AI back-end networking spend to hit $1T by 2030: report Data center switch spending for AI back-end networks is projected to reach $1 trillion by 2030, with scale-up interconnectivity accounting for more than half the total market, according to a report from Dell'Oro Group. Nvidia's NVLink currently dominates the scale-up segment, but Dell'Oro Group expects Ethernet to become the leading scale-up technology by the end of the decade as hyperscalers adopt alternatives like Ultra Accelerator Link (UALink) for custom GPUs. The analyst firm noted that multiple networking technologies will coexist, creating multibillion-dollar opportunities, and that co-packaged optics (CPO) switches are expected to hit shelves this year. Data center switch spending for AI back-end networks is set to skyrocket to $1 trillion over the next five years, with scale-up interconnectivity accounting for more than half the total market by 2030. Projections from Dell’Oro Group https://www.delloro.com/market-research/data-center-infrastructure/data-center-switch-ai-back-end-networks/ suggest the interconnect inside the rack is becoming as commercially significant as the network between racks, with demand being so high that scale-up is carving out its own multihundred-billion-dollar category. Nvidia’s NVLink currently dominates the scale-up scene, while demand for its switches has soared by triple digits https://www.sdxcentral.com/news/nvidia-tops-ethernet-data-center-switch-charts-by-revenue-as-sales-soar/ this year alone. The firm leads to such an extent that Nvidia CEO Jensen Huang crowned his company https://www.sdxcentral.com/news/nvidia-thinks-it-has-dethroned-cisco-as-the-worlds-biggest-networking-company/ “the largest networking company in the world.” Dell’Oro Group anticipate that rival scale-up technologies will challenge Nvidia’s dominance, with hyperscalers beginning to turn toward Ethernet https://www.sdxcentral.com/news/ethernet-dominates-ai-networking-as-switch-sales-double-but-infiniband-rebounds/ and Ultra Accelerator Link UALink https://www.sdxcentral.com/news/ualink-consortium-takes-another-swing-at-nvidias-nvlink-supremacy-with-specification-20/ alternatives to support the deployment of custom or simply non-Nvidia graphic processing units GPUs . “We expect Ethernet to become the leading scale-up technology by the end of the decade and to remain the long-term leader across the broader AI back-end networking market, including scale-up, scale-out, and scale-across,” Sameh Boujelbene, VP at Dell’Oro Group, noted. A Dell’Oro Group forecast published in February said that spending on switches for AI back-end networks alone would surpass $100 billion by 2030 https://www.sdxcentral.com/news/ai-back-end-networks-to-drive-data-center-switch-spending-past-100b-by-2030/ . Despite Boujelbene and Dell’Oro Group’s anticipation for continued Ethernet dominance, the analyst firm said that multiple networking technologies and architectures will coexist, creating “multibillion-dollar opportunities” as operators balance their stack’s performance requirements and ecosystem preferences. An emerging networking technology set to shake up wider back-end buildout is the rise of more “advanced” pluggables and optics options like near-packaged optics NPO and co-packaged optics CPO , which the analyst firm said will gain early traction. CPO switches are expected to hit shelves this year in the form of CPO iterations of its Spectrum-X https://www.sdxcentral.com/analysis/inside-spectrum-x-nvidias-ethernet-networking-platform/ and Quantum-X https://www.sdxcentral.com/news/lambda-offers-glimpse-into-nvidias-cpo-switch-enabling-more-tokens-through-substantial-power-savings/ lines, and Dell’Oro Group analysts suggest the emerging segment will ramp over time as both the tech and supply chain mature. “The unprecedented pace of investment in accelerated computing to support the next wave of AI journey – driven by agentic and physical AI applications – is fueling a rapid expansion of the AI networking market, including scale-up, scale-out, and scale-across,” Boujelbene explained. “We have substantially raised our revenue and port-shipment forecasts since our January 2026 outlook to reflect both stronger demand for accelerated compute and higher infrastructure costs. Despite this significant upward revision, we believe the market will remain supply-constrained rather than demand-constrained over our five-year forecast horizon.”