# AI capex — X 热门讨论 (2026-10-11 06:21 UTC)
## @LearningEleven (Sekhar) · 10-11 03:32 · ♥270 ↻25 💬15 Timepass talk on Sunday
1. Unimech - Hyper growth aspirations with complications
Unimech is a Bengaluru-based integrated precision-engineering company. It manufactures aero tooling, ground-support equipment, precision components, engineered metallic bellows and complex assemblies for aerospace, defence, energy, semiconductor and industrial sectors.
As on June 30, 2026, the company operated five manufacturing facilities, equipped with over 155+ CNC machines that are operating at ~60% utilisation, providing headroom for growth without proportionate capex in the near term.
Differentiated niche: Unimech is a qualified-SKU compounder (6,300+ SKUs, 41 customers, 96% exports) in aero tooling and precision parts. It has captured only 25–40% of the ~1,500 tools per engine programme, so existing programmes still offer room to grow.
Multiple vectors beyond tooling: LEAP engine stands (~120 shipped), Lam Research (31+ parts qualified, ₹1–1.2bn by FY29), an ASML opportunity ($50–70mn over 3–4 years), FACC ($7.5mn) and nuclear (NPCIL) all add to the tooling base.
The ₹1000cr FY29 target (50%+ CAGR): Although nothing is guaranteed, company expects, tooling ₹450–500cr + precision ₹250–260cr + Hobel ₹150–180cr + Kanoo JV ₹80–100cr sums to ₹930–104cr. Hitting the low end of every bucket gives ~₹9.3bn, so the target needs near-full delivery on all four, including two unproven ones.
Hobel is a pricey bet: Hobel was bought for ₹450bn in cash, against FY29 revenue of ₹170–200bn (~2.5x sales). Its largest customer is ~70% of revenue, and AS9100 isn’t due until April 2027, so payback depends on customer diversification and data-centre genset demand.
Margin and working capital: EBITDA is guided down from ~36% to ~32%, which management attributes to R&D and a mix shift to precision and nuclear. Those same segments carry 220–400 working capital days versus ~100–110 for tooling. A blended 170–180 days by FY29 looks hard to reconcile with that mix.
Nuclear is the swing factor: The ₹87cr refurbishment order and ~4 new tenders in 4–6 months give a visible ramp (₹38–40cr in FY27, ~₹100cr by FY29). Management itself treats ₹100cr as a ceiling because of low margins and a 350–400 day cycle, so it adds revenue but dilutes returns.
Execution and balance-sheet risks: The Saudi JV needs ~$30mn of capex, with a Feb–Mar 2027 commissioning date and local approvals still pending. On the other side, 60% CNC utilisation gives capex-light headroom. The key tracking metrics are asset turnover toward 2.5x, working capital days, and order inflow beyond the ₹280cr book (₹180cr Unimech, ₹100cr Hobel).
2. GMM Pfaudler – An infusion of B-positive blood! Credits: Artha Vichaar Capital
GMM Pfaudler was incorporated as Gujarat Machinery Manufacturers Limited (GMM) in 1962 and got its present name in 1987 after acquisition of 40% stake by Phaudler Inc.. GMM Pfaudler is a diversified global engineering company with a heritage of more than 140 years. The Company brings together strong engineering capabilities, a cost-efficient global manufacturing footprint, and deep process know-how to help customers improve performance, reliability, and efficiency in process industries worldwide.
With 20 manufacturing facilities across four continents and a strong local presence, the Company operates through four distinct global divisions: Corrosion-Resistant Technologies (CRT), Process Performance Technologies (PPT), Heavy Engineering Technologies (HET), and Process System Technologies (PST), to serve customers worldwide.
Record order book: The backlog is ₹2,289 Cr, and order intake is ~₹1,000 Cr a quarter. Orders are forecast to rise from ₹3,714 Cr (FY26) to ₹5,540 Cr by FY29E, taking revenue from ₹3,524 Cr to ₹5,155 Cr (~13% CAGR). The plants have room, so no new glass-lined capex is needed.
Pharma is the main growth engine: Indian CDMOs and CMOs are on a capex upcycle, and US onshoring adds more demand. GMM holds over half of Indian glass-lined demand and ~40% globally, and pharma buyers rarely switch because the vendor is written into regulatory filings. Q1 FY27 orders were up 23% in CRT and 64% in PPT.
Demand is diversifying beyond chemicals: Chemicals fell from 58% to 35% of orders. Nuclear (₹130 Cr of NPCIL orders, India going from ~9 GW to ~22 GW by FY32), defence acid-recovery (a EUR 33.7m order, with USD 20–30m a year targeted) and mining (SEMCO) now take 42% of orders. A chemical capex recovery from FY28 would be pure upside.
Margin expansion from 11.4% to 15%: EBITDA is forecast to rise from ₹403 Cr to ₹772 Cr by FY29E. The drivers are operating leverage on payroll (29% to 25% of revenue), ~₹45 Cr a year of German restructuring savings, no repeat of FY26’s loss-contract provisions, and partial recovery in glass-lined pricing.
Below-EBITDA repair is the biggest profit lever: The tax rate should fall from 49% toward 30% as the loss-making entity structure is simplified. Finance cost should fall from ₹123 Cr to ~₹58 Cr through debt repayment and refinancing at 6–7% instead of ~9.9%. Together these add ~₹143 Cr to FY29E profit.
Earnings and returns compound: Adjusted EPS goes from ₹20 to ₹90 by FY29E and ROCE from 12.8% to 22.4%. Net debt turns to net cash by FY29E (₹354 Cr of net debt now), and free cash flow recovers to ~₹270 Cr.
3. Pitti Engineering - Everything inside the motor, under one roof
Pitti makes the metal parts that go inside electric motors and generators. These are the laminations (thin steel sheets stacked to form the motor core), castings (moulded metal housings) and machined parts (castings finished to exact sizes), plus shafts and full assemblies. Customers include Cummins, Marathon, Nidec, ABB, CG, Siemens etc.
Selling more: Lamination volumes were ~19,200 tons in Q1 (+19%), and Management raised the FY27 target to 82,000 tons from 78,000 tons.
Extra space already built: A ₹150 Cr expansion has just started operating. It lifts the sheet-metal (lamination) capacity to 108,000 tons a year, and also adds casting and machining capacity. Sheet-metal utilisation is 73% (up from 70%), machining is 86% (up from 82%) and casting and fabrication is 72%.
At the 80% utilisation management considers efficient, there is room for only about 8,000 tons more, which is why the next expansion is already planned.
Moving to higher-value products: Instead of selling only loose laminations, Pitti now sells finished rotor-shaft and stator assemblies that combine lamination, casting and machining. These assemblies grew about 37% in volume, against about 16% for loose laminations. The growth comes from data centres, mining, off-highway vehicles, wind power and special industrial uses. EBITDA per ton is expected to be more than double for these higher value products!
Casting and machining is the fastest-growing part: Casting capacity is now 24,000 tons, and the FY27 casting target goes up to about 17,000 tons from 16,000. Machined parts are growing over 50%, because customers want a finished part, not a raw casting. Machining is limited by how fast new machines arrive, not by orders. The ₹290 Cr project will take machining capacity to about 1.08 million machine hours, in steps over the coming quarters.
Many industries and a global shift to India: Traction motors and railways revenue share stands at 28%, power generation 15%, industrial and commercial 12%, mining and oil & gas 10% (doubled from 5%), special motors 9%, data centres 5%, renewables 3% and others 17%. Companies are moving production from China and Europe to India because of high European energy and labour costs. Some of what Pitti supplies to global companies’ Indian plants is later exported, which management calls indirect exports. North American railway modernisation is another source of demand. Management is cautious about data centres, saying AI-related spending may not last, so it is not building capacity far ahead for them.
Exports should recover: Exports were flat at about ₹139 Cr (against about ₹137 Cr last year), while domestic sales grew 23%. Management expects direct exports to rise each quarter from Q2 to Q4 as the new capacity ramps up.
Profits should grow faster than sales: The Q1 adjusted EBITDA margin was 16.8%, held back because staff and costs for the new plants are already in place while volumes have not yet caught up. Management guides FY27 EBITDA of about ₹370 Cr. Next year it expects turnover above ₹2,500 Cr at a 17–17.2% margin, assuming 90,000 tons and no further lamination capacity. Within about three years it expects margins of 18–18.5%. The tax rate should be about 25% for the full year (Q1 was lower because of a one-off deferred tax effect). A state incentive of about ₹40 Cr a year is expected from next year, but not this year, and about ₹70 Cr of past incentives is still to be received in the next 9–12 months.
Size ambition: current capacity supports about ₹2,500 Cr of revenue. Adding the ₹290 Cr project and a further ₹400 Cr (a ₹200 Cr Bangalore plant and ₹200 Cr of equipment) would support ₹3,000–3,300 Cr. The Bangalore plant is not yet approved by the board.
Risks: Net debt is ₹491 Cr with more capex coming, mining and data-centre demand can reverse quickly, and the margin gains are so far only guidance.
4. The Rising Cancer Burden
Cancer remains one of the world's biggest healthcare challenges, with around 20+ million new cases reported anually. The global cancer burden is projected to rise to 35 million annual cases by 2050, driving the need for more effective and targeted treatments.
For decades, cancer treatment has largely relied on chemotherapy, radiation and surgery. Chemotherapy works by killing rapidly dividing cells, including cancer cells, but it can also damage healthy cells, leading to side effects such as hair loss, fatigue and nausea. While these treatments have saved millions of lives, treating cancer effectively without causing significant damage to the rest of the body has remained a major challenge.
Targeted therapies and immunotherapy have improved treatment outcomes, but not every patient responds, and cancers can eventually develop resistance.
Antibody-Drug Conjugates (ADCs) are changing the dynamics by combining the targeting ability of antibodies with the cancer-killing power of chemotherapy.
Think of them as guided missiles: the antibody identifies specific proteins on cancer cells and delivers a potent drug payload more directly to the tumour, potentially reducing damage to healthy cells. The technology is evolving rapidly, with future growth driven by new targets, more effective payloads, better linkers and the development of ADCs for additional cancer types and earlier stages of treatment.
The next frontier includes combination therapies and next-generation ADC designs, making this an exciting area of innovation for the pharmaceutical industry. However, ADCs are not side-effect-free, and their success depends on the cancer type, target and individual patient.
The ADC field reached a defining inflection point in 2025. According to the Beacon ADC database maintained by Hanson Wade, a total of 2,334 distinct antibody-based drug conjugates were tracked globally as of January 5, 2026, representing a 20% YoY increase in traditional ADCs and an extraordinary 88% YoY increase in novel ADC formats.
Most strikingly, 130 new ADC candidates progressed into clinical development in 2025, a 49% increase over the prior year and the highest single-year clinical entry figure ever recorded.
Among Indian CDMO names, Piramal Pharma, Cohance Lifesciences, Anthem Biosciences, Sai Life, Shilpa Medicare, Syngene, and Laurus Labs have established or are building capabilities linked to ADCs in one form or another.
5. Gravita India
Gravita India is facing near-term execution challenges as geopolitical disruptions have affected scrap availability. Management explained that 15–20% of its scrap imports came from the Gulf region, with shipments from other countries also affected by transit disruptions. Lead volumes have suffered, and management cautioned that Q2 FY27 could also see an impact, although it expects to protect EBITDA margins through better realisations and product mix. Q1 FY27 EBITDA per tonne stood at ₹24,181 for lead, ₹25,175 for aluminium, ₹10,197 for plastic and ₹55,151 for copper. Meanwhile, copper capacity utilisation remains around 50%, and elevated working capital is another monitorable.
All of this has led to a cooling off in valuations. At current levels, the stock is trading below its 5-year median valuations.
However, the near-term story, including the remainder of FY27, appears challenging. It’s not a bargain yet, but valuations are certainly looking more reasonable.
The longer-term opportunity remains meaningful. Gravita's acquisition of Rashtriya Metal Industries (RMIL) expands its presence in copper and copper alloys, while its new 29,400-tonne copper recycling facility and plans to double copper capacity to around 60,000 tonnes over three years provide additional growth avenues. Management targets copper EBITDA per tonne of ₹70,000–75,000 over the next 2–3 years, supported by debottlenecking, better procurement and integration.
With total installed capacity targeted to rise from 4.97 lakh tonnes to over 8 lakh tonnes by FY29, Gravita has room to scale, provided scrap sourcing normalises, utilisation improves and returns on invested capital strengthen. Management's 25–30% annual PAT growth aspiration is encouraging, but execution remains key.
Could Gravita offer a more attractive opportunity post Q2 results? We’ll have to wait and watch!
That's all for this edition. Have a great Sunday!
Disclaimer: None or buy or sell recommendations. This publicly available information is shared for learning and education purposes. https://x.com/LearningEleven/status/2109124902028538056
## @vijaythirumalai (Vijay Thirumalai) · 10-11 03:13 · ♥56 ↻18 💬6 To all the financially & arithemtically challenged people asking how Market cap is converted to cash, here is the napkin math
1/ Jio net block is ~$45B, Star link need not spend so much as much of access infrastructe is moved to orbit, lets say incremental capex for SpaceX is $30B
2/ Operating Loss is say $1.5 B ( Rs 15,000 cr /year), so 5 years of operating loss at tops is $7.5B
3/ Say by 2032, Star link gets to Jio number of $30B of revenues and $18B of EBIDTA , you are loooking at $150B of EV for Starlink Inida business
4/ Discounting back, PV of 2032 EV ~$100 B
5/ Basically $40B of Capex + Opex gets Musk $180B of EV+ the ability to Lord over 1.2 B Indians & huge security issues
5/ $40B of capex is less than what Musk spent for Cursor $60B + Markets will reward him with the price increase right way
6/ In essence Musk needs to spend < 1% market cap to build $150 B of Market cap in 5-6 years just from Indian business
7/ This $150 B of EV should have been typically shared between Jio and Airtel and going out of the country ( bascially hits your MF SIP performance as both are huge part of NIFTY)
8/ And what do we get in return for letting go for $150B Market cap, non stop , sanctimonious advise from the Lords on how to run our country + zero benefits of so called Rural connectivity
9/ To rub it all you give up the most important pieces of digital infra ( after having already given up cloud, Social media, AI to US companies) for zero gain
10/ Also this is not the same as $META / $MSFT investment in Jio, those were minority strategic investments, do more that, don't let this megalomaniac in India and do the same of what you did with East India company
Now tell me why again we need SpaceX in India apart from your random nonsense of customer surplus argument? > 引用 @vijaythirumalai: See so many brain dead takes on Starlink saying there should be more competition to Airtel & Jio
Let me give you a quick Capitalism 101 class, on what would happen
1/ Combined Jio + Airtel market cap is ~ 24 L cr - say $240 b
2/ SpaceX market cap is $2.3 T ( 10X combined MCap of Jio and Airtel)
3/ Moment Starlink is launched in India, Market cap would correct by 10-20% ( like what happened in US), say $24B
4/ Musk can infinitely undercut Jio and Airtel forever, they have $2.4T of Mcap to burn
5/ Non Zero proababilty of Airtel &/Jio going bankrupt
6/ Worst Part is ZERO upside for India- majority of zip codes from Kashmir to Kanyakumari have been covered & (apart from some made up use cases of some guy stuck in the mountains of Arunachal Pradesh for whom Star link is required)
7/ Worse case after Bankrupting Airtel and Jio, Musk can incease the price by 10X with no competition and we will be sitting ducks
8/ We have already lost the AI, Big Tech war, telecom is a basic necessity/ utility and we can't afford to lose this for ZERO gain to India
9/ If we want FDI, do what Jio did, get META, MSFT as investors, don't get an operator
You are welcome! https://x.com/vijaythirumalai/status/2109120295206596912