article
Data centers are back in strategy
Compute placement has returned to the executive agenda because AI economics, data gravity and power availability now shape product choices.
Emil Shirokikh · Published September 6, 2026 · Updated September 23, 2026 · 2 min read
Abstract
The cloud did not make physical infrastructure irrelevant. It abstracted it until workload scale and accelerator scarcity made location, cooling and network design visible again.
Place workloads deliberately
AI accelerators, power availability, data gravity and egress have made physical placement visible again. The answer is rarely one location for every workload.
Cost model
Segment training, inference, storage and regulated workloads. Model utilization, energy, egress, latency, capacity commitments and migration cost for each.
Decision record
Approve a mixed-placement map with assumptions that finance and engineering can revisit.
Challenge the conclusion
Most organisations should not build a facility. They should understand the infrastructure economics beneath their contracts.
Use this in a working session
Stress the model at low utilization and high growth, then identify which commitment becomes irreversible first.
BELTO editorial analysis. It does not describe a client engagement or claim a commercial result.
References
2 sourcesAuthor
Emil ShirokikhFounder
Founder of Belto Inc. Writes on engineering, venture building and applied intelligence.
Related
Read next
Most AI pilots should be killed sooner
The polite fiction around enterprise AI is that every pilot teaches something. Many teach only that nobody defined the decision, owner or failure cost.
AI agents need boundaries, not personalities
The market is decorating automation with human traits while neglecting permissions, reversibility and evidence.
A computer-vision demo is not a system
Benchmark accuracy says little about glare, drift, maintenance, latency or the operator who must act on an alert.