The unprecedented memory shortage has caught IT leaders off guard amid skyrocketing hardware price as supply stretching out to 2027 has all but been gobbled up. But of all the firms looking to capitalize on the capacity shortage, Broadcom is eyeing it as a means to entice customers to stick with its VMware Cloud Foundation (VCF) platform as a software-driven “systems approach” to surviving the memory crunch.

Speaking with SDxCentral, Sabina Anja, chief technologist for Broadcom’s VCF division, framed the platform as a means to help customers make constrained memory and storage budgets go further through software intelligence instead of waiting for or overpaying for new hardware.

“It’s a systems approach of combining the power of software to make the most out of the hardware that you’re given. That way, there’s a shortage – great, how can I squeeze the most out of what I have, which is something that your CFO and your CIO will love as a statement, because that’s what they always want,” Anja said.

Among the existing offerings in VCF 9.0 Anja highlighted as a means to supplant hardware spend with software optimization was non-volatile memory express (NVMe) tiering, which treats local NVMe devices as a secondary memory tier. Essentially, this intelligently moves inactive (cold) memory pages to the lesser tier, ensuring optimal capacity is reserved for more expensive dynamic random access memory (DRAM).

Anja argued NVMe tiering’s debut in VCF last year was a proactive play, claiming: “We saw this coming.”

“For the past two years, since the Broadcom acquisition, a lot of what we've done is build out these software-defined solutions by thinking more in terms of the platform,” Anja said, noting that VCF has converged into a software stack that's able to commoditize the hardware behind it. “We kind of want to get to a point where I don't care what you buy behind this; what you care about is how you leverage what you have.”

The memory shortage has compounded existing questions around costs at a time when some long-time VMware customers are already questioning their position amid licensing changes around the VCF platform.

But while some customers are mulling their stack costs for the next few years, Anja and the VCF team argue that the memory shortage is actually proof that its existing customers haven't been getting full value from what they already pay for.

Anja acknowledged as much, noting that most customers aren't hovering over virtual machined (VM) allocations from 2017 asking whether they're still the right size, and that the shortage is now the forcing function to make them look.

“Most [customers] are just trying to get through their days and the projects and the work they need to do," Anja said. "So now you can recapture a lot of stale resources. You can have that level of visibility. You can understand how your resources are being used. And while it's not the biggest temptation to go back in time and fix the past. You can also do that going forward.”

Beyond clearing up existing VM waste, with memory and server costs spiking, Anja said most of the CIOs and IT leaders she speaks to are under pressure not to slow down AI projects.

“Whether you're a bank or pharma, somebody has an AI demand on you, and guess what? That's a massive requirement of memory [and] storage,” Anja said. “The CIO mandate sometimes is like: 'we need to show that we can be more productive with AI,' so they're not willing to compromise on the AI.”

Instead, Anja argued the crunch is forcing a more disciplined approach to how new workloads are deployed in the first place – something she claimed VCF is designed to enforce.

“You can guardrail it in a lot of ways, and say, 'not bigger than this much of a GPU slice, no more than this much memory, no more than this much storage provisioning,'” Anja said, advocating for “conservative deployment” by default and only granting more resources by exception.

That discipline, she contends, goes beyond simple means to an end like NVMe tiering, adding: “The ability to control a lot of these deployments is a lot more important than just the one feature.”

The Broadcom exec drew a parallel to data center power consumption: a problem nobody optimized aggressively until AI workloads started threatening local power grids. The memory crunch, she suggested, is a similar forcing function.

“Good challenges can produce good engineering,” Anja said. “When resources are unlimited, you're not forced to think twice about your architecture.”