AI inference workloads to surpass model training by 2027
JLL forecasts that AI inference workloads will surpass training workloads by 2027, necessitating a shift toward distributed edge computing and low-latency network infrastructure. This transition highlights the growing strategic importance of edge-based processing and fibre connectivity for real-time video analytics and other streaming-related applications.
Key Takeaways
- AI inference workloads are projected to surpass model training by 2027, flipping the current infrastructure paradigm.
- Global data center capacity is expected to nearly double to 200 gigawatts by 2030, with AI representing 50% of that capacity.
- Inference performance depends on metropolitan network connectivity, shifting strategy from pure compute to distributed fiber infrastructure.
- Edge AI deployment reduces latency for real-time video analytics and digital twins by processing data closer to the end user.
Why It Matters
The transition to inference-dominant AI shifts the technical bottleneck from raw GPU power to network latency and edge density. For the streaming industry, this move supports the deployment of real-time video analytics and personalized ad-insertion at scale without the round-trip delay of centralized clouds. As inference moves to metropolitan and telco edge sites, CDN operators and low-latency network providers become essential orchestrators of the AI stack. Watch for a rise in 'AI-native' mobile network investments, such as Nokia's recent partnerships, to support this distributed demand.
Additional Context
The shift toward distributed inference aligns with fiscal trends seen across the hardware and data center sectors. Per NVIDIA's February 2025 earnings report, demand for its Blackwell architecture is increasingly driven by 'inference time scaling,' with the data center segment reaching a record $35.6 billion in quarterly revenue. NVIDIA’s CEO Jensen Huang noted that reasoning models consume 100x more compute than traditional queries, further accelerating the need for high-throughput infrastructure. This is also reflected in the data center colocation market; according to an August 2025 release from Equinix, AI workloads now drive approximately 60% of their largest deals as enterprises move models into production environments.
Simultaneously, the video delivery segment is seeing specialized infrastructure investments to meet these needs. Per SNS Insider in June 2026, the AI video analytics market is projected to reach $64.48 billion by 2035, growing at a 22.8% CAGR as real-time monitoring and automated decision-making move to the edge. Market leaders like Cisco and Genetec are already integrating AI-powered monitoring platforms with cloud-native analytics to reduce latency. Additionally, Deloitte reported in late 2025 that social video platforms now dominate daily media consumption, creating massive live-data streams that require the decentralized inference capabilities JLL forecasts for 2027.
Read full article at digitaljournal.com
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source