My study, in one breath: AI ke har data centre me GPUs itni dense pack ho rahe hain ki garmi (heat) ek dum se phat rahi hai. Purana "air cooling" (pankhe/AC) ab kaafi nahi — kyunki ek rack ki power 10kW se 50–130kW tak jaa rahi hai. Iska seedha matlab: liquid cooling aur immersion cooling ab "optional" nahi, "compulsory" ban rahi hai. Aur jo companies ye cooling banaati hain — woh GPUs ke peeche chhupi hui "picks & shovels" hain.
Yeh ek detailed STUDY hai, advice nahi. Niche US ke top 5 aur India ke top 5 (listed) cooling-linked stocks ka analysis hai, aur future perspective me ye theme kitni zaroori hai — sab record pe, date ke saath.
Main koi SEBI/SEC-registered advisor nahi hoon. Main ek founder hoon jo technology trends ko obsessively track karta hai. Yeh blog "tips" nahi — ek research-backed study hai, public aur dated. Maine pehle silicon photonics pe likha tha (chip ke andar data movement); yeh usi AI-infra story ka doosra hissa hai — jab itna data move hoga aur itne chips chalenge, to garmi sabse bada physical problem banegi. Aur garmi ka business = cooling.
Ek line yaad rakhna: "Compute grows, but heat governs." Chip kitna bhi powerful ho, agar usko thanda nahi rakha to woh throttle ho jaata hai ya jal jaata hai. Isiliye cooling AI ka silent gatekeeper hai. Chalo deep dive karte hain.
Heat ka Problem — Aakhir Itni Garmi Kyun?
Har chip bijli (electricity) ko kaam me badalta hai, aur uska ek hissa hamesha garmi banke nikalta hai. Normal computer me ye garmi chhoti hoti hai — pankha kaafi. Lekin AI ka GPU ek chhote heater jaisa hota hai, aur data centre me hazaaron GPUs ek saath, ek hi rack me, kandhe se kandha milake baithe hote hain.
Numbers samjho: pehle ek server rack ~5–10 kilowatt (kW) power kheenchta tha. AI ke aane ke baad ek rack 50, 80, yahan tak ki 130+ kW tak jaa raha hai. Itni power = itni hi garmi ek chhoti si jagah me. Yeh ab "thoda garam" nahi — yeh "ek choti bhatti" wali situation hai. Aur agar garmi nahi nikli, to:
- Chip throttle kar deta hai (speed kam — paisa jala, kyunki mehenga GPU full speed nahi de raha).
- Hardware fail ho sakta hai (lakhon ka GPU kharab).
- Bijli ka bill phat jaata hai (cooling pe hi data centre ki ~30–40% energy lag sakti hai).
Isiliye cooling sirf "comfort" nahi — ye seedhe profit, uptime aur GPU ki life se juda hua hai. Jo data centre garmi sahi se handle karega, wahi AI ki race jeetega.
Cooling Tech ke Types — Air se Liquid se Immersion tak
Cooling ek "ladder" hai — jaise-jaise garmi badhti hai, upar wali tech chahiye hoti hai:
- 1. Air cooling (purana tarika): pankhe + AC se hawa thandi karo. Kaam karta hai jab tak rack ~15–20 kW ho. AI ke liye ab kaafi nahi.
- 2. Rear-door heat exchangers: rack ke peeche ek liquid-cooled darwaza — air ko exit pe thanda karta hai. Beech ka step.
- 3. Direct-to-chip liquid cooling: seedha chip ke upar ek metal plate (cold plate) jisme paani/coolant behta hai, garmi seedha utha leta hai. Yahi abhi ka mainstream AI solution ban raha hai.
- 4. Immersion cooling: poora server ek special non-conductive liquid me dubo diya jaata hai. Sabse efficient, sabse advanced — abhi early adoption me, par future bada.
Mera point: industry tezi se air se liquid (direct-to-chip) ki taraf shift ho rahi hai, aur aage immersion bhi aayega. Jo companies ye liquid-cooling systems, cold plates, CDUs (coolant distribution units), heat exchangers aur poora thermal setup banaati hain — unke liye ye ek multi-year tailwind hai.
Boom Kyun, Aur Kyun ABHI?
Ek line me: AI data centre buildout. Jab tak naye AI data centre ban rahe hain (aur ye to abhi shuruaat hai), tab tak rack-density badhegi, aur badhti density = badhti cooling demand. Air cooling ki physical limit aa chuki hai, isiliye liquid cooling ab choice nahi, majboori hai.
Doosra trigger: energy cost + sustainability. Cooling data centre ki sabse badi non-compute energy hai. Jo bhi liquid/immersion cooling se power bachata hai, woh seedha crore-rupaye bachata hai aur carbon goals bhi hit karta hai. Isiliye hyperscalers (bade cloud players) cooling pe seriously invest kar rahe hain.
Mera reading: cooling, AI infra ki sabse "boring" par sabse structural layer hai. GPUs glamorous hain, par cooling ke bina woh chal hi nahi sakte. Jab demand structural ho, leading cooling companies multi-year run pe hoti hain.
Kyun Koi Seedha Nahi Bata Raha (The "Khichdi")
Cooling consumer-facing nahi hai — koi flashy product nahi. Ye data centre ke andar chhupi "plumbing" hai. Media aur retail ka poora dhyaan NVIDIA/GPUs pe hai, kyunki woh "face" hai. Lekin har GPU ko zinda rakhne wali cooling — uspe spotlight bahut kam.
Yahi har bade tech cycle me hota hai: gold rush me sona dhoondhne wale (GPUs) headlines me hote hain, par "picks & shovels" bechne wale (cooling, power, optical) bhi quietly fortune banate hain. Data centre cooling us picks-and-shovels layer ka ek under-the-radar hissa hai. Jab tak ye "mainstream story" banegi, shuruaati move ho chuka hoga.
🇺🇸 US — Top 5 Cooling-Linked Stocks (Study)
⚠️ Yeh buy/sell recommendation NAHI hai — sirf "kaun is space me lead kar raha hai" ka educational map. Har naam ke saath apna research zaroor karna.
- 1. Vertiv Holdings (VRT) — data-centre thermal & power management ka pure-play leader. Liquid cooling, CDUs, rack systems — AI cooling demand ka sabse direct beneficiary. Is theme ka "poster child".
- 2. Modine Manufacturing (MOD) — thermal management specialist; data-centre cooling (incl. coolant distribution & air-to-liquid) me tezi se badh raha hai. Ek classic "old industrial → AI cooling" re-rating story.
- 3. Comfort Systems USA (FIX) — mechanical/HVAC contractor jo data centres ke liye cooling install karta hai. AI data-centre construction boom ka direct order-book beneficiary.
- 4. Trane Technologies (TT) — bada HVAC/thermal player; data-centre cooling solutions (chillers, thermal management) me strong exposure. Stable compounder type.
- 5. nVent Electric (NVT) — electrical enclosures + liquid cooling solutions (cold plates, CDUs). Cooling + power dono ka exposure ek hi naam me.
Honourable mentions: Johnson Controls (JCI) & Carrier (CARR) (HVAC giants, data-centre cooling push), Eaton (ETN) (power + cooling), aur Super Micro (SMCI) (liquid-cooled servers — par woh server play zyada hai).
🇮🇳 India — Top 5 Cooling-Linked Stocks (Honest Study)
Saaf baat: India me abhi koi bada pure data-centre liquid-cooling wala public company nahi hai — ye advanced tech mostly US ke paas hai. Lekin India ka leverage HVAC + cooling + thermal + data-centre infra me hai, jo isi AI/data-centre wave pe ride karta hai. Ye companies cooling-equipment, chillers, aur data-centre HVAC supply karti hain:
- 1. Blue Star (BLUESTARCO) — India ka leading HVAC & commercial cooling player; data-centre cooling projects me active. Sabse direct India "cooling" exposure.
- 2. Voltas (VOLTAS) — Tata group; AC, cooling & commercial refrigeration. Brand + scale + data-centre/commercial cooling demand ka beneficiary.
- 3. Amber Enterprises (AMBER) — HVAC components & contract manufacturing ka leader; cooling supply-chain ka "picks-and-shovels" India version.
- 4. Johnson Controls-Hitachi Air Conditioning India (JCHAC / "Hitachi") — premium HVAC & cooling, global parent ki tech ke saath — commercial/data-centre cooling exposure.
- 5. Kirloskar Brothers (KIRLOSBROS) — pumps & fluid systems; liquid cooling ka matlab paani/coolant pumping — yeh us "fluid movement" layer ka India play hai (thoda indirect, par structural).
Watch-also: Schneider Electric Infrastructure (power + cooling infra), data-centre operators/REIT-type plays, aur Indian semiconductor/data-centre mission ke aage badhne se thermal demand structurally badhegi. Mera maanna: India ka 1–2 saal ka play pure cooling-chip nahi, balki HVAC + thermal + fluid infra hai. Isiliye main inhe alag risk-bucket me rakhta hoon — same wave, alag profile.
8 Criteria — Jinki Wajah Se Is Theme Me Dum Hai
- 1. Physics ki majboori. Garmi ko nikalna padta hai — ye fashion nahi, thermodynamics hai. Air ki limit aa chuki, liquid hi raasta.
- 2. Rack density exponential. 10kW → 130kW per rack. Jitni density, utni cooling. Linear solution exponential demand nahi sambhalta.
- 3. Energy = sabse bada kharcha. Cooling data-centre ki ~30–40% energy. Efficiency = direct profit + carbon goals. Hyperscalers majbooran invest karenge.
- 4. Liquid cooling roadmap on hai. Industry air → direct-to-chip → immersion shift kar rahi hai. Supply chain ko follow karna majboori.
- 5. Picks-and-shovels durability. GPU war me koi bhi jeete, cooling sabko chahiye. Diversified, structural demand.
- 6. High barrier to entry. Thermal engineering, reliability, certifications — raaton-raat copy nahi hota. Leaders ka moat strong.
- 7. Hyperscaler capex record pe. AI infra spending historic; cooling us capex ka growing hissa.
- 8. Under-owned / under-covered. Retail ka dhyaan GPUs pe; cooling layer pe spotlight kam = kam premium = early entry ka mauka (contrarian point).
History Rhymes — Aisा Pehle Bhi Hua
- 2010s cloud buildout: jab cloud bada hua, data-centre infra (power + cooling) companies ne quietly strong multi-year run diya — kyunki har server ko power aur cooling chahiye tha.
- 2023–2025 AI/GPU wave: pehle GPUs (NVIDIA) chale, phir memory (HBM), phir networking/optical. Har wave me spotlight "next layer" pe shift. Cooling + power us "physical infra" layer hai jise abhi tak full credit nahi mila.
Pattern same: bottleneck → demand → leaders ko multi-year tailwind. Garmi ab woh bottleneck hai jo har AI data centre face kar raha hai.
Future Perspective — Ye Kitna Zaroori Hai?
Yeh sabse important section hai. Aage chal ke AI sirf badhega — bade models, zyada GPUs, zyada data centre. Iska seedha matlab: garmi ka problem har saal aur bada hoga. Aur jaise-jaise rack density badhegi, air cooling poori tarah retire ho jaayega aur liquid/immersion default ban jaayega.
Future me cooling teen reasons se aur critical hogi:
- Power/grid pressure: data centres itni bijli kheench rahe hain ki grids pe load aa raha hai. Efficient cooling = kam waste = grid pe kam pressure. Sarkar/regulator bhi efficiency push karenge.
- Sustainability rules: carbon aur water-usage pe norms aayenge — efficient cooling compliance ka hissa ban jaayega.
- Edge + AI everywhere: chhote local data centres (edge) bhi badhenge — har jagah thermal management chahiye.
Isiliye mera maanna: cooling AI infra ka "permanent tax" hai — jab tak compute hai, cooling chahiye. Yeh ek fad nahi, ek structural, multi-decade need hai. Jo isme leaders hain, woh long-term tailwind pe baithe hain.
Main Galat Kahan Ho Sakta Hoon (Invalidation)
- AI capex slowdown — agar hyperscalers spending kam karein (recession/over-build), to cooling demand cycle slip kar jaata hai.
- Breakthrough chip efficiency — agar chips achanak bahut kam garmi paida karne lagein (naya architecture), to cooling intensity kam ho sakti hai.
- Timeline lamba — liquid/immersion mainstream hone me 1–2 ki jagah 3–4 saal lagein, to "1–2 year" wala part galat.
- Valuation already rich — kuch leaders (jaise Vertiv) pehle hi bahut daud chuke ho sakte hain; high valuation pe entry pe achha business bhi bura trade ban sakta hai.
- Commoditization — agar cooling sasta/commodity ban jaaye, to margins gir sakti hain.
📋 Study Scorecard — 2027–2028 Me Tick Karenge
Aaj 29 June 2026 hai. ~1–2 saal baad wapas aana aur tick karna:
Yaad rakhna — maine ye 29 June 2026 ko likha tha, jab spotlight sirf GPUs pe thi. Galat nikla to isi page pe likhunga ki kahan chuka.
Main Personally Kya Dekh Raha Hoon (Watchlist, Advice Nahi)
- Rack density trends — kitni tezi se kW-per-rack badh raha hai.
- Liquid vs immersion adoption — kaun production me la raha hai, kaun roadmap dikha raha hai.
- Cooling companies ki data-centre revenue mix — traditional HVAC se data-centre ki taraf shift kitna tez.
- Hyperscaler capex commentary — har quarter spending badh raha hai ya ruk raha hai.
- India data-centre buildout + HVAC order books — Blue Star/Voltas/Amber ke margins aur orders.
- Valuations — kahीं leaders pehle se overheated to nahi.
Mera personal principle wahi: emergency fund (6 mahine) kabhi haath mat lagao, udhaar leke speculation kabhi mat karo, aur foreign/thematic stocks me position chhoti rakho — chahe study kitni bhi strong lage.
Bottom Line
Aaj 29 June 2026 ko, jab duniya sirf GPU ka naam le rahi hai, mera nazariya ek layer aage hai — silicon photonics ke baad ab cooling. AI compute jitna badhega, garmi utni hi badhegi, aur usko handle karne wali tech (liquid + immersion cooling) "optional" se "compulsory" ban jaayegi. Iska exposure US me Vertiv, Modine, Comfort Systems, Trane, nVent me hai, aur India me Blue Star, Voltas, Amber, JCHAC, Kirloskar Brothers (HVAC + thermal + fluid infra) me.
Phir se — yeh ek study hai, advice nahi. Main future me sochta hoon, aaj ke shor me nahi. Ye theme ab record pe hai, date ke saath, logic ke saath, aur galat hone ki shart ke saath. 1–2 saal baad milte hain, scorecard tick karne. ❄️🔥
Useful BillCraft Tools
Apne investments aur tax ko plan karne ke liye free Indian tools:
- Capital Gains Calculator — US/Indian stock sale pe LTCG/STCG nikaalo
- SIP vs Lumpsum Calculator — thematic bets me staggered entry plan karo
- Goal-Based Investment Calculator — target amount ke liye plan banao
- Income Tax Calculator 2026 — foreign income/gains pe tax samjho
Agar ye study aapko ek "layer aage" sochne pe majboor kar gayi — chahe agree karo ya disagree — to ise share karo. Aur mera silicon photonics wala blog bhi padhna mat bhoolna. 🔖
My study, in one breath: In every AI data centre, GPUs are being packed so densely that the heat is suddenly exploding. Old-school "air cooling" (fans/AC) is no longer enough — because a single rack's power is climbing from 10kW to 50–130kW. The direct meaning: liquid cooling and immersion cooling are no longer "optional," they're becoming "compulsory." And the companies that build this cooling — they are the hidden "picks & shovels" sitting behind the GPUs.
This is a detailed STUDY, not advice. Below is an analysis of the top 5 US and top 5 Indian (listed) cooling-linked stocks, and how essential this theme is from a future perspective — all on the record, with a date.
I am not a SEBI/SEC-registered advisor. I'm a founder who tracks technology trends obsessively. This blog isn't "tips" — it's a research-backed study, public and dated. I wrote earlier about silicon photonics (data movement inside the chip); this is the second part of that same AI-infra story — when this much data moves and this many chips run, heat becomes the biggest physical problem. And the business of heat = cooling.
Remember one line: "Compute grows, but heat governs." No matter how powerful a chip is, if it isn't kept cool it throttles or burns out. That's why cooling is AI's silent gatekeeper. Let's do a deep dive.
The Heat Problem — Why So Much Heat in the First Place?
Every chip turns electricity into work, and a part of it always comes out as heat. In a normal computer this heat is small — a fan is enough. But an AI GPU is like a little heater, and in a data centre thousands of GPUs sit together in a single rack, shoulder to shoulder.
Understand the numbers: earlier, a server rack drew ~5–10 kilowatts (kW) of power. After AI arrived, a single rack is going up to 50, 80, even 130+ kW. That much power = that much heat in one tiny space. This isn't "a little warm" anymore — it's a "small furnace" situation. And if the heat doesn't escape:
- The chip throttles (speed drops — money burned, because an expensive GPU isn't running at full speed).
- Hardware can fail (a GPU worth lakhs gets damaged).
- The electricity bill explodes (cooling alone can take ~30–40% of a data centre's energy).
That's why cooling isn't just "comfort" — it's tied directly to profit, uptime and GPU life. Whichever data centre handles heat properly will win the AI race.
Types of Cooling Tech — From Air to Liquid to Immersion
Cooling is a "ladder" — as the heat rises, you need the tech a step higher:
- 1. Air cooling (the old way): cool the air with fans + AC. Works as long as a rack is ~15–20 kW. For AI it's no longer enough.
- 2. Rear-door heat exchangers: a liquid-cooled door behind the rack — cools the air on its way out. The intermediate step.
- 3. Direct-to-chip liquid cooling: a metal plate (cold plate) sits right on top of the chip with water/coolant flowing through it, lifting heat away directly. This is becoming today's mainstream AI solution.
- 4. Immersion cooling: the entire server is submerged in a special non-conductive liquid. The most efficient, the most advanced — still in early adoption, but a big future.
My point: the industry is rapidly shifting from air to liquid (direct-to-chip), and immersion will come next. For the companies that build these liquid-cooling systems, cold plates, CDUs (coolant distribution units), heat exchangers and the whole thermal setup — this is a multi-year tailwind.
Why a Boom, and Why NOW?
In one line: the AI data centre buildout. As long as new AI data centres are being built (and this is just the beginning), rack density will keep rising, and rising density = rising cooling demand. Air cooling has hit its physical limit, so liquid cooling is no longer a choice — it's a compulsion.
The second trigger: energy cost + sustainability. Cooling is a data centre's biggest non-compute energy. Whoever saves power with liquid/immersion cooling directly saves crores of rupees and hits carbon goals too. That's why hyperscalers (the big cloud players) are investing seriously in cooling.
My reading: cooling is AI infra's most "boring" but most structural layer. GPUs are glamorous, but without cooling they simply can't run. When demand is structural, the leading cooling companies are on a multi-year run.
Why Nobody Is Saying It Straight (The "Khichdi")
Cooling isn't consumer-facing — there's no flashy product. It's the "plumbing" hidden inside the data centre. Media and retail attention is entirely on NVIDIA/GPUs, because that's the "face." But the cooling that keeps every GPU alive — the spotlight on it is very low.
This happens in every big tech cycle: in a gold rush, the gold-seekers (GPUs) make the headlines, but the ones selling "picks & shovels" (cooling, power, optical) also quietly make a fortune. Data centre cooling is an under-the-radar part of that picks-and-shovels layer. By the time it becomes a "mainstream story," the early move will already be done.
🇺🇸 US — Top 5 Cooling-Linked Stocks (Study)
⚠️ This is NOT a buy/sell recommendation — just an educational map of "who is leading in this space." Do your own research with every name.
- 1. Vertiv Holdings (VRT) — the pure-play leader in data-centre thermal & power management. Liquid cooling, CDUs, rack systems — the most direct beneficiary of AI cooling demand. The "poster child" of this theme.
- 2. Modine Manufacturing (MOD) — thermal management specialist; growing fast in data-centre cooling (incl. coolant distribution & air-to-liquid). A classic "old industrial → AI cooling" re-rating story.
- 3. Comfort Systems USA (FIX) — a mechanical/HVAC contractor that installs cooling for data centres. A direct order-book beneficiary of the AI data-centre construction boom.
- 4. Trane Technologies (TT) — a big HVAC/thermal player; strong exposure in data-centre cooling solutions (chillers, thermal management). A stable compounder type.
- 5. nVent Electric (NVT) — electrical enclosures + liquid cooling solutions (cold plates, CDUs). Exposure to both cooling + power in a single name.
Honourable mentions: Johnson Controls (JCI) & Carrier (CARR) (HVAC giants, data-centre cooling push), Eaton (ETN) (power + cooling), and Super Micro (SMCI) (liquid-cooled servers — but that's more of a server play).
🇮🇳 India — Top 5 Cooling-Linked Stocks (Honest Study)
Straight talk: India doesn't yet have a big pure data-centre liquid-cooling public company — that advanced tech mostly sits with the US. But India's leverage is in HVAC + cooling + thermal + data-centre infra, which rides the same AI/data-centre wave. These companies supply cooling-equipment, chillers, and data-centre HVAC:
- 1. Blue Star (BLUESTARCO) — India's leading HVAC & commercial cooling player; active in data-centre cooling projects. The most direct India "cooling" exposure.
- 2. Voltas (VOLTAS) — Tata group; AC, cooling & commercial refrigeration. A beneficiary of brand + scale + data-centre/commercial cooling demand.
- 3. Amber Enterprises (AMBER) — leader in HVAC components & contract manufacturing; the India version of the cooling supply-chain "picks-and-shovels."
- 4. Johnson Controls-Hitachi Air Conditioning India (JCHAC / "Hitachi") — premium HVAC & cooling, with the global parent's tech — commercial/data-centre cooling exposure.
- 5. Kirloskar Brothers (KIRLOSBROS) — pumps & fluid systems; liquid cooling means water/coolant pumping — this is the India play on that "fluid movement" layer (a bit indirect, but structural).
Watch-also: Schneider Electric Infrastructure (power + cooling infra), data-centre operators/REIT-type plays, and as the Indian semiconductor/data-centre mission moves ahead, thermal demand will rise structurally. My view: India's 1–2 year play isn't pure cooling-chip, but HVAC + thermal + fluid infra. That's why I keep them in a separate risk-bucket — same wave, different profile.
8 Criteria — Why There's Substance in This Theme
- 1. The compulsion of physics. Heat must be removed — this isn't fashion, it's thermodynamics. Air has hit its limit, liquid is the only road.
- 2. Rack density is exponential. 10kW → 130kW per rack. The more density, the more cooling. A linear solution can't handle exponential demand.
- 3. Energy = the biggest expense. Cooling is ~30–40% of a data-centre's energy. Efficiency = direct profit + carbon goals. Hyperscalers will be forced to invest.
- 4. The liquid cooling roadmap is on. The industry is shifting air → direct-to-chip → immersion. Following the supply chain is a compulsion.
- 5. Picks-and-shovels durability. Whoever wins the GPU war, everyone needs cooling. Diversified, structural demand.
- 6. High barrier to entry. Thermal engineering, reliability, certifications — not copied overnight. The leaders' moat is strong.
- 7. Hyperscaler capex on the record. AI infra spending is historic; cooling is a growing slice of that capex.
- 8. Under-owned / under-covered. Retail attention is on GPUs; low spotlight on the cooling layer = lower premium = a chance for early entry (the contrarian point).
History Rhymes — This Has Happened Before
- 2010s cloud buildout: when cloud grew big, data-centre infra (power + cooling) companies quietly delivered a strong multi-year run — because every server needed power and cooling.
- 2023–2025 AI/GPU wave: first GPUs (NVIDIA) ran, then memory (HBM), then networking/optical. In every wave the spotlight shifts to the "next layer." Cooling + power is that "physical infra" layer that hasn't yet gotten full credit.
Same pattern: bottleneck → demand → a multi-year tailwind for leaders. Heat is now that bottleneck every AI data centre is facing.
Future Perspective — How Essential Is This?
This is the most important section. Going forward, AI will only grow — bigger models, more GPUs, more data centres. The direct meaning: the heat problem will get bigger every year. And as rack density rises, air cooling will retire completely and liquid/immersion will become the default.
In the future, cooling will be even more critical for three reasons:
- Power/grid pressure: data centres are drawing so much electricity that load is building on grids. Efficient cooling = less waste = less pressure on the grid. Governments/regulators will also push efficiency.
- Sustainability rules: norms on carbon and water-usage will come — efficient cooling will become part of compliance.
- Edge + AI everywhere: small local data centres (edge) will also grow — thermal management is needed everywhere.
That's why my view: cooling is AI infra's "permanent tax" — as long as there's compute, you need cooling. This isn't a fad, it's a structural, multi-decade need. Those who are leaders in it are sitting on a long-term tailwind.
Where I Could Be Wrong (Invalidation)
- AI capex slowdown — if hyperscalers cut spending (recession/over-build), the cooling demand cycle slips.
- Breakthrough chip efficiency — if chips suddenly start producing far less heat (a new architecture), cooling intensity could drop.
- Long timeline — if liquid/immersion takes 3–4 years instead of 1–2 to go mainstream, the "1–2 year" part is wrong.
- Valuation already rich — some leaders (like Vertiv) may have already run a lot; entering at a high valuation can turn even a good business into a bad trade.
- Commoditization — if cooling becomes cheap/a commodity, margins can fall.
📋 Study Scorecard — To Tick in 2027–2028
Today is 29 June 2026. Come back in ~1–2 years and tick these:
Remember — I wrote this on 29 June 2026, when the spotlight was only on GPUs. If it turns out wrong, I'll write on this very page where I missed.
What I'm Personally Watching (Watchlist, Not Advice)
- Rack density trends — how fast kW-per-rack is climbing.
- Liquid vs immersion adoption — who's bringing it into production, who's just showing a roadmap.
- Cooling companies' data-centre revenue mix — how fast the shift from traditional HVAC to data-centre is.
- Hyperscaler capex commentary — whether spending is rising or stalling each quarter.
- India data-centre buildout + HVAC order books — the margins and orders of Blue Star/Voltas/Amber.
- Valuations — whether the leaders are already overheated.
My personal principle is the same: never touch your emergency fund (6 months), never speculate with borrowed money, and keep your position small in foreign/thematic stocks — no matter how strong the study looks.
Bottom Line
Today, on 29 June 2026, when the world is only naming GPUs, my view is one layer ahead — after silicon photonics, now cooling. The more AI compute grows, the more heat grows, and the tech that handles it (liquid + immersion cooling) will go from "optional" to "compulsory." The exposure to this is in the US via Vertiv, Modine, Comfort Systems, Trane, nVent, and in India via Blue Star, Voltas, Amber, JCHAC, Kirloskar Brothers (HVAC + thermal + fluid infra).
Again — this is a study, not advice. I think into the future, not into today's noise. This theme is now on the record, with a date, with logic, and with the condition for being wrong. We'll meet in 1–2 years, to tick the scorecard. ❄️🔥
Useful BillCraft Tools
Free Indian tools to plan your investments and tax:
- Capital Gains Calculator — work out LTCG/STCG on a US/Indian stock sale
- SIP vs Lumpsum Calculator — plan a staggered entry into thematic bets
- Goal-Based Investment Calculator — build a plan for a target amount
- Income Tax Calculator 2026 — understand tax on foreign income/gains
If this study pushed you to think "one layer ahead" — whether you agree or disagree — then share it. And don't forget to read my silicon photonics blog too. 🔖
मेरी स्टडी, एक साँस में: AI के हर data centre में GPU इतने सघन (dense) पैक हो रहे हैं कि गर्मी (heat) एकदम से फट रही है। पुराना "air cooling" (पंखे/AC) अब काफ़ी नहीं — क्योंकि एक rack की power 10kW से 50–130kW तक जा रही है। इसका सीधा मतलब: liquid cooling और immersion cooling अब "optional" नहीं, "compulsory" बन रही है। और जो companies यह cooling बनाती हैं — वे GPU के पीछे छुपी हुई "picks & shovels" हैं।
यह एक विस्तृत STUDY है, सलाह नहीं। नीचे US के top 5 और India के top 5 (listed) cooling-linked stocks का विश्लेषण है, और future perspective में यह theme कितनी ज़रूरी है — सब रिकॉर्ड पर, तारीख़ के साथ।
मैं कोई SEBI/SEC-registered advisor नहीं हूँ। मैं एक founder हूँ जो technology trends को जुनूनी ढंग से ट्रैक करता है। यह ब्लॉग "tips" नहीं — एक research-backed study है, सार्वजनिक और तारीख़ के साथ। मैंने पहले silicon photonics पर लिखा था (chip के अंदर data movement); यह उसी AI-infra कहानी का दूसरा हिस्सा है — जब इतना data move होगा और इतने chips चलेंगे, तो गर्मी सबसे बड़ी physical समस्या बनेगी। और गर्मी का business = cooling।
एक लाइन याद रखना: "Compute grows, but heat governs." Chip कितना भी powerful हो, अगर उसे ठंडा नहीं रखा तो वह throttle हो जाता है या जल जाता है। इसीलिए cooling AI का silent gatekeeper है। चलो deep dive करते हैं।
गर्मी की समस्या — आख़िर इतनी गर्मी क्यों?
हर chip बिजली (electricity) को काम में बदलता है, और उसका एक हिस्सा हमेशा गर्मी बनकर निकलता है। सामान्य computer में यह गर्मी छोटी होती है — पंखा काफ़ी। लेकिन AI का GPU एक छोटे heater जैसा होता है, और data centre में हज़ारों GPU एक साथ, एक ही rack में, कंधे से कंधा मिलाकर बैठे होते हैं।
नंबर समझो: पहले एक server rack ~5–10 किलोवाट (kW) power खींचता था। AI के आने के बाद एक rack 50, 80, यहाँ तक कि 130+ kW तक जा रहा है। इतनी power = इतनी ही गर्मी एक छोटी सी जगह में। यह अब "थोड़ा गरम" नहीं — यह "एक छोटी भट्टी" वाली situation है। और अगर गर्मी नहीं निकली, तो:
- Chip throttle कर देता है (speed कम — पैसा जला, क्योंकि महँगा GPU full speed नहीं दे रहा)।
- Hardware fail हो सकता है (लाखों का GPU ख़राब)।
- बिजली का बिल फट जाता है (cooling पर ही data centre की ~30–40% energy लग सकती है)।
इसीलिए cooling सिर्फ़ "comfort" नहीं — यह सीधे profit, uptime और GPU की life से जुड़ा हुआ है। जो data centre गर्मी सही से handle करेगा, वही AI की race जीतेगा।
Cooling Tech के प्रकार — Air से Liquid से Immersion तक
Cooling एक "सीढ़ी" है — जैसे-जैसे गर्मी बढ़ती है, ऊपर वाली tech चाहिए होती है:
- 1. Air cooling (पुराना तरीक़ा): पंखे + AC से हवा ठंडी करो। काम करता है जब तक rack ~15–20 kW हो। AI के लिए अब काफ़ी नहीं।
- 2. Rear-door heat exchangers: rack के पीछे एक liquid-cooled दरवाज़ा — हवा को exit पर ठंडा करता है। बीच का step।
- 3. Direct-to-chip liquid cooling: सीधा chip के ऊपर एक metal plate (cold plate) जिसमें पानी/coolant बहता है, गर्मी सीधा उठा लेता है। यही अभी का mainstream AI solution बन रहा है।
- 4. Immersion cooling: पूरा server एक special non-conductive liquid में डुबो दिया जाता है। सबसे efficient, सबसे advanced — अभी early adoption में, पर future बड़ा।
मेरा point: industry तेज़ी से air से liquid (direct-to-chip) की तरफ़ shift हो रही है, और आगे immersion भी आएगा। जो companies ये liquid-cooling systems, cold plates, CDUs (coolant distribution units), heat exchangers और पूरा thermal setup बनाती हैं — उनके लिए यह एक multi-year tailwind है।
Boom क्यों, और क्यों अभी?
एक लाइन में: AI data centre buildout। जब तक नए AI data centre बन रहे हैं (और यह तो अभी शुरुआत है), तब तक rack-density बढ़ेगी, और बढ़ती density = बढ़ती cooling demand। Air cooling की physical limit आ चुकी है, इसीलिए liquid cooling अब choice नहीं, मजबूरी है।
दूसरा trigger: energy cost + sustainability। Cooling data centre की सबसे बड़ी non-compute energy है। जो भी liquid/immersion cooling से power बचाता है, वह सीधे करोड़ों रुपये बचाता है और carbon goals भी hit करता है। इसीलिए hyperscalers (बड़े cloud players) cooling पर seriously invest कर रहे हैं।
मेरा reading: cooling, AI infra की सबसे "boring" पर सबसे structural layer है। GPU glamorous हैं, पर cooling के बिना वे चल ही नहीं सकते। जब demand structural हो, तो leading cooling companies multi-year run पर होती हैं।
क्यों कोई सीधा नहीं बता रहा (The "खिचड़ी")
Cooling consumer-facing नहीं है — कोई flashy product नहीं। यह data centre के अंदर छुपी "plumbing" है। Media और retail का पूरा ध्यान NVIDIA/GPU पर है, क्योंकि वही "face" है। लेकिन हर GPU को ज़िंदा रखने वाली cooling — उस पर spotlight बहुत कम।
यही हर बड़े tech cycle में होता है: gold rush में सोना ढूँढने वाले (GPU) headlines में होते हैं, पर "picks & shovels" बेचने वाले (cooling, power, optical) भी चुपचाप fortune बनाते हैं। Data centre cooling उस picks-and-shovels layer का एक under-the-radar हिस्सा है। जब तक यह "mainstream story" बनेगी, शुरुआती move हो चुका होगा।
🇺🇸 US — Top 5 Cooling-Linked Stocks (Study)
⚠️ यह buy/sell recommendation नहीं है — सिर्फ़ "कौन इस space में lead कर रहा है" का educational map। हर नाम के साथ अपना research ज़रूर करना।
- 1. Vertiv Holdings (VRT) — data-centre thermal & power management का pure-play leader। Liquid cooling, CDUs, rack systems — AI cooling demand का सबसे direct beneficiary। इस theme का "poster child"।
- 2. Modine Manufacturing (MOD) — thermal management specialist; data-centre cooling (incl. coolant distribution & air-to-liquid) में तेज़ी से बढ़ रहा है। एक classic "old industrial → AI cooling" re-rating story।
- 3. Comfort Systems USA (FIX) — mechanical/HVAC contractor जो data centres के लिए cooling install करता है। AI data-centre construction boom का direct order-book beneficiary।
- 4. Trane Technologies (TT) — बड़ा HVAC/thermal player; data-centre cooling solutions (chillers, thermal management) में strong exposure। Stable compounder type।
- 5. nVent Electric (NVT) — electrical enclosures + liquid cooling solutions (cold plates, CDUs)। Cooling + power दोनों का exposure एक ही नाम में।
Honourable mentions: Johnson Controls (JCI) & Carrier (CARR) (HVAC giants, data-centre cooling push), Eaton (ETN) (power + cooling), और Super Micro (SMCI) (liquid-cooled servers — पर वह server play ज़्यादा है)।
🇮🇳 India — Top 5 Cooling-Linked Stocks (Honest Study)
साफ़ बात: India में अभी कोई बड़ी pure data-centre liquid-cooling वाली public company नहीं है — यह advanced tech ज़्यादातर US के पास है। लेकिन India का leverage HVAC + cooling + thermal + data-centre infra में है, जो इसी AI/data-centre wave पर ride करता है। ये companies cooling-equipment, chillers, और data-centre HVAC supply करती हैं:
- 1. Blue Star (BLUESTARCO) — India का leading HVAC & commercial cooling player; data-centre cooling projects में active। सबसे direct India "cooling" exposure।
- 2. Voltas (VOLTAS) — Tata group; AC, cooling & commercial refrigeration। Brand + scale + data-centre/commercial cooling demand का beneficiary।
- 3. Amber Enterprises (AMBER) — HVAC components & contract manufacturing का leader; cooling supply-chain का "picks-and-shovels" India version।
- 4. Johnson Controls-Hitachi Air Conditioning India (JCHAC / "Hitachi") — premium HVAC & cooling, global parent की tech के साथ — commercial/data-centre cooling exposure।
- 5. Kirloskar Brothers (KIRLOSBROS) — pumps & fluid systems; liquid cooling का मतलब पानी/coolant pumping — यह उस "fluid movement" layer का India play है (थोड़ा indirect, पर structural)।
Watch-also: Schneider Electric Infrastructure (power + cooling infra), data-centre operators/REIT-type plays, और Indian semiconductor/data-centre mission के आगे बढ़ने से thermal demand structurally बढ़ेगी। मेरा मानना: India का 1–2 साल का play pure cooling-chip नहीं, बल्कि HVAC + thermal + fluid infra है। इसीलिए मैं इन्हें अलग risk-bucket में रखता हूँ — same wave, अलग profile।
8 Criteria — जिनकी वजह से इस Theme में दम है
- 1. Physics की मजबूरी। गर्मी को निकालना पड़ता है — यह fashion नहीं, thermodynamics है। Air की limit आ चुकी, liquid ही रास्ता।
- 2. Rack density exponential। 10kW → 130kW per rack। जितनी density, उतनी cooling। Linear solution exponential demand नहीं सँभालता।
- 3. Energy = सबसे बड़ा ख़र्चा। Cooling data-centre की ~30–40% energy। Efficiency = direct profit + carbon goals। Hyperscalers मजबूरन invest करेंगे।
- 4. Liquid cooling roadmap on है। Industry air → direct-to-chip → immersion shift कर रही है। Supply chain को follow करना मजबूरी।
- 5. Picks-and-shovels durability। GPU war में कोई भी जीते, cooling सबको चाहिए। Diversified, structural demand।
- 6. High barrier to entry। Thermal engineering, reliability, certifications — रातों-रात copy नहीं होता। Leaders का moat strong।
- 7. Hyperscaler capex रिकॉर्ड पर। AI infra spending historic; cooling उस capex का growing हिस्सा।
- 8. Under-owned / under-covered। Retail का ध्यान GPU पर; cooling layer पर spotlight कम = कम premium = early entry का मौक़ा (contrarian point)।
History Rhymes — ऐसा पहले भी हुआ
- 2010s cloud buildout: जब cloud बड़ा हुआ, data-centre infra (power + cooling) companies ने चुपचाप strong multi-year run दिया — क्योंकि हर server को power और cooling चाहिए था।
- 2023–2025 AI/GPU wave: पहले GPU (NVIDIA) चले, फिर memory (HBM), फिर networking/optical। हर wave में spotlight "next layer" पर shift। Cooling + power वह "physical infra" layer है जिसे अभी तक full credit नहीं मिला।
Pattern same: bottleneck → demand → leaders को multi-year tailwind। गर्मी अब वह bottleneck है जो हर AI data centre face कर रहा है।
Future Perspective — यह कितना ज़रूरी है?
यह सबसे important section है। आगे चलकर AI सिर्फ़ बढ़ेगा — बड़े models, ज़्यादा GPU, ज़्यादा data centre। इसका सीधा मतलब: गर्मी का problem हर साल और बड़ा होगा। और जैसे-जैसे rack density बढ़ेगी, air cooling पूरी तरह retire हो जाएगा और liquid/immersion default बन जाएगा।
Future में cooling तीन कारणों से और critical होगी:
- Power/grid pressure: data centres इतनी बिजली खींच रहे हैं कि grids पर load आ रहा है। Efficient cooling = कम waste = grid पर कम pressure। सरकार/regulator भी efficiency push करेंगे।
- Sustainability rules: carbon और water-usage पर norms आएँगे — efficient cooling compliance का हिस्सा बन जाएगी।
- Edge + AI everywhere: छोटे local data centres (edge) भी बढ़ेंगे — हर जगह thermal management चाहिए।
इसीलिए मेरा मानना: cooling AI infra का "permanent tax" है — जब तक compute है, cooling चाहिए। यह एक fad नहीं, एक structural, multi-decade need है। जो इसमें leaders हैं, वे long-term tailwind पर बैठे हैं।
मैं ग़लत कहाँ हो सकता हूँ (Invalidation)
- AI capex slowdown — अगर hyperscalers spending कम करें (recession/over-build), तो cooling demand cycle slip कर जाता है।
- Breakthrough chip efficiency — अगर chips अचानक बहुत कम गर्मी पैदा करने लगें (नया architecture), तो cooling intensity कम हो सकती है।
- Timeline लंबा — liquid/immersion mainstream होने में 1–2 की जगह 3–4 साल लगें, तो "1–2 year" वाला part ग़लत।
- Valuation already rich — कुछ leaders (जैसे Vertiv) पहले ही बहुत दौड़ चुके हो सकते हैं; high valuation पर entry पर अच्छा business भी बुरा trade बन सकता है।
- Commoditization — अगर cooling सस्ता/commodity बन जाए, तो margins गिर सकती हैं।
📋 Study Scorecard — 2027–2028 में Tick करेंगे
आज 29 जून 2026 है। ~1–2 साल बाद वापस आना और tick करना:
याद रखना — मैंने यह 29 जून 2026 को लिखा था, जब spotlight सिर्फ़ GPU पर थी। ग़लत निकला तो इसी page पर लिखूँगा कि कहाँ चूका।
मैं Personally क्या देख रहा हूँ (Watchlist, Advice नहीं)
- Rack density trends — कितनी तेज़ी से kW-per-rack बढ़ रहा है।
- Liquid vs immersion adoption — कौन production में ला रहा है, कौन roadmap दिखा रहा है।
- Cooling companies की data-centre revenue mix — traditional HVAC से data-centre की तरफ़ shift कितना तेज़।
- Hyperscaler capex commentary — हर quarter spending बढ़ रहा है या रुक रहा है।
- India data-centre buildout + HVAC order books — Blue Star/Voltas/Amber के margins और orders।
- Valuations — कहीं leaders पहले से overheated तो नहीं।
मेरा personal principle वही: emergency fund (6 महीने) कभी हाथ मत लगाओ, उधार लेकर speculation कभी मत करो, और foreign/thematic stocks में position छोटी रखो — चाहे study कितनी भी strong लगे।
Bottom Line
आज 29 जून 2026 को, जब दुनिया सिर्फ़ GPU का नाम ले रही है, मेरा नज़रिया एक layer आगे है — silicon photonics के बाद अब cooling। AI compute जितना बढ़ेगा, गर्मी उतनी ही बढ़ेगी, और उसको handle करने वाली tech (liquid + immersion cooling) "optional" से "compulsory" बन जाएगी। इसका exposure US में Vertiv, Modine, Comfort Systems, Trane, nVent में है, और India में Blue Star, Voltas, Amber, JCHAC, Kirloskar Brothers (HVAC + thermal + fluid infra) में।
फिर से — यह एक study है, advice नहीं। मैं future में सोचता हूँ, आज के शोर में नहीं। यह theme अब रिकॉर्ड पर है, तारीख़ के साथ, logic के साथ, और ग़लत होने की शर्त के साथ। 1–2 साल बाद मिलते हैं, scorecard tick करने। ❄️🔥
उपयोगी BillCraft Tools
अपने investments और tax को plan करने के लिए free Indian tools:
- Capital Gains Calculator — US/Indian stock sale पर LTCG/STCG निकालो
- SIP vs Lumpsum Calculator — thematic bets में staggered entry plan करो
- Goal-Based Investment Calculator — target amount के लिए plan बनाओ
- Income Tax Calculator 2026 — foreign income/gains पर tax समझो
अगर यह study आपको एक "layer आगे" सोचने पर मजबूर कर गई — चाहे agree करो या disagree — तो इसे share करो। और मेरा silicon photonics वाला blog भी पढ़ना मत भूलना। 🔖