25 May 2026 , 06:22 PM
When hyperscalers started signing billion-dollar supply contracts with Indian fibre makers, the market did not just reprice individual companies it reframed an entire sector. This is an expert look at the mechanics of that re-rating: what changed, what investors are paying for, and whether the story supports facts.
| Metric | Figure |
| Hyperscaler capex 2026 | $700B+ (62% jump from 2025) |
| Global data centre fibre demand growth | +76% YoY in 2025 (CRU) |
| Fibre price recovery | +70% from $3.70 to $6.30/fkm |
| India data centre capacity by 2030 | ~10 GW (from ~1.5 GW in 2025) |
Every market re-rating has a moment when a number becomes impossible to ignore. For India’s telecom infrastructure sector, that number was 36.
As STL’s CEO of Optical Networking, Rahul Puri, put it plainly: AI-focused data centres require approximately 36 times more fibre than traditional CPU-based racks, to handle the massive data volumes and high-speed connectivity required by GPU clusters. For a sector that had spent the better part of a decade drowning in Chinese-led oversupply and single-digit margins, that ratio was not just a statistic it was a complete revision of the demand curve.
The context that makes this number explosive: the world’s five largest technology companies Amazon, Alphabet, Meta, Microsoft, and Oracle have collectively committed $660–700 billion in capital expenditure for 2026 alone, nearly doubling the $388 billion spent in 2025. Amazon leads at a projected $200 billion. Alphabet follows at $175–185 billion. Meta at $115–135 billion. Microsoft at $110–120 billion. Approximately 75% of this aggregate capex is directed at AI infrastructure.
| Company | 2025 Capex | 2026 Projected | YoY Change |
| Amazon | $125B | $200B | +60% |
| Alphabet | $91B | $175–185B | +98% |
| Meta | $72B | $115–135B | +74% |
| Microsoft | $90B | $110–120B | +28% |
| Oracle | — | ~$50B | — |
| Combined | $388B | ~$660–700B | ~+77% |
Source: Company earnings calls Q1 2026, Epoch AI, Futurum Research. ~75% directed at AI infrastructure per CreditSights.
Epoch AI data shows that combined capital expenditure at these five companies has been growing at an average annual rate of 72% since Q2 2023 the moment AI spending shifted from gradual growth to a steep acceleration. By Q4 2025, these five companies were spending a combined $140 billion in a single quarter.
That money flows through layers. It buys GPUs and TPUs from Nvidia and Broadcom. It buys land and power infrastructure. It buys cooling systems. And critically for our story it buys the fibre optic cables that connect thousands of GPU servers to each other within a data centre, and data centres to the world. Data centre fibre demand grew roughly 76% year-on-year in 2025 and is projected to account for 30% of total global fibre demand by 2027. In 2024, that figure was below 5%. The velocity of this shift is the core of the investment case.
“The fibre era is over. The hyperscaler era is just beginning.” the unspoken thesis behind every order announcement in this sector since late 2024.
To understand why stock prices moved the way they did, one must first understand that AI did not simply increase demand for the same kind of fibre that telcos use. It created demand for a fundamentally different product category at higher density, higher specification, and higher margin.
| Application | Fibre count per cable | Purpose | Margin profile |
| Traditional telecom | 24 96 fibre | 5G backhaul, broadband | Low (commoditised) |
| Early cloud DC | 144 432 fibre | Server-to-switch links | Moderate |
| AI GPU clusters | 3,456 6,912 fibre | GPU-to-GPU interconnect | High (specialised) |
The 72-GPU nodes such as Nvidia’s Blackwell require 16 times more fibre than traditional cloud switch racks, according to Corning’s SVP John McGirr. STL’s own CEO puts the comparison at 36x for AI-optimised versus traditional CPU server racks. These are not incremental upsells they represent a complete rearchitecting of what a data centre’s cabling plant looks like.
Within a single AI training cluster, hundreds or thousands of GPU servers must communicate with each other at terabits per second with sub-millisecond latency. The back-end network fabric that enables this sometimes called the GPU interconnect fabric requires ultra-high-density cabling, often pre-terminated with specialised multi-fibre connectors, routed through overhead cable trays with precision. A single large-scale GPU cluster can require hundreds of thousands of fibre strands just to interconnect its nodes.
This shift created a bifurcated market. Telecom-grade fibre (G.652D) remained commoditised. AI-grade fibre (G.657 bend-insensitive, ultra-low-loss) became scarce and high-value. The producers who could supply the latter at scale found themselves holding pricing power for the first time in years.
Optical cable usage for AI applications grew roughly 138% in 2024 and was expected to increase by a further 77% in 2025. When compared with data centre demand for non-AI applications, this is staggeringly higher.
Demand alone does not create a re-rating. What transforms a demand story into a structural repricing of an entire sector is the collision of that demand with a supply wall and in optical fibre, that wall is made of glass.
Every optical fibre cable starts with a glass preform a rod of ultra-pure silica that is drawn into thin fibre at temperatures exceeding 2,000°C. Preforms account for approximately 70% of fibre cost and represent the most technically demanding part of the supply chain. Crucially, new preform capacity takes 18 to 24 months to build. That is not a logistical constraint it is a physical one. No amount of capital can accelerate the timeline meaningfully.
| Indicator | Status / Figure | Implication |
| Global fibre price recovery | $3.70 → $6.30/fkm (+70%) | Margin expansion for producers |
| Delivery lead times | Up to 60 weeks (vs 15–20 week norm) | Sellers’ market; volume secured upfront |
| Chinese manufacturer order book | Booked into early 2027 | Opens door for Indian suppliers |
| New preform capacity timeline | 18–24 months to build | Supply cannot respond quickly to demand |
| Annual supply-demand gap estimate | ~100M core-km | Underpins multi-year contract structures |
| Meta–Corning deal | $6 billion (January 2026) | Validates hyperscaler lock-in logic |
Sources: Tom’s Hardware, CRU, DigiTimes, Guotai Junan Securities, industry filings.
The implication for Indian producers is significant. With Chinese manufacturers running at full capacity and facing anti-dumping duties in the US and Europe, and with Western producers like Corning locked into exclusive arrangements with their largest customers, Indian companies found themselves with a rare window the opportunity to sell into a supply-constrained market at improving prices, backed by multi-year contracts that eliminated the pricing volatility they had suffered for a decade.
HFCL’s ₹580 crore investment in a glass preform facility is not just backward integration it is an insurance policy against this very bottleneck. A company that can produce its own preforms is insulated from the supply crunch that is currently forcing smaller competitors to quote long lead times or lose orders entirely.
One manufacturer had already sold all of its fibre inventory through the year 2026. Lead times for fibre optic cables have reached up to sixty weeks the longest since the early 2000s build-out cycle.
In capital-goods businesses, order books are not just revenue pipelines they are statements of competitive positioning. Each order carries information about which customers trust which suppliers, at what scale, and for how long. The orders this sector has accumulated since late 2024 are worth reading carefully.
Multi-year supply of optical connectivity products to a global hyperscaler. FY27–FY29 allocation. Product: ultra-high-density and IBR cable series (up to 6,912 fibres). This order does not represent a commodity cable sale it is a technology supply agreement for AI-grade product. The PAL structure signals the hyperscaler is treating STL as a strategic partner, not a spot vendor. The total potential value confirmed at approximately USD 1.11 billion based on prevailing selling prices. STL’s consolidated order book now stands at over ₹17,000 crore.
What it means: The PAL (Product Award Letter) is a pre-contract designation that hyperscalers use when they have selected a preferred supplier and are finalising volume allocation. Receiving a PAL is not the same as a purchase order it is a stronger signal, because it means the hyperscaler has evaluated competing suppliers and committed directionally. The market read this correctly and repriced STL accordingly.
Long-term supply agreement for Optical Fiber Cables, signed March 2026. The long-term structure is the critical signal here: hyperscalers are not buying on spot. They are locking in supply years in advance because they understand the preform bottleneck. A long-term deal transfers price and availability risk to the supplier but also guarantees revenue visibility that the market had not previously assigned to HFCL.
What it means: Revenue visibility is the single most powerful driver of valuation multiple expansion in capital-goods companies. When a company moves from a business where 80% of next year’s revenue is uncertain to one where a multi-year contract covers a significant portion of capacity, the market applies a higher multiple to that revenue. That is the mechanical explanation for HFCL’s re-rating.
HTL Limited secured purchase orders in April 2026, followed by export orders in May 2026. The subsidiary-level order flow confirms that the thesis is not concentrated in one contract it is distributed across HFCL’s manufacturing footprint, validating breadth of operations.
What it means: When a parent company’s headline contract is followed by subsidiary-level orders of meaningful size, it signals that the business is operating at scale across multiple customer relationships not winning a single large contract and otherwise running at low utilisation.
Order book of approximately ₹5,226 crore across Cable and EPC segments as of March 2026. Extended planned investment in speciality optical fibre cables to ₹101 crore for data centre supply.
What it means: The “speciality” designation matters. Vindhya is not simply expanding standard cable production it is targeting the data-centre-specific product category, signalling intentional positioning in the higher-margin AI supply chain. The EPC segment adds another dimension: as India’s domestic data centre campuses are constructed, the engineering, procurement, and construction contracts that deliver the fibre infrastructure within those facilities become a distinct revenue stream.
256,000+ towers across all 22 telecom circles. 5G BTS base reached approximately 531,000 as of Q4 FY26. Revenue ₹8,101 crore; net profit ₹1,792 crore in Q4 FY26. ICRA upgraded the company to AAA (Stable) in November 2025.
What it means: Indus Towers is a different kind of AI trade. The tower-as-infrastructure thesis for AI is indirect but structurally sound. As AI inference migrates to the network edge serving real-time queries from users on 5G networks the physical tower site becomes compute-adjacent infrastructure. Operators running AI services on 5G networks will require edge nodes co-located at or near tower sites. This is an early-stage but growing revenue stream that the market has begun to price as an optionality premium on top of the core tower business.
Stock returns are not random noise they encode market beliefs at specific points in time. The divergence in 1-year, 6-month, and 1-month returns across this sector reveals which companies were recognised first, which are being discovered now, and what the market is still debating.
| Company | LTP | Today | 1 MM | 6 M | 1 Year | Re-rating signal |
| Sterlite Tech (STL) | ₹463.20 | +5.00% | +57.66% | +310.03% | +544.58% | Early mover; PAL order catalyst |
| HFCL | ₹163.03 | +10.00% (UC) | +50.39% | +108.28% | +75.40% | Multi-order momentum; still re-rating |
| Valiant Communications | ~₹1,100+ | +30.16% | +30.16% | +62.50% | +166.56% | Niche; profit turnaround catalyst |
| Indus Towers | ₹439.90 | +1.82% | — | — | ~+40% | Steady compounder; AAA upgrade + Vi resolution |
| Vindhya Telelinks | ₹1,795.00 | +13.01% | +17.94% | +3.95% | +6.62% | Late mover; capital rotation now |
The return sequence matters enormously. STL’s +544% 1-year return includes the period before the hyperscaler order the market began pricing in the possibility of a large win before it was formally announced. This is the market’s forward-looking mechanism at work. The 6-month return of +310% corresponds almost precisely to the order announcement and subsequent analyst upgrades. The 1-month return of +57% suggests continued institutional buying after the fact late entrants still absorbing the news.
HFCL’s pattern tells a different story: a more modest 1-year return of +75.4% that dramatically accelerates to +108% over 6 months and +50% in a single month, including today’s upper circuit. This is a stock that the market was slower to believe, but has now started pricing with conviction. The sequence of order announcements the $1.10B long-term agreement, then the subsidiary orders, then the export orders has functioned as a drip feed of validation that has progressively narrowed the gap between management guidance and market belief.
HFCL hit upper circuit today for a specific reason. The market is reacting not just to the cumulative order book it is recognising that HFCL has simultaneously secured AI supply chain revenue, initiated backward integration into glass preforms, and opened a defence manufacturing vertical. Three separate re-rating triggers compressing into a single price move.
Vindhya’s +6.62% 1-year return contrasted with today’s +13.01% move illustrates sector rotation in real time. When a theme matures and the large-cap beneficiaries are already well-discovered, capital rotates to names with similar sector exposure but lower valuations. Vindhya offers direct exposure to the same order dynamics speciality OFC for data centres, a growing EPC book but has not yet been fully priced for the AI premium. That mismatch, not any specific news catalyst, is what is driving today’s move.
Valiant Communications‘ +166.56% 1-year return is a micro-cap re-rating story that runs slightly differently. The company’s profit turnaround from a net loss of ₹0.53 crore in Q3 FY25 to a net profit of ₹6.05 crore in Q3 FY26, with revenues rising 165% was the catalyst. But the nature of what Valiant makes is the structural underpinning: quantum-safe networking equipment, EMP-protected data storage, and Data Diode cybersecurity devices serve the exact intersection of AI infrastructure and defence where margins are highest and competition is lowest.
Before the AI inflection, this sector was valued on earnings multiples and those multiples were low, because earnings were thin, predictable, and commoditised. A typical OFC manufacturer trading at 10–15x earnings was priced as a commodity manufacturer, because that is precisely what it was.
The re-rating has changed the valuation framework entirely. Investors are now asking a different set of questions:
| Old question | New question |
| What is the earnings multiple? | What is the order book as a multiple of annual revenue? |
| What is the EBITDA margin? | What % of revenue is locked in via multi-year contracts? |
| What is the competition from China? | Is the company protected by anti-dumping duties or tech specification? |
| What is the capex plan? | Does the company have preform capacity? (vertical integration) |
| Who are the telecom customers? | Which hyperscalers are anchor customers, and what are the terms? |
This framework shift explains returns far better than any macro factor. STL’s re-rating was not driven by GDP growth or interest rates it was driven by a single contract that answered every one of the new framework questions affirmatively: a hyperscaler anchor customer, multi-year structure, AI-specific product specifications, and a 50M fkm manufacturing capacity providing credibility.
The concept of switching costs normally associated with software businesses has entered the hardware supply chain in a meaningful way. A hyperscaler building a data centre campus over two to three years cannot easily change its fibre supplier mid-construction. Connector specifications, cable tray designs, and pre-terminated assemblies are all customised for a specific product. This creates a form of lock-in that gives Indian manufacturers a revenue certainty premium they have never enjoyed before.
The market is not just paying for current earnings. It is paying for the probability that the next three years of order books will be filled by hyperscalers who cannot afford to switch suppliers mid-build.
The hyperscaler export order is one engine. India’s own data centre buildout is the second and it is only beginning to fire.
India’s installed data centre capacity stood at approximately 1,300 MW at end-2025. Deloitte and NITI Aayog jointly estimate that this will scale to approximately 10 GW by 2030 a seven-fold expansion in five years. At facility build costs of $5.5–8.0 million per MW (excluding IT hardware), that implies $55–80 billion in construction-phase spending alone. Every megawatt of that capacity needs fibre, towers, and telecom interconnection.
| Metric | 2024 | 2025 | 2026E | 2030E |
| Installed DC capacity | ~1,000 MW | ~1,300 MW | ~1,700 MW | ~10,000 MW |
| AI-optimised DC revenue | $589M | — | — | $3.55B (CAGR 35.1%) |
| Total sector revenue | — | — | — | $45.7B (by 2033) |
| Hyperscaler India investment | $30B+ announced since 2023 (Microsoft, Google, Amazon, Meta) |
Sources: KPMG India DC Report 2026, Deloitte TMT Predictions 2026, JLL, Wright Research.
The DPDP Act (India’s Digital Personal Data Protection Act) functions as a structural accelerant. Restrictions on cross-border data transfers are pushing global enterprises to establish physical data residency in India meaning they cannot simply run Indian operations from Singapore or Frankfurt. Every such enterprise becomes a buyer of domestic data centre capacity, and consequently a source of demand for domestic fibre and tower infrastructure.
The geographic expansion pattern also matters for towers. Mumbai accounts for roughly 50–55% of current data centre capacity due to submarine cable landings and existing power infrastructure. But land constraints and power costs are pushing new capacity toward Hyderabad, Pune, Chennai, and Tier-2 cities. This geographic spread is precisely the pattern that benefits Indus Towers with a presence in all 22 telecom circles, the company is structurally positioned to provide backhaul and edge infrastructure regardless of where the next data centre campus lands.
Customer concentration. Both STL and HFCL’s billion-dollar orders appear to flow from single hyperscaler anchor clients. Non-renewal, volume renegotiation, or a pause in AI capex from that counterparty could sharply reset earnings expectations. Multi-year structures provide some protection; they are not ironclad guarantees.
Valuation vs execution. STL at +544% over one year has compressed the margin of safety for new investors. The order book validates the revenue story; execution risk delivery timelines, working capital intensity, quality specifications remains. A single quarter of weak execution can trigger significant impact in a re-rated stock.
Technology risk. Corning and Sumitomo hold proprietary technology in hollow-core fibre and ultra-low-loss ribbon cable. If hyperscalers migrate toward next-generation fibre types that Indian manufacturers cannot yet produce, the competitive positioning of domestic players could erode faster than current contracts expire.
AI capex cyclicality. The sector was commoditised as recently as 2023. If global AI capex pauses due to model efficiency improvements reducing compute needs, regulatory intervention, or a hyperscaler earnings reset the demand driver disappears rapidly. The structural re-rating of this sector is durable only if AI infrastructure spending sustains its current trajectory.
Raw material exposure. Companies without in-house preform capacity remain exposed to the global glass shortage. The industry-wide supply-demand gap of approximately 100 million core-km annually means input costs can spike unpredictably. Only HFCL, with its ₹580 crore preform investment, is actively building structural protection here.
Late-cycle entry risk. Vindhya’s recent surge and HFCL’s upper circuit today may attract momentum capital chasing the sector theme. Retail investors entering at these levels in names where the institutional re-rating has already occurred bear asymmetric risk: the upside requires continued order flow; the downside is a reversion toward pre-AI multiples.
The structural case is not in question. The question for investors entering now is simpler and harder: how much of a multi-year thesis are you paying for today?
The optical fibre and telecom infrastructure sector did not simply grow it changed what kind of business it is. Companies that sold commodity cables at thin margins to price-sensitive telcos are now signing multi-year, billion-dollar supply contracts with the world’s most capitalised technology companies. That is not a cyclical upturn. It is a category migration.
The stock prices reflect this, imperfectly and with varying lags depending on which company investors focused on first. STL moved early because the order announcement was unambiguous. HFCL is still moving because its order cadence has been additive each successive announcement removes a layer of investor scepticism. Vindhya is moving today because capital eventually finds the last repriced asset in a re-rated sector.
What ties all of it together is the hyperscaler capex machine $700 billion in annual spending that must find its way through GPUs, data centres, power infrastructure, cooling systems, and at the very end of that chain, the fibre that connects all of it. Indian companies have positioned themselves to capture a meaningful slice of that last mile. The market has priced the first act of that story. The second act execution, margin expansion, and the India domestic buildout is what will determine whether these returns compound or consolidate.
This piece is for informational and educational purposes only and does not constitute investment advice or a recommendation to buy or sell any security. All data sourced from publicly available company filings, KPMG India Data Centre Report 2026, Deloitte TMT Predictions 2026, CRU, Epoch AI, IEEE ComSoc Technology Blog, Tom’s Hardware, Mckingsey and other cited publications. Past stock returns do not guarantee future performance. Investors should conduct their own research or consult a financial advisor before making any investment decisions. Investments in securities market are subject to market risks. Read all the related documents carefully before investing. Market prices as of May 25, 2026.
Related Tags
IIFL Customer Care Number
(Gold/NCD/NBFC/Insurance/NPS)
1860-267-3000 / 7039-050-000
IIFL Capital Services Support WhatsApp Number
+91 9892691696
Download The App Now
Follow us on
2026, IIFL Capital Services Ltd. All Rights Reserved
IIFL Capital Services Limited - Stock Broker SEBI Regn. No: INZ000164132 (Member ID - NSE: 10975 BSE: 179 MCX: 55995 NCDEX: 01249), DP SEBI Reg. No. IN-DP-185-2016, PMS SEBI Regn. No: INP000002213, IA SEBI Regn. No: INA000000623, Merchant Banker SEBI Regn. No. INM000010940, RA SEBI Regn. No: INH000000248, BSE Enlistment Number (RA): 5016, AMFI-Registered Mutual Fund Distributor & SIF Distributor
ARN NO : 47791 (Date of initial registration – 17/02/2007; Current validity of ARN – 08/02/2027), PFRDA Reg. No. PoP 20092018, IRDAI Corporate Agent (Composite) : CA1099
This Certificate Demonstrates That IIFL As An Organization Has Defined And Put In Place Best-Practice Information Security Processes.
