Telecom Reseller / Technology Reseller News
Telecom Reseller
Podcast
Episodes
Listen, download, subscribe
Blaize and Nokia Target Real-World Edge AI with Hybrid Inference for APAC, Podcast
Doug Green, Publisher of Technology Reseller News, spoke with Dinakar Munagala, CEO & Co- Dinakar Munagala Founder of Blaize, and Joseph Sulistyo, SVP of Corporate Marketing, about Blaize’s push to make AI inference practical outside the data center—and why a new strategic collaboration with Nokia is designed to accelerate that shift, especially across Asia Pacific. Blaize positions itself as an AI computing company built around a purpose-built, fully programmable processor architecture it calls a graph streaming processor, paired with software intended to simplify development of “real-world” AI. Munagala framed the company’s focus as practical AI inference for environments like smart factories, smart cities, agriculture, defense, and other edge and hybrid deployments where latency, power, thermal limits, and operating conditions are non-negotiable. A centerpiece of the discussion was Blaize’s announcement that Nokia is strengthening edge AI capabilities through a strategic collaboration with Blaize to deliver hybrid inference solutions across APAC. Munagala and Sulistyo described the move as a signal that AI’s next phase isn’t only about large-scale training in centralized data centers, but about deploying inference where outcomes are realized—near cameras, sensors, machines, and field infrastructure. In their view, Nokia’s global reach in networking, automation, and integration creates a path to deliver end-to-end solutions that combine connectivity and compute for real deployments, not demos. Sulistyo emphasized the economics driving hybrid inference: cost-sensitive, power-constrained environments often cannot justify a single “monolithic” compute approach. Instead, he argued, the market is moving toward heterogeneous architectures—mixing different compute types to hit performance targets while controlling total cost of ownership. In APAC, he noted, the scale of deployments makes marginal savings meaningful, and hybrid designs become an operational requirement, not a preference. The conversation also connected edge inference to public-sector and community outcomes. Both executives highlighted smart-city use cases—such as traffic management, tolling, and first-responder automation—where real-time inference can improve accuracy and responsiveness while reducing labor-intensive processes. They extended that point to rural and underserved regions, arguing that “smart city” also includes municipalities and regional governments, where automation and analytics can unlock revenue (e.g., tolls and fines) while improving safety. Doug pushed on definitions and practicality, prompting Munagala to describe edge inference as compute performed as close as possible to the sensor—for example, processing video near a camera mounted on a pole, at a toll booth, or in a factory—so systems can detect events and respond with low latency. He added that some deployments may route inference to nearby on-prem servers or regional data centers, depending on architecture and proximity, and Blaize aims to support these variations with a common hardware/software platform. Blaize also addressed the “AI energy speed bump” impacting communities and operators—particularly where power availability and cost are constrained. Munagala said low power is foundational to Blaize’s design goals and argued that purpose-built inference architectures can reduce the burden associated with power-hungry AI approaches. Sulistyo added that the broader infrastructure conversation increasingly includes cooling realities (air and liquid) and the need to match the deployment environment to the right compute profile. To ground “real-world AI” in examples, the guests pointed to deployments including license plate recognition in complex, variable conditions and traffic anomaly detection (identifying behavior that deviates from normal flow). They described these as compute-intensive workloads that must run reliably outdoors and under harsh conditions, where latency and endurance matter as much as accuracy. They also discussed retail analytics as another example of edge inference delivering measurable business outcomes by connecting what happens in-store to revenue-driving decisions. Looking ahead, Munagala described the Nokia collaboration as a model for additional partnerships that bring inference solutions into production environments at scale. Sulistyo noted APAC is the initial focus, with other regions expected to follow based on demand, proof points, and the prioritization of specific use cases. To learn more about Blaize and its technology, visit https://www.blaize.com/.
Telecom Reseller / Technology Reseller News RSS Feed
