Sterlite Technologies’ record order book reveals the fibre boom underneath AI infrastructure

Artificial intelligence infrastructure is generally discussed through three scarce resources: advanced semiconductors, data-centre space and electricity. A fourth requirement sits physically between the computing equipment and is receiving far less public attention: the increasingly dense optical-fibre infrastructure required to move enormous quantities of data among GPUs, switches, racks, buildings and geographically separated data centres.

GPU clusters depend on extremely high-speed communication between thousands of processors. As clusters grow, the network connecting those processors becomes more complex, fibre counts increase and the physical cabling architecture must fit enormous connectivity into limited rack and pathway space. The result is an infrastructure requirement that expands alongside the artificial-intelligence computing cycle even though it attracts a fraction of the attention directed toward processors and power generation.

Sterlite Technologies Limited (NSE: STLTECH) has become an unusually visible beneficiary of that trend. The company reported a record ₹18,618 crore open order book at the end of Q1 FY27, compared with ₹7,309 crore at the end of FY26. Quarterly revenue reached ₹1,910 crore and EBITDA increased to ₹397 crore, with management attributing part of the margin improvement to a larger contribution from data-centre business.

A single hyperscaler award announced in May was valued at more than $1 billion and covers optical-connectivity products for artificial-intelligence data-centre buildouts. Sterlite Technologies has separately announced a ₹960 crore multi-year domestic optical-fibre cable order and continued expanding its U.S. data-centre portfolio.

Those numbers suggest that artificial intelligence’s physical infrastructure chain extends much further than chip manufacturers. The companies building the connectivity fabric around processors may become increasingly important as clusters scale and data-centre operators struggle to fit more bandwidth into increasingly dense physical environments.

Why do AI data centres need more fibre than conventional facilities?

Traditional enterprise computing often concentrated relatively modest processing requirements within individual servers. Artificial-intelligence training workloads distribute enormous calculations across large clusters of specialised accelerators that must continuously exchange data, making networking performance directly relevant to overall computing efficiency.

That communication requirement increases network bandwidth and the number of high-speed connections running between racks and switches. As GPU density rises, copper connections become less practical over many distances because of bandwidth, power and signal-loss constraints, increasing the role of optical technology.

Sterlite Technologies has told investors that GPU-dense racks can require dramatically more fibre than conventional CPU-based configurations and that fibre density inside artificial-intelligence data centres can be materially higher than in traditional facilities. Those estimates originate from the company and should therefore be understood as management’s market framing rather than independent industry forecasts, but the underlying physical direction is clear.

The engineering challenge is not merely installing more cables. Operators need compact, low-loss and rapidly deployable fibre assemblies that can be managed inside increasingly dense facilities without creating unacceptable complexity.

That is why products such as pre-terminated assemblies and high-fibre-count ribbon cables are becoming strategically relevant. Greater computing density increases the value of connectivity systems that can deliver more bandwidth without consuming a proportionate increase in physical space.

What is different about fibre inside the data centre and fibre connecting data centres?

Artificial-intelligence infrastructure uses fibre at several layers. Inside the facility, fibre connects switches, racks and computing clusters, while extremely dense environments demand compact assemblies capable of carrying large numbers of fibres through constrained overhead or underfloor pathways.

Between buildings or campuses, data-centre interconnect networks carry enormous quantities of traffic between facilities. Hyperscalers may distribute computing, storage and backup capacity across multiple structures, making high-capacity external fibre essential.

Longer-distance networks connect data centres with metropolitan, national and international backbones, while telecom operators and broadband networks ultimately connect users and enterprises to cloud services. The AI fibre opportunity therefore overlaps with a much older connectivity cycle rather than existing as an isolated market.

India is simultaneously expanding rural broadband through BharatNet, which is designed to provide optical-fibre connectivity across more than 600,000 villages. A manufacturer serving data centres, telecom operators and national broadband projects can consequently benefit from several infrastructure cycles at once.

Why is Sterlite Technologies moving closer to finished connectivity systems?

Commodity fibre manufacturing can be cyclical. Prices and margins respond to industry capacity, telecom investment and inventory cycles, making undifferentiated volume less attractive than specialised connectivity products.

Sterlite Technologies has been moving further toward the data-centre endpoint through products such as its Neuralis portfolio, high-density pre-terminated systems and hollow-core fibre development. The strategic logic is to capture more value from the engineered system rather than depending exclusively on kilometres of fibre sold.

In March 2026, the company announced what it described as India’s first hollow-core fibre cable for data-centre applications. Unlike conventional fibre, which transmits light primarily through glass, hollow-core designs guide much of the signal through air, potentially reducing latency. Sterlite Technologies said its implementation can enable signals to travel around 46% faster than through conventional solid-glass fibre.

Commercial adoption of hollow-core technology remains an evolving market rather than a guaranteed mass deployment, but the development illustrates how optical manufacturers are trying to differentiate through performance rather than pure capacity. The company has also launched plenum-rated fibre trunk assemblies for the U.S. market after achieving relevant fire-safety compliance for portions of its high-density ribbon portfolio.

The strategic objective is clear: move from supplying kilometres of fibre toward supplying more of the engineered connectivity system surrounding AI computing. If successful, that shift could produce a business with greater technological differentiation and potentially better economics than commodity-oriented optical manufacturing.

Why is U.S. manufacturing becoming more important to optical suppliers?

Sterlite Technologies announced plans in May to invest up to $100 million in the United States to expand manufacturing capacity for data-centre and telecom connectivity, with the company expecting the investment to create approximately 400 to 500 jobs.

Local production matters for several reasons. Hyperscale customers want resilient supply chains and shorter delivery times, while U.S. industrial policy and procurement considerations increasingly reward domestic manufacturing.

The shift also reduces the risk of treating AI infrastructure as an export-only opportunity from India. A global manufacturer may need production footprints close to large customers if data-centre construction continues at the current pace.

For Sterlite Technologies, the U.S. investment complements manufacturing operations across several regions and could help convert large hyperscaler agreements into a more durable customer base. It also raises the stakes because manufacturing expansion requires capital, customer concentration can create negotiating power for hyperscalers, and rapid technology changes can make specialised products obsolete faster than conventional telecom cables.

Could fibre become an AI bottleneck in the same way electricity has?

Fibre is unlikely to become a constraint in exactly the same way as electricity because optical cable can be manufactured and deployed more flexibly than new power generation, substations or transmission infrastructure. Data centres cannot operate without adequate grid capacity, while fibre supply can generally respond more rapidly if manufacturing capacity, materials and installation labour are available.

The more relevant constraint may be physical and operational complexity. As fibre counts increase, the difficulty of packaging, installing, testing and maintaining thousands of optical connections rises, while space inside ducts, racks and cable pathways becomes increasingly valuable.

This creates demand for higher-density cables, factory-terminated assemblies and architectures designed specifically for GPU clusters. Operators have an economic incentive to reduce installation time, connection errors and physical congestion as each generation of computing infrastructure concentrates more bandwidth inside the same footprint.

In that sense, the AI opportunity is less about a global shortage of glass and more about a shortage of efficient ways to connect extraordinarily dense computing environments. Suppliers that solve those engineering problems may capture more value than companies competing primarily on undifferentiated fibre volume.

What does Sterlite Technologies’ share-price surge tell investors?

Market enthusiasm has been extraordinary. Sterlite Technologies traded as high as ₹865.90 on September 9 after a sharp 2026 rally, and the stock closed around ₹894.70 on September 11 following another strong session. Market reporting has described the year-to-date gain as several hundred percent, reflecting intense investor optimism around its hyperscaler orders, improving profitability and AI-data-centre positioning.

That strength creates a different problem from the one Sterlite Technologies faced when optical-fibre markets were weak. Expectations are now high, meaning investors increasingly need evidence that the expansion in orders will translate into sustained revenue growth, margins and free cash flow rather than simply larger headline contract values.

Revenue must convert from the expanded order book, margins must remain resilient, hyperscaler concentration must be managed and capital investment must generate appropriate returns. A $1 billion award is commercially significant, but the ultimate valuation depends on how much profitable cash flow the contract produces over its life.

Investors should also distinguish between the structural AI connectivity opportunity and short-term market enthusiasm around anything carrying an artificial-intelligence label. The underlying infrastructure logic is nevertheless compelling because every new GPU cluster creates network connections, every campus-scale data centre needs links between buildings, and every cloud region requires long-haul connectivity to customers and other facilities.

Artificial intelligence may therefore create an unusual chain of beneficiaries stretching from semiconductor fabrication to electricity generation, cooling, construction and optical networking. GPUs perform the calculations, but the processors become far less useful if data cannot move among them quickly and efficiently enough, making optical fibre one of the least glamorous but potentially most indispensable physical layers of the AI investment cycle.