October 27, Bangkok, At Telecoms World Asia 2026 held in Bangkok today, Prasad K S, Vice President of Pre-Sales at Tejas Networks, a subsidiary of India’s Tata Group, delivered a keynote speech titled Networks for AI: AI-ready transport and access infrastructure. As a major player in the optical communications sector, Tejas outlined a clear roadmap for the evolution of AI-driven optical networks in the presentation.
That said, when this technical presentation is viewed alongside Tejas’ recent capital market performance, a key question emerges: how can the enormous market potential of AI-enabled optical networks translate into tangible profits on the books of optical equipment vendors?
I. Reassessing the Core Logic: The Speed of Light Cannot Be Beat, but Its Waste Can Be Stopped
The core thesis of Prasad’s on-stage presentation can be summed up in one line: For AI, Latency is the new bandwidth.
Traditional optical networks are designed to pursue throughput expansion alone. In the context of large-model distributed training and compute clusters (DCI), network jitter and microsecond-scale latency directly erode the efficiency of costly GPU resources. Tejas put forward the view that “the speed of light cannot be surpassed, yet we can stop wasting light time”, and offered a full-stack technical solution:
Breakthroughs at the physical layer: Push for 1.6T single-wavelength rates, C+L wide-spectrum expansion, and adopt hollow-core fiber, CPO/LPO and other decoupling technologies to cut physical latency caused by optical-electrical-optical (O-E-O) conversions.
AI embedded across every layer (AI Inside Each Layer):
DWDM layer: Leverage machine learning to accurately estimate OSNR and implement Margin Harvesting. By trimming conservative safety margins over existing fiber infrastructure, the approach directly boosts usable bandwidth by 20%–40%.
IP/MPLS and PON layers: Use predictive traffic engineering (Predictive TE) to avoid congestion, and deliver intelligent fault diagnosis and experience tuning on the access side.
Progressive Agentic Operations: Build a network-wide digital twin. The initial phase adopts a read-only mode with human-approval automation loops to guarantee smooth migration for live networks.
II. From Technical Blueprint to Financial Statements: Three Gaps Holding Back AI Monetization
Tejas’ AI optical networking solution represents cutting-edge industry thinking. Still, placed against the commercial realities of optical equipment suppliers, three barriers stand in the way of capturing AI dividends:
Fragmented revenue streams between legacy and new businessesThe bulk of revenue and orders for optical vendors such as Tejas still stem from conventional telecom operator infrastructure projects, including BSNL’s 4G/5G and national broadband rollout across India. AI-focused 1.6T DWDM, hollow-core fiber and compute-network gear remain largely in proof-of-concept (PoC) and trial deployments with top-tier overseas carriers and hyperscalers. While AI technologies build long-term technical moats, they have not yet become the engine for near-term revenue growth.
Timing mismatch between heavy R&D spending and margin pressureAI optical networks impose stringent requirements on hardware architectures (130+ GBd symbol rates, open line systems) and software capabilities (network digital twins, LLM copilots). Vendors must sustain heavy investment in R&D and high-spec production capacity. Before high-end AI gear ships at scale, steep R&D costs squeeze gross margins and may even lead to periodic losses.
Valuation reset in capital marketsInvestors once priced in aggressive expectations for “AI compute + optical communications”, fuelling sharp rallies in Tejas and peer stocks. As markets recognise that commercial mass procurement of AI equipment lags years behind concept validation, capital has refocused on fundamentals. Tejas’ recent share price pullback reflects this adjustment to reconcile long-term AI expectations with near-term legacy financial results.
III. Conclusion: How Vendors Turn AI into Real Profits
As illustrated by Prasad’s presentation at Telecoms World Asia, AI is not merely bringing cost and efficiency gains to optical networks — it is reshaping business models. Equipment vendors must complete three transitions to unlock genuine profit generation:
Advanced hardware: Deploy 1.6T / DCI high-speed optical hardware to tap compute interconnect use cases and secure high-value orders.
Software SaaS monetization: License AI modules such as Margin Harvesting and Predictive Operations as standalone software to lift overall gross margins.
Global market expansion: Leverage Open Line Systems and AI operation assistants to break incumbent vendor lock-in and capture high-margin overseas and enterprise markets.
Tejas’ keynote paints an exciting vision for the future of optical networking. For the optical communications industry as a whole, however, the real test that decides winners in the AI wave lies in moving these sophisticated compute-era technologies beyond exhibition halls and trial phases, converting them into order backlogs and healthy gross margins.