Webinar: Freedom to Evolve: AI-Ready Optical Transport from DCI to Ultra-Long-Haul
Date & Time - 16 Oct 2025, 11am EST
Duration - 60 min
What You’ll Learn
AI is no longer experimental, it powers critical operations from autonomous driving to healthcare diagnostics. But while GPUs capture headlines, the real bottleneck is the network. AI workloads generate massive, deterministic, and latency-sensitive traffic flows that overwhelm legacy packet-switched architectures, which introduce jitter, congestion, and unpredictable latency.
In this session, PacketLight's Rena Erez and AlixPartners' Andrej Danis explain why the network, not the GPU, is today's real AI bottleneck, and how open DWDM and OTN backbones deliver the jitter-free, encrypted, high-capacity transport AI workloads need, from data center interconnect to ultra-long-haul.
Speaker
Rena Erez, Marketing Director
PacketLight Networks
Marketing Executive with a proven track record in driving growth across SaaS, telecom infrastructure, and AI. With leadership roles at NASDAQ-listed companies and innovative startups, Rena brings deep expertise in translating complex technology into business value. Currently at PacketLight, Rena helps organizations build AI-ready optical transport strategies that meet performance requirements of any type of organization and help them scale securely and efficiently.
Who will benefit from this webinar?
- Network architects & planners
- Engineering & operations teams
- Executives seeking ROI and scalability
Key Takeaways
- Why the network is the new AI bottleneck - Research from Meta and Stanford found GPUs sitting idle up to 56% of training time due to data transport limits, not compute.
- The 2026 data center demand picture - 98% of GPU-powered data centers are in growth mode, but 57% of operators have limited demand visibility beyond one year.
- The six biggest challenges in AI networks - bandwidth, jitter, cross-region latency, dynamic workloads, compliance, and visibility.
- How optical transport solves them - DWDM scaling to 51.2 Tbps per fiber pair, OTN's deterministic latency, ROADM-based flexibility, and Layer-1 encryption with full throughput.
- What "Freedom to Evolve" means - building networks that scale on demand without compromising budget, security, or performance.
FAQs
1. Is the webinar recording available?
Yes. Watch the full on-demand recording above, anytime.
2. What is "Freedom to Evolve" in optical networking?
PacketLight's approach to building networks that scale on demand, modular, standards-based DWDM and OTN infrastructure that grows with AI traffic without forklift upgrades or vendor lock-in.
3. Why is the network the bottleneck for AI workloads, not the GPU?
AI training needs GPUs perfectly synchronized. Legacy packet networks introduce jitter, queuing, and retransmission delays that stall GPUs and waste compute cycles. Meta and Stanford research found GPUs idle up to 56% of the time due to transport limits, not compute.
4. What are the six biggest challenges in AI networks?
Bandwidth growth, jitter and GPU stalls, cross-region latency, dynamic workload volatility, compliance and encryption, and lack of real-time visibility.
5. How do DWDM and OTN solve AI network bottlenecks?
DWDM scales fiber capacity up to 51.2 Tbps per pair. OTN adds deterministic latency and forward error correction for lossless delivery. Together they eliminate jitter and keep GPUs fully utilized, from DCI to ultra-long-haul.
6. Who should watch this webinar?
Network architects, engineering and operations teams, and executives evaluating how to scale AI infrastructure with predictable performance and ROI.
