CEO Tan: ‘AI Networking Demand Continues To Be Strong’

‘AI networking demand continues to be strong because networking is becoming critical as LLMs continue to evolve in intelligence and compute clusters have to grow bigger. The network is the computer, and our customers are facing challenges as they scale to clusters beyond 100,000 compute nodes,’ said Broadcom President and CEO Hock Tan on the company’s earnings call.

The AI semiconductor has become a big driver of Broadcom’s growth and will only increase in importance, President and CEO Hock Tan said Wednesday.

Tan, speaking to investment analysts during Broadcom’s third fiscal quarter 2025 quarterly financial conference call, called out growth in the company’s AI-focused XPU AI accelerator and its networking technology as especially important drivers.

Indeed, Tan, in emphasizing the importance of Broadcom’s networking business, borrowed a quote from Sun Microsystems Chief Scientist John Gage who said back in the mid-1980s that “the network is the computer.”

[Related: Scale Computing CEO On Broadcom-VMware Fallout, Nutanix Competition, Veeam]

“AI networking demand continues to be strong because networking is becoming critical as LLMs continue to evolve in intelligence and compute clusters have to grow bigger,” he said. “The network is the computer, and our customers are facing challenges as they scale to clusters beyond 100,000 compute nodes.”

Tan said Broadcom is addressing the “difficult challenge” of creating substantial bandwidth to share memory across multiple GPUs or XPUs within a rack.

“Today’s AI racks scale up to a mere 72 GPUs and 28.8 terabit-per-second bandwidth using [Nvidia’s] proprietary NVLink,” he said. “On the other hand, earlier this year, we launched Tomahawk 5 with Open Ethernet, which can scale up to 512 compute nodes for customers using XPUs.”

When it comes to scaling across multiple data centers, Tan said the June launch of Broadcom’s new Tomahawk 6 Ethernet-based 102.4 terabit-per-second switch flattens networks to two tiers instead of the previous three, resulting in lower latency and power consumption. That is coupled with the company’s Jericho3 internet routers, launched about two years ago.

“And today, we have launched our next-generation Jericho4 Ethernet fabric router with 51.2 terabits per second, gate buffering intelligence, [and] intelligent congestion control to handle classes beyond 200,000 compute nodes crossing multiple data centers,” he said

The biggest challenge to deploying larger clusters of compute for GenAI will be in networking, Tan said.

And for the past 20 years, what Broadcom has developed for Ethernet networking is entirely applicable to the challenges of scale up, scale out and scale across in generative AI, he said.

When it comes to Broadcom’s XPU AI accelerator business, Tan said demand for custom AI accelerators from the company’s three current customers continues to grow as each of them journeys at their own pace toward compute self-sufficiency.

Tan also said that Broadcom continues to gain share with those three customers and has already identified a fourth customer with over $10 billion worth of orders for AI racks based on Broadcom’s XPUs as a “qualified” customer. Tan said the other three potential customers in this space are already talking with Broadcom.

Tan did not mention any customer or potential customer by name.

“Reflecting this, we now expect the outlook for fiscal 2026 AI revenue to improve significantly from what we had indicated last quarter,” he said.

In total, Tan said Broadcom continues to make steady progress in growing its AI revenue for its fourth fiscal quarter 2025 and forecasts AI semiconductor revenue to be approximately $6.2 billion, up 66 percent year on year.

“Looking beyond what we’re just reporting this quarter, with robust demand from AI, bookings were extremely strong, and our current consolidated backlog for the company hit a record $110 billion,” he said. “Q3 semiconductor revenue was $9.2 billion as year-on-year growth accelerated to 26 percent. And this accelerated growth was driven by AI semiconductor revenue of $5.2 billion, which is up 63 percent year on year, and extends the trajectory of robust growth to 10 consecutive quarters.”

Demand for Broadcom’s non-AI semiconductors continues to be slow, with third fiscal quarter 2025 revenue of $4 billion, which was flat sequentially, Tan said.

“While broadband showed strong sequential growth, enterprise networking and server/storage were down sequentially,” he said. “Wireless and industrial were flat quarter on quarter as we expected. In contrast, in Q4, driven by seasonality, we forecast non-AI semiconductor revenue to grow low double digits sequentially to approximately $4.6 billion. Broadband, server/storage and wireless are expected to improve, while enterprise networking remains down quarter on quarter.”

Broadcom’s infrastructure software segment saw third fiscal quarter 2025 revenue of $6.8 billion, up 17 percent year over year, above its outlook of $6.7 billion as bookings continued to be strong, Tan said

Tan said he was most excited about what the company is doing with VMware.

“After two years of engineering development by over 5,000 developers, we delivered on a promise when we acquired VMware,” he said. “We released VMware Cloud Foundation Version 9.0, a fully integrated cloud platform which can be deployed by enterprise customers on-prem or carried to the cloud. It enables enterprises to run any application workload, including AI workloads, on virtual machines and on modern containers. This provides the real alternative to public cloud.”

Broadcom expects fiscal fourth quarter infrastructure software revenue to be approximately $6.7 billion, up 15 percent year on year.

During the question-and-answer period following prepared remarks, after an analyst asked Tan about expectations for customer adoption of VMware Cloud Foundation, Tan said that over 90 percent of its VMware vSphere customers have also purchased VCF.

“[But] because we have sold them on this and they have bought licenses to deploy doesn’t mean they are fully deployed,” he said. “Here comes the other part of our work, which is to take these 10,000 customers, a big chunk of them who have [bought] the vision of a private cloud on-prem, and working with them to enable them to deploy and operate it successfully on their infrastructure on-prem. That’s the hard work over the next two years that we see happening.”

At the same time, Tan said Broadcom sees VCF-based private cloud expanding across those customers’ data centers.

“That’s the second phase of my VMware story,” he said. The first phase was, convince people to convert from perpetual subscriptions [and in] so doing purchase VCF. Second phase now is to make that [VCF purchase] create the value they look for in private clouds on their premises, on their IT data centers. That’s what’s happening.”

Responding to that same analyst who asked whether increased VMware Cloud Foundation sales would impact the company’s hardware sales, Tan said that won’t happen.

“In fact, as they virtualize the data centers, we consciously accept the fact that we are commoditizing the underlying hardware in the data centers, commoditizing servers, commoditizing storage, commoditizing even networking, and that’s fine,” he said. “And by so commoditizing, we’re actually reducing the cost of investments in hardware in data centers for enterprises.”

VMware has about 300,000 customers, Tan said.

“We see the top 10,000 as being people where it makes a lot of sense to derive a lot of value in deploying private clouds using VCF,” he said. “We now are looking at whether the next 20,000 or 30,000 midsized companies see it the same way. Stay tuned. I’ll let you know.”