Broadcom touted a flurry of product and service updates at its VMware Explore conference in Las Vegas this week including a “private AI cloud” and an “AI factory” – ways to let customers run validated LLMs and SLMs alike without hitch on their own bare metal, with a range of token monitoring, and “AI metrics observability” capabilities baked in.
Touting a range of new and forthcoming “private AI” services, Broadcom said it would be offering an AI gateway for model governance across on-prem and cloud environments with “intelligent prompt routing, token and usage rate-limiting, and application authorization” for those who haven’t already bought or built their own equivalent on open-source.
Among the other updates: “A new partnership with MetalSoft to deliver integrated heterogeneous bare metal automation for VCF that drops bare-metal provisioning time from weeks to minutes. The integration will help IT provision or repave physical servers from multiple vendors directly through the VCF management console, unifying the software and hardware lifecycle into a single operational model and eliminating the need for vendor-specific tools for hardware and firmware management.”
Then there’s “TrueSource by Broadcom.” That’s a portfolio/library of secure and hardened container images that also spans “clean-room builds of the broader Java ecosystem, Python, and Node.js” software and “trusted artifacts, support, and deployment expertise for PostgreSQL, RabbitMQ, MySQL, and Valkey data engines” – making Broadcom the latest company to enter a goldrush for secure OSS components amid mounting supply chain security risk. All will be human-audited, it said.
(Broadcom cited a recent study by 1Password that showed AI patches tend to cause a lot of issues, saying a mere “26 percent of 6,000 AI-generated patches fixed vulnerabilities without breaking applications. That’s right: These days you pay extra to ensure a hallucinated AI patch doesn't break your production database/hypervisor environment…)
The catchphrase of choice across Broadcom’s flurry of press releases was getting from “metal to model” faster (it’s currently “slow, complex, and expensive,” said VMware cloud foundation division CPO Paul Turner in a canned statement) and bringing models “to the data”...
To that end it has validated a range of LLMs and SLMs to run without breaking things on its VCF proposition, underpinned by support for mixed CPU and GPU compute across AMD, Intel, and NVIDIA . These include Google’s Gemma 4 family of models, China’s GLM 5.2 and Qwen 3.8 (27B) models, as well as Japan’s cotomi model from NEC, it said.
"Many VMware customers are still focused on a more immediate question: how to balance support, licensing costs and long-term platform strategy," commented Shane O’Rourke, VP, global support services, VMware at Spinnaker Support.
He added: "Customers will ultimately evaluate whether these AI capabilities justify the additional investment required..."
Broadcom has an updated product compatibility lists for VMware infrastructure, including VMware Cloud Foundation, here.