Locaweb's Data Center Planning Under Hardware Supply Pressure

Capacity planning used to be a math problem: forecast demand, order servers, wait a bit. That's getting harder. When memory is scarce and delivery windows stretch, the forecast has to be made earlier, and the fallback plan has to exist before the order is placed. Locaweb, a cloud hosting and digital services provider based in Brazil, is working through exactly that shift, and its experience shows how a service provider can respond without quietly cutting corners on what customers receive.

Why the board is looking at infrastructure again

Gustavo Gomes Salviano, Locaweb's chief technology officer, put it plainly: growing demand "brings the infrastructure challenge again to the board." In his words, the company now needs to "think more quickly and strategically in the infrastructure than we did before."

That's a notable admission for a company of Locaweb's age. It began as a hosting and email provider for small businesses, then widened into e-commerce, marketplace and financial solutions. Each new line of business adds different workloads, traffic patterns and uptime expectations, so the infrastructure underneath has to flex in ways a simple shared-hosting stack never did. Salviano's point is that the portfolio has grown faster than the old assumptions about how far ahead to plan.

Anticipating demand instead of reacting to it

The first change Salviano described is temporal. "First of all is to anticipate the demands," he said. "We anticipate … more than we've done before." Forecasting further out is only useful if the purchase and deployment timeline can absorb it, and that's where hardware availability becomes the binding constraint.

Memory is the component drawing the most attention in these conversations, alongside cost and delivery schedules. Locaweb and Dell have been talking about these supply issues since last year, looking at how to secure the right configurations in time for demand that's already arriving. Salviano said the team is re-engineering some of its infrastructure requirements specifically to deal with lead times and supply chain length. The goal, as he framed it, is to change the dependency without taking functionality away from products or customers.

That last clause is the hard part. Swapping in a different configuration is easy on a spreadsheet. Proving that it handles the same performance and capacity profile under real workloads is where the engineering work sits. Teams in this position typically run alternatives through the same benchmarks and failure scenarios as the original design before committing, because a cheaper or faster-to-ship box that degrades a customer-facing service just moves the problem.

Why the vendor relationship matters in practice

Salviano credited the longstanding relationship with Dell for making those comparisons workable. Dell's familiarity with Locaweb's environment lets the two sides judge alternatives against the performance and capacity the services actually need, rather than against generic specifications. "When we talk about our needs, our technical needs, they'll understand our products," he said. "They know what runs there."

That's a useful reminder for other buyers. Supply shortages tend to push teams toward whichever vendor has inventory, but a substitution made without deep knowledge of the workload is a gamble. Here the substitution decisions run through engineers who already know the stack. Diego Puerta, Dell's general manager for Brazil, joined Salviano in the discussion with theCUBE's John Furrier, which covered capacity planning, supply constraints and the engineering collaboration behind Locaweb's growth.

AI raises the stakes on the same hardware

The infrastructure question is getting sharper because Locaweb is bringing AI into its own operations and products. The company is building tools that let nontechnical staff put together solutions for their own operational problems. Salviano said employees are already using those tools to improve business processes, with IT supporting the work rather than owning every build.

That shifts load in a direction that's harder to forecast. When business users create their own workflows, demand for compute and storage stops tracking the IT roadmap and starts tracking how quickly teams adopt the tools. Capacity planning has to account for that bottom-up growth, which is one reason Salviano keeps returning to anticipation as the central discipline.

Puerta framed the broader point this way: "It's not an IT transformation; it's a business transformation leveraging IT. It must come from the leadership. IT is the enabler because what AI is doing is amplifying the possibilities." For Dell, he said, helping customers build the foundation for enterprise AI starts with understanding their business objectives, not with the hardware catalog.

What to watch

My read is that Locaweb's approach is less about any single hardware decision and more about treating supply as a planning input from the start. A few things will show whether it holds up. First, whether the alternative configurations keep service performance flat under peak load, since that's where substitutions usually reveal their flaws. Second, whether the earlier forecasting stays accurate as AI tools spread across nontechnical teams, which is a demand curve that's still taking shape. Third, how much of this depends on the Dell relationship specifically, and whether the engineering knowledge that makes substitution safe can be reproduced with other suppliers.

For infrastructure leaders, the takeaway is practical. Lead times now belong in the architecture conversation, not just the procurement one, and the people who know your workloads best need to be in the room when configurations change.