Think Grid AI compute service launches with 27% lower costs

The launch of Think Grid AI compute introduces a high-density alternative for streaming platforms and AI-native firms requiring heavy inference and voice application support. By providing dedicated bare-metal nodes instead of virtualized slices, the service addresses the latency and throughput bottlenecks often found in containerized cloud environments. This move signals a shift toward specialized, regional infrastructure providers competing directly with global hyperscalers on cost transparency and hardware utilization. As streaming companies integrate more generative AI for personalization and metadata, the ability to run multiple models on shared hardware via ILM orchestration could significantly reduce operational overhead. Watch for Think to announce specific global expansion sites beyond Riyadh to gauge its impact on international data sovereignty requirements.

The launch of Think Grid enters a market where NVIDIA's Blackwell architecture is already being deployed across telecom and enterprise AI infrastructure at scale. In June 2026, Ericsson launched its AI in RAN commercial software subscription claiming up to 20% higher downlink throughput across more than 15 live deployments, demonstrating that GPU-accelerated compute is moving from experimental pilots into production network operations. Nokia and Indosat Ooredoo Hutchison announced a GPU-accelerated AI-RAN partnership in Indonesia on June 8, expanding the Nokia-NVIDIA architecture already adopted by T-Mobile US, SoftBank, and Vodafone. These deployments validate the same Blackwell-class hardware that Think Grid packages for enterprise customers, suggesting that bare-metal GPU pooling approaches like ILM orchestration could find traction among operators seeking dedicated compute without hyperscaler abstraction layers.

The competitive dynamics between Ericsson and Nokia illustrate how vendors are positioning AI infrastructure as a strategic differentiator rather than a commodity. Light Reading reported that Nokia's entire RAN strategy is now built on its partnership with Nvidia, cemented by a $1 billion investment, with Layer 1 RAN functions designed to run on Nvidia's CUDA platform and GPUs. Meanwhile, Nokia combined with AWS and Databricks to build a telco AI control layer at DTW Ignite, positioning its Autonomous Network Fabric as an operating system spanning radio, core, transport, and service domains. Nokia claims operators using its autonomous networks portfolio are achieving automation rates above 90%, service delivery times under four hours, and up to 85% reduction in slice rollout time. For Think, these moves confirm that the market for specialized AI compute is fragmenting: some buyers want integrated vendor stacks, while others seek flexible bare-metal access to accelerators without platform lock-in.

On the technical side, the divergence between Ericsson and Nokia approaches highlights the trade-offs Think Grid is targeting with its ILM orchestration layer. , running on AWS via Amazon Bedrock with more than 20 cloud-native AI applications across OSS and BSS functions. The company acknowledges its stack is cloud-agnostic in principle but notes the most mature tooling remains on AWS. , framing the choice as a CFO-versus-CTO decision. Think Grid's pitch of dedicated bare-metal nodes with pooled GPU memory sits between these approaches, offering hardware-level performance without requiring operators or enterprises to commit to a single vendor's orchestration framework.