Open in interactive viewer → charts, metric popovers & call review

Tata Communications Builds an AI Stack — and Quietly Clears a ₹1,658 Crore Tax Cloud

Record EBITDA growth and a near-breakeven digital portfolio look like a real inflection — until you strip out the currency tailwind.
TATACOMM.NS · Earnings Call · 2026-10-15

Profitable growth, delivered — with an asterisk

For three consecutive calls, CEO Ganesh Lakshminarayanan and CFO Siddhartha Mundra have told investors the same thing: forget topline vanity, the only goal is profitable growth. This quarter, Q2 FY2027, they finally have the receipts. Consolidated revenue came in at INR 6,688 crore, up 9.6% year-over-year; data revenue at INR 5,810 crore, up 12.2%; and EBITDA at INR 1,422 crore, up 21.2% — which Ganesh calls “the highest year-on-year EBITDA growth ever achieved and since the recent M&As.” — Ganesh Lakshminarayanan, CEO and MD · 2026-10-15 PAT after minority interest jumped 86% to INR 341 crore. The centerpiece is the digital portfolio. It reported an EBITDA margin of negative 3.6%, versus negative 6.9% in Q1, with losses “less than half of what was in Q1.” — Ganesh Lakshminarayanan, CEO and MD · 2026-10-15 Management says it is "very close" to breakeven on the digital book. Analyst Balaji Subramanian flagged exactly this on the call — after a long time, strong digital revenue growth accompanied by real margin expansion. But the CFO's own qualifier is the part worth holding onto. The 12.2% data-revenue growth is reported, not real: “Normalizing for the same, data revenues saw a growth of about 6% on a year-over-year basis.” — Siddhartha Mundra, CFO · 2026-10-15 Roughly half the headline is a strengthening dollar. Underlying connectivity growth is mid-single-digit — the same range that has kept this a slow compounder rather than a fast one.

The unified AI stack: the genuinely new thing

What changed this quarter is the narrative architecture. Ganesh — six months into the job, with a transformation chief now in place — laid out a three-layer "unified AI stack": a sovereign hardware foundation to run models securely; an enterprise context layer that turns enterprise data into intelligence (built on the Commotion acquisition); and an action layer of AI agents that execute across voice, messaging and digital channels. Two marquee logos were trotted out: Tata Consumer, where an AI field-sales assistant called SalesSathi handles multilingual voice; and HDFC ERGO, where voice AI agents now run policy-servicing journeys in multiple Indian languages. The framing is deliberate.

If data is not transformed to intelligence, there is only hallucination. All that intelligence, if it doesn't transform into actions, it's like having the smartest and highest-paid employee in your company with no real output.

Ganesh Lakshminarayanan, CEO and MD · 2026-10-15
This is where the keywords turn company-unique. The AI stack and the ontology layer — a customer-owned context layer that Tata calls its real moat — are not sector boilerplate. Nor is the pricing model: a SaaS-like platform fee plus consumption, spread across network minutes, GPU usage and software. The Commotion capability is being sold as the connective tissue between enterprise knowledge and execution.

Fair point. I think you should think of ourselves as putting these multiple fabrics together to stitch a nice suit... the fabric makers will get their portion. But the suit is what the customer wears.

Ganesh Lakshminarayanan, CEO and MD · 2026-10-15

Where the tape disagrees with the story

Here is the uncomfortable part. Tata Communications is betting its next leg of growth on AI-driven data center buildout — DC-to-DC connectivity, subsea, metro fiber, GPU clusters. It cites a projection of 70 petabytes of traffic demand by 2036 versus two or three today, and a multi-terabyte win with a global neocloud provider for GPU-cluster interconnection in India. Its Order book has now grown double-digit for what management calls 2.5 years. But the market's AI data center complex is under pressure. Over the last 90 and 30 days, the broad AI data centers theme is one of the tape's worst decliners — roughly -6% with the vast majority of constituent tickers negative. Related clusters — high bandwidth memory, PCIe Gen 6, co-packaged optics, high-performance compute — show the same negative lurch. So the company is leaning harder into a theme the tape is currently voting against. That divergence is either early-cycle opportunity or wishful thinking; order-book conversion will be the arbiter. The other shared theme — data center buildout as a capital cycle — shows up across today's reporters too, from PLD to FAST to POLYCAB. That is consensus, not differentiation. Tata's edge has to come from the stack layer, not the real estate.

The buried headline: a ₹1,658 crore tax cloud

Halfway through the CFO's prepared remarks, almost as an aside, comes the item that may matter more to intrinsic value than any AI slide: an in-principle agreement with the Canada Revenue Agency on a longstanding tax matter, against a contingent liability that stood at INR 1,658 crore as of March 31, 2026. Terms are still being finalized, and the market got almost no detail in Q&A. Clearing an overhang of that size — on a company earning INR 341 crore a quarter — is a balance-sheet event masquerading as a footnote.

The margins, and what management is careful to say

Backing out the disclosed digital margins, the implied core connectivity margin comes to roughly 44.8%. Analyst Vibhor Singhal asked directly whether that is sustainable; Sid's answer was a flat yes, with “no one-offs or there are no such events that are impacting our Q2 performance in any way.” — Siddhartha Mundra, CFO · 2026-10-15 That is a strong statement — the high-EBITDA core connectivity business is carrying the AI capex build. The NR to GR ratio — net revenue over gross revenue, i.e. the quality of the sales mix — is the mechanism management credits for the digital turn. But the notes of caution are now carefully hedged. On whether the long-written-off voice business revives on the back of AI agents, Ganesh demurred: “it is too early to say... voice is becoming the execution layer for AI agents more and more.” — Ganesh Lakshminarayanan, CEO and MD · 2026-10-15 The media business remains cyclical and softer year-over-year, with management leaning on an Olympics next year as the next catalyst. Net debt ended at INR 11,197 crore, 2.17x EBITDA; free cash flow swung to positive INR 212 crore; ROCE ticked to 15.9%. All tidy. The bull case now rests on one proposition: that the stack conversion turns a double-digit profitable growth order book into double-digit constant-currency revenue — the number that has not yet materialized.