TD Power Systems — the generator maker the AI boom found

The State of Play

TD Power Systems makes the generators that bolt onto other people’s turbines and engines, and in FY26 the people who make those turbines and engines — for US data centers, mostly — could not get enough of them. Consolidated income rose 44% to ₹1,878 crore, order inflows jumped 51% to ₹2,238 crore, and the company beat not just its original FY26 guidance of ₹1,500 crore but the raised mid-year guidance of ₹1,800 crore. Management now guides ₹2,400 crore-plus for FY27, sees ₹3,000–3,200 crore of capacity through FY28, and has announced its biggest strategic bet yet: moving up from 50 MW-class machines into large generators of up to 200 MW.

The Company

TDPS builds custom AC generators — steam- and gas-turbine generators up to 250 MVA, hydro up to 45 MVA, plus engine-driven gensets, traction machines and an emerging motors line — every machine engineered to order. Its customers are the handful of prime-mover OEMs that dominate the world’s gas engine and turbine market (“we work with all the big ones except for one”), and its pitch is simple arithmetic: comparable quality to European and Japanese rivals at roughly three-quarters of the cost, from three plants in Tumkur, Karnataka, with a Turkish subsidiary as a tariff-era pressure valve and a new UK design centre developing its own large-machine designs.

The long arc is a genuine transformation: TDPS lost money on 2–3% margins through FY15–FY18, then re-rated operationally — revenue tripled from ₹594 crore (FY21) to ₹1,856 crore (FY26), operating margins settled near 18%, ROCE hit 33%, and the balance sheet is essentially debt-free (₹18 crore of borrowings against ₹1,041 crore of reserves). The one governance wrinkle: promoter holding has more than halved in three years, from 58.4% to 26.9%, in two big step-downs (September 2023 and June 2025), absorbed mostly by FIIs, who went from 2.6% to 26.7%. The MD said on the August 2025 call that no further selling is planned “for the next 24 months at least” — and the stake has held flat since June 2025. The stock trades at a P/E of 86 and 19x book.

The Story So Far

Q1 FY26 (call: August 7, 2025) — a tariff bomb, defused with a spreadsheet

The call happened one day after the US announced an additional 25% tariff on India, taking the headline rate toward 50%, and MD Nikhil Kumar spent most of it walking analysts through the arithmetic of why it didn’t matter much. The quarter itself was strong — standalone income up 36% to ₹363 crore, PAT up 51%, order inflows up 32% with a 66% export mix. The tariff logic, in his ballpark numbers: a TDPS machine at an index cost of 100 competes with a European one at 125; with tariffs, routing through the company’s Turkish subsidiary lands TDPS at 125–130 versus the European product at ~150 — “the gap gets reduced to some extent, but not to an extent where it jeopardizes our cost advantage.” And because a power plant is only 3–4% of an AI data-center’s total cost, customers — ultimately the Microsofts and Metas — would simply absorb it.

“Even if the forecasts are 80% correct, we expect to exceed the guidance made for this year and next year.” — Nikhil Kumar, on the ₹1,500 crore FY26 / ₹1,800 crore FY27 guidance

The growth engine was named plainly: gas turbine and gas engine generators for the US and Europe, driven by data centers, grid stabilisation, and a global power-equipment shortage. The third Tumkur plant (₹120 crore, per the FY25 annual report) would commission through Q2–Q3, lifting capacity toward ₹2,000 crore of revenue, stretchable to ₹2,400 crore.

Q2 FY26 (call: October 31, 2025) — “completely turbocharged”

The first promised guidance upgrade arrived on schedule. H1 consolidated income rose 42%; Q2 order inflow hit ₹524 crore, up 45%, with the half-year mix at 76% exports. FY26 guidance went from ₹1,500 to ₹1,800 crore; FY27’s “tentative” ₹1,800 became “minimum ₹2,000 crore,” indicated at ₹2,000–2,200 crore. The gas business was, in Kumar’s phrase, “completely turbocharged and demand has increased dramatically.” Asked about an AI bubble, he quoted the Wall Street Journal back at the analyst: big tech was raising 2026 spend, and OEM customers were already discussing FY27 and FY28 capacity.

The honest footnotes: gross profit dipped ~250 basis points on product mix (promised to normalise by Q4 — it did); employee costs rose 35% YoY as the third plant staffed up; and free cash flow was thin — “as long as we have this high growth, I don’t see us generating huge amounts of free cash flow” — with retained earnings going into inventory for the ramp. A 20%+ EBITDA margin was named as a goal, deliberately not a commitment. New seeds: a first-ever oil & gas order from ADNOC, railway trial units for the US and Europe, and the UK-designed 40–50 MW generator to be offered to a customer by January.

Q3 FY26 (call: January 30, 2026) — record inflows, firm forecasts to 2030

The December quarter printed an all-time-record order inflow of ₹656 crore, up 61%, with 84% of it exports; the pending order book for generators and motors had more than doubled in 24 months. Nine-month consolidated income rose 36%. The third plant was declared operational on December 18, slightly behind the original Q2–Q3 schedule but done. FY27 guidance moved up again, to “₹2,200-plus crores,” explicitly labelled conservative against an order run-rate heading for ₹575–600 crore a quarter.

The demand-durability question got its most concrete answer of the year:

“There is a clear demand profile… firm forecast from OEMs up to 2030… That’s a good visibility that we have right now, four years. More than that, I don’t think anyone can expect in the capital goods business.” — Nikhil Kumar, Q3 FY26 call

On the AI-slowdown narrative: “Absolutely not. Absolutely not… Nobody is cutting it down.” Copper inflation was being passed through via renegotiated prices (“we have absolutely no problem in getting the price increases”), cushioned by cheaper booked inventory, and a rupee/euro tailwind was due to hit the P&L from Q4. Peak revenue on the existing asset base was sized at ₹2,600–2,800 crore, with no bulk capex through FY28.

Q4 FY26 (call: May 15, 2026) — the beat, a brutal customer, and a 200 MW ambition

FY26 closed at ₹1,878 crore of consolidated income (+44%) and ₹236 crore of PAT (+36%) — past both versions of the year’s guidance. FY26 order inflow reached ₹2,238 crore (+51%), the book ₹1,973 crore, exports 79% of inflows. The quarter’s blemish was a one-off: a Turkish contract delayed when components from India got stuck in transit for six weeks drew a heavy liquidated-damages penalty from a customer management called “exceptionally brutal,” cutting the quarter’s gross margin by about 3 points. Guidance for FY27 moved up a third time, to ₹2,400 crore-plus with “an extremely high probability” of further raises; FY28 expectations were framed at ₹3,000–3,200 crore — “conservative.”

The strategic news was the large-generator program: heavy investment in rotor manufacturing and large machining capability for machines up to 200 MW — competing with Siemens, Mitsubishi, Toshiba — aimed at power generation, large AI data centers, combined-cycle plants, and potentially SMR nuclear. Machine tools take 15–16 months to arrive, so the ramp comes in calendar 2028; the detailed plan and capex number were promised within 2–3 months. Alongside: a signed capacity commitment with engine-maker INNIO running to 2030 (prime-mover OEMs are on average doubling capacity by then; INNIO plans to triple), a delivery to “projects such as SpaceX,” NPCIL induction-motor orders, and a newly hired CEO — Deepak, ex-L&T Mitsubishi — to professionalise the management layer. The new risk Kumar volunteered himself: factories now run so hot that an equipment breakdown is the main threat. “This is a new normal for us.”

The ledger: said vs. delivered

This is the cleanest scorecard of the season. Beat: FY26 revenue — guided ₹1,500 crore in the annual report, raised to ₹1,800 mid-year, delivered ₹1,878 consolidated; every “we will exceed” proved true. Kept: margins within the promised ±0.5% band (18.14% vs 17.46%, ex the Turkey one-off); the gross-margin recovery by Q4; copper pass-through; no order cancellations from tariffs, exactly as claimed in August; no promoter selling since the June 2025 step-down; third plant operational, one month late. Pending: the 20%+ EBITDA goal (18.1% achieved — still a goal); the large-generator plan details (due ~August 2026); the UK-designed 40–50 MW machine’s commercial orders. Worth noting: the repeated guidance raises were the opposite pattern of guidance-walking-down — each “conservative” number was beaten — though it bears remembering the whole engine runs on one structural assumption: that US/European data-center and grid demand for gas-fired generation holds through 2030, which management treats as firm OEM-contracted fact.

Where Things Stand

FY27 starts with ₹2,000 crore of the ₹1,973 crore order book executable in-year, a stated production ramp to ~₹600 crore a quarter, and guidance of ₹2,400 crore-plus that management has all but said it will raise. Margins are guided flat-to-better ex-Turkey; copper and a possible commodity spike are the visible cost risks, the rupee a tailwind. Capex stays modest (₹50 crore a year) until the large-generator investment lands — that plan, due in months, is the next structural thing to read, since it takes TDPS from supplying the data-center boom’s periphery to competing with Siemens-class incumbents at the 100–200 MW core. The valuation (P/E ~86) already assumes the boom is durable; management’s own four-year OEM forecasts agree with it. The Q1 FY27 print — promised to be “better than Q4” — is the first checkpoint.

The Four Checks

1. Quality and moat. A good business with a real but narrow moat. The advantage is qualification and cost: TDPS is designed into the supply chains of nearly every major prime-mover OEM (“all the big ones except for one”), each machine engineered to the customer’s spec, with comparable quality to European and Japanese rivals at roughly three-quarters of the cost — and management says there’s “no room for new entrants” because generator demand is set by engine and turbine capacity, where TDPS already holds signed commitments (INNIO to 2030). But the MD himself drew the limit: “it’s difficult for us to get better pricing than what we currently have… we are not going to make the same kind of money that the turbine guys make.” This is a well-positioned component supplier riding its customers’ boom, not a toll collector — and the whole engine rests on one assumption, that gas-fired data-center demand holds through 2030. Decent edge, contestable position, cyclical end-market.

2. Returns on incremental capital and runway. The returns are excellent and recently earned: ROCE of 33% and ROE of 25% as of the June 2026 snapshot, up from roughly 2% in FY15, with operating margins re-rated from 3% to a steady 18%. The reinvestment itself is mostly working capital — fixed capex runs a modest ₹50 crore a year against ₹1,856 crore of revenue, while inventory and receivables absorb the retained earnings (cash conversion cycle ~170–190 days) and convert at those 30%-plus returns. The runway is real but dated: firm OEM forecasts to 2030, capacity sized for ₹3,000–3,200 crore by FY28, and a large-generator program (up to 200 MW, ramping calendar 2028) that opens a Siemens-class market — at unproven returns. High rate, four visible years of road, and a single structural assumption underneath.

3. Capital allocation for the stage. Rational for a business compounding at 33% ROCE: nearly everything is reinvested. The dividend payout has fallen from ~16% to 7% as growth accelerated (yield 0.10%), borrowings sit at ₹18 crore against ₹1,041 crore of reserves, there is no buyback history visible in the data, and the spending has been pointed — ₹120 crore for the third Tumkur plant (delivered, one month late), a UK design centre that feeds the large-generator ambition, and a CEO hire to professionalise the layer below the MD. The blemishes are minor and mostly historical: the DFPS subsidiary investment is now 100% provided for, and free cash flow is thin by design (“as long as we have this high growth, I don’t see us generating huge amounts of free cash flow”). The promoter halving his stake from 58% to 27% in three years is a shareholder-register event, not a company-cash decision — but it is the one allocation signal pointing the other way.

4. Price. Demanding. As of the June 2026 snapshot the stock trades at ₹1,163 — a P/E of 76 and 17.1 times book — even after FY26 delivered 44% income growth and a 36% PAT increase. That multiple already capitalises the guidance ladder: ₹2,400 crore-plus in FY27, ₹3,000–3,200 crore in FY28, margins holding near 18%, and the large-generator bet paying off from FY29. Management has beaten every number it has set this cycle, which is exactly what the price assumes will continue; a custom-machinery business with admitted limited pricing power, earning ₹15 of EPS, is being valued as if the data-center gas boom is a permanent fact rather than a four-year forecast. Cheap it is not.

Sources

• Concall transcripts (4): Q1 FY26 (Aug 7, 2025), Q2 FY26 (Oct 31, 2025), Q3 FY26 (Jan 30, 2026), Q4 FY26 + full-year (May 15, 2026) — BSE filings, converted to markdown.
• Annual reports (3): FY23, FY24, FY25 sections — FY25’s was substantive (third-plant ₹120 crore investment, UK design centre, the original ₹1,500 crore FY26 guidance, segment narrative); FY23/FY24 thinner.
• Screener.in snapshot: consolidated quarterly and annual tables, ratios, shareholding — fetched 2026-06-05 (logged-out session).
• Research files: vault/Sources/Earnings/TD Power Systems Ltd/ — raw transcripts, AR sections, snapshot, per-document digests (not published).