Broadcom CEO Hock Tan rejects “agentic AI replaces software” worry, points to VMware demand

Broadcom CEO Hock Tan used a blunt line to address a growing anxiety in enterprise software circles: the idea that “agentic AI” could reduce or eliminate the need for traditional software licenses by automating what those products do. During the company’s fiscal second-quarter earnings call on June 3, Tan told investors, “We’re not seeing it.”

The comment directly challenged a “Saaspocalypse”-style narrative that has been circulating among software executives and investors, which holds that increasingly autonomous AI systems will take over workflows that currently require paid tools. In Broadcom’s case, Tan argued that the opposite is happening, at least for the infrastructure software VMware provides, and that demand is tied to enterprises expanding compute capacity to support AI workloads.

Broadcom has positioned VMware as a layer of software that helps organizations deploy, run, and manage workloads on virtualized infrastructure, including the environments where AI workloads run. In the same call discussion, Tan pointed to momentum in VMware growth driven by the volume of CPU cores and GPUs being sold together for AI builds, and he said Broadcom expects that demand to continue for “multiple quarters” as AI capacity ramps up.

Financially, Broadcom’s results on June 3 were dominated by AI semiconductor growth but also underscored why investors have focused on the company’s broader software platform. In its fiscal Q2 2026 results, Broadcom reported consolidated revenue of $22.187 billion, up 48% year over year, along with AI semiconductor revenue of $10.8 billion, up 143% year over year. The company also reported adjusted EBITDA of $15.244 billion (69% of revenue) and free cash flow of $10.262 billion, according to its earnings materials.

The company’s near-term guidance continued that AI-forward trajectory. Broadcom guided third-quarter fiscal 2026 revenue to approximately $29.4 billion, and it expected AI semiconductor revenue to grow to $16.0 billion, which the company’s guidance implied as more than a doubling versus the prior year period. Broadcom did not provide a separate, AI-agent-specific metric for VMware in the disclosed top-line guidance, but Tan’s remarks were designed to connect the software outlook to the same AI infrastructure build-out driving Broadcom’s chips and networking.

Part of Broadcom’s message appears in how it has recently framed VMware’s product roadmap around production AI in private cloud. VMware Cloud Foundation 9.1 is described by Broadcom as a secure, cost-effective private cloud infrastructure platform for production AI workloads, with updates intended to simplify scaling, lifecycle operations, and governance as organizations add more AI-capable infrastructure.

In that framing, VMware’s value is not just “more automation,” but the operational layer enterprises still need when software is managing sensitive data, enforcing policies, coordinating security, and interoperating with heterogeneous hardware. That is the central tension in the agentic-AI licensing debate: if agents can execute tasks, enterprises may still need the underlying infrastructure software that keeps those systems reliable, governed, and performant in production.

A key caveat is that Broadcom’s public disclosures tied to the earnings period did not quantify how much of VMware’s performance, pricing, or renewal timing is specifically attributable to agentic AI versus broader AI infrastructure expansion. Tan’s “We’re not seeing it” comment was also reported through third-party transcripts of the call, and Broadcom’s formal materials are careful about uncertainties, including demand for data center virtualization products and customer acceptance of its software strategy.

For the market, the next test is whether VMware’s momentum holds as new AI deployments move from experimentation to broader production rollouts. Investors are likely to watch Broadcom’s quarterly commentary on VMware demand, any discussion of infrastructure software bookings, and whether product updates aimed at “production AI” continue to translate into durable customer spending rather than displacement pressure on enterprise software budgets.