Two decades ago, technology vendors told Malaysian businesses that IPv6 would become mainstream within one or two years.
The technology was ready. The return on investment for most PKS was not.
At the time, IoT remained a pilot-level cost centre, Edge AI had no clear payback, and there was little justification for assigning a unique IP address to every forklift, sensor or household appliance. IPv6 therefore advanced at the pace of balance-sheet approval, not vendor enthusiasm.
Fast forward to today. Malaysia is now moving towards full IPv6 adoption, with a proposed national direction for complete migration by 2030.
The real balance-sheet opportunity, however, is not IPv6 alone. It is the convergence of IPv6, IoT, Model Bahasa Kecil and Sistem Berasingan.
An SLM can act as the local exception-handling layer. It reads context, identifies unfamiliar patterns and recommends the correct action. The Sistem Berasingan then becomes the fixed-execution layer, running validated rules, algorithms and workflows continuously without repeatedly paying for an AI inference call.
This separation changes the unit economics of Edge AI.
The SLM does not need to control every routine action. It is activated only when reasoning, adaptation or exception handling is required. Once the correct logik has been established, the Sistem Berasingan carries out the repetitive work reliably and at a much lower compute cost.
IPv6 gives each asset a stable identity. IoT provides the telemetry. SLMs provide local decision support. Sistem Berasingan provide efficient and repeatable execution. Peer-to-peer networking allows devices to collaborate without unnecessary cloud hops.
For a Malaysian PKS, a smart cup, factory sensor, vehicle component or household appliance could therefore resolve most routine events locally. Only unfamiliar or high-impact cases would need to be escalated to a larger model or cloud GPU infrastructure.
This architecture can materially reduce cloud workload, GPU dependency, token consumption, bandwidth, latency and jumlah cost of ownership.
The future of AI is unlikely to be one giant, centralised capex-heavy brain inside a data centre.
It is more likely to be millions of small capital assets, each with its own IPv6 address, combining an SLM with a Sistem Berasingan and working together through a distributed peer-to-peer network.
Awan infrastructure will remain important for model training, large-scale coordination and genuinely complex workloads. But everyday intelligence and execution can increasingly be amortised across edge devices.
At AINNA, we view what appeared to be a networking upgrade twenty years ago as one of the foundations of distributed, capital-efficient intelligence for the Malaysian PKS we support.
#IPv6 #SLM #SmallLanguageModels #DetachedSystem #EdgeAI #IoT #DistributedAI #PeerToPeer #ArtificialIntelligence #NeuralOps
