‘We expect this to represent an AI revenue, serviceable addressable market, or SAM, for XPUs and network in a range of $60 [billion] to $90 billion in fiscal 2027 alone. We are very well positioned to achieve a leading market share in this opportunity and expect this will drive a strong ramp from our 2024,” Broadcom CEO Hock Tan told investors during the company’s fourth-quarter earnings call.

Broadcom CEO Hock Tan (pictured) said the company has a “massive” opportunity with three existing hyperscale customers—and two more potential hyperscale customers—which plan to buy millions of Broadcom’s XPUs to accelerate AI workloads.

“We see our opportunity over the next three years in AI as massive,” he said. “Specific hyperscalers have begun their respective journeys to develop their own custom AI accelerators, or XPUs, as well as network these XPUs with open and scalable Ethernet connectivity for each of them. This represents a multiyear, not a quarter-to-quarter, journey.”

Tan steered the semiconductor through a strong fourth quarter with 11 percent growth in its legacy products year over year. Tan’s remarks about the “massive” AI bets that the company’s big three hyperscale customers are making sent Broadcom’s shares rocketing 24 percent higher Friday and saw its valuation top $1 trillion.

[RELATED: Broadcom CEO Hock Tan: VMware Bookings ‘Continued To Accelerate,’ While AI Chip Sales Surge]

“In 2027 we believe each of them plans to deploy 1 million XPU clusters across a single fabric,” he said. “We expect this to represent an AI revenue, serviceable addressable market, or SAM, for XPUs and network in a range of $60 [billion] to $90 billion in fiscal 2027 alone. We are very well positioned to achieve a leading market share in this opportunity and expect this will drive a strong ramp from our 2024.”

Tan teased investors that more large hyperscalers are kicking Broadcom’s XPU tires, which could “compound” the company’s already heady revenue projections.

“We have been selected by two additional hyperscalers and are in advanced development for their own next-generation, AI XP use. We have line of sight to develop these prospects into revenue-generating customers before 2027,” Tan said. “So the reality going forward for this company is that the AI semiconductor business will rapidly outgrow the non-AI semiconductor business.”

Broadcom shares closed last week at $224.80, up $44.14, driving its market cap to more than $1.05 trillion.

VMware, which the company did not own in this time frame last year, contributed $3.85 billion in revenue for the quarter. When factoring in those sales, Broadcom achieved 51 percent growth in the comparable quarter.

Broadcom completed its $69 billion acquisition of VMware in November 2023.

For the full fiscal year, VMware earned about $13.8 billion in revenue. VMware ended its last full fiscal year as an independent company with annual revenue of $13.35 billion in March 2023.

Broadcom’s revenue including VMware was 44 percent higher.

Tan said he has cut VMware’s expenses in half since taking over the company a year ago, dropping quarterly operating costs from $2.4 billion to $1.2 billion in that time frame and boosting its margins from 30 percent to 70 percent.

He said VMware was installed on 21 million CPU cores last quarter versus 19 million CPU cores in the previous quarter, and 4,500 of the company’s largest 10,000 customers have converted their virtualization estates to Broadcom’s bundle of software called VMware Cloud Foundation.

Market research firm Forrester expects VMware customers to “shrink their deployments by 40 percent in favor of alternatives this coming year.

It expects that increased migration to the public cloud, on-premises alternatives and new architectures will drive this reduction, analysts noted in its “Predictions 2025: Technology Infrastructure and Operations” report.

On the earnings call Broadcom CFO Kirsten Spears said going forward Broadcom will no longer break out the company’s virtualization revenue separately from its mainframe and other legacy software business.