Four companies. One year. Seven hundred and twenty-five billion dollars.
That is what Amazon, Microsoft, Google's parent Alphabet, and Meta are collectively spending on Artificial Intelligence infrastructure in 2026 alone — according to projections compiled by analysts tracking their capital expenditure guidance. It is not a typo. It is not a cumulative figure spread over a decade. It is a single year's worth of physical spending on servers, chips, data centres, and the power to run them.
Put it in context: $725 billion is larger than the entire annual GDP of Switzerland. It is roughly 2.5% of America's total economic output — spent by just four companies, on just one technology. And Goldman Sachs estimates cumulative AI capex from these four firms alone will cross $5.3 trillion between 2025 and 2030.
The world has never seen an infrastructure build of this speed, scale, and concentration. And the question that serious investors are starting to ask — quietly, carefully — is whether the revenue will ever match the ambition.
The Hardware Heist Nobody Named Correctly
The financial media has spent three years calling this an "AI rally" or a "tech rally." That framing is imprecise. Look at where the money is actually flowing, and a cleaner description emerges: this is a hardware rally.
Before the pandemic, the Information Technology sector made up a modest slice of the MSCI Emerging Markets Index — the benchmark that tracks equities across developing economies globally. Today, technology hardware and semiconductors account for roughly 40-45% of that same index. Taiwan Semiconductor Manufacturing Company (TSMC) — the Taiwanese foundry that physically manufactures chips for Apple, Nvidia, and virtually every other major chip designer — now accounts for approximately 14-15% of the entire MSCI EM index on its own. That makes it the single largest stock weight in the index in three decades.
TSMC, Samsung Electronics, and SK Hynix together account for over 30% of MSCI EM. Three companies. One sector. Nearly a third of an entire global index.
Meanwhile, traditional IT services and software — the companies building the intelligence that runs on AI rather than the machines that power it — have quietly retreated to their lowest share of the index since the Global Financial Crisis of 2008. The picks-and-shovels trade, at an extreme that would make a 19th-century Gold Rush prospector envious.
What the Trillion-Dollar Number Is Actually Measuring
Here is a nuance that rarely makes it into the headlines. A significant portion of AI capital expenditure figures is nominal — meaning they reflect what is being paid, not what is being received.
GPUs (Graphics Processing Units — the high-performance chips that power AI model training) and HBM, or High-Bandwidth Memory, are the two components where supply has been genuinely constrained. When a small number of manufacturers — primarily Nvidia for GPUs and SK Hynix for HBM — control the supply of the components the entire industry needs, prices go up. Sharply. Persistently.
This means that when a cloud company announces $50 billion in infrastructure spending, a meaningful portion of that number is simply paying inflated prices for the same volume of chips they would have ordered anyway. The actual expansion in AI compute capacity is real — but it is smaller than the dollar figures suggest. Some of what reads as a boom is, in part, supply-chain inflation wearing a technology story's clothes.
The Arithmetic Problem
Now for the uncomfortable question buried inside the extraordinary numbers.
Every dollar spent on AI infrastructure multiplies through the supply chain. Chip designers, foundries, memory makers, data-centre builders, power providers, cooling specialists, and finally the cloud platforms that rent all of it out to businesses. Trace the multiplier through the full chain and the industry will eventually need to generate between $3 trillion and $4 trillion in annual revenue to justify today's investment levels.
Global IT spending — every technology purchase made by every bank, hospital, retailer, government, and factory on earth — is forecast by Gartner at $6.37 trillion for 2026. That figure was revised upward four times this year alone, largely because AI infrastructure is pulling spending forward at an unprecedented pace.
To justify its current investment, the AI industry needs to capture half to two-thirds of all global technology spending. Not a slice of the market. Not a significant share. The majority. The path to profitability is not proven. It is not even defined clearly enough to be tested yet.
This does not mean the bet fails. Transformative technologies have often looked economically incoherent at the moment of maximum investment before delivering returns that made the spending look cheap in hindsight. But it does mean investors who treat AI infrastructure spending as automatic proof of AI profitability are skipping the most important step in the analysis.
Scale vs History: A Comparison
| Era | Sector | Build Timeline | Outcome |
|---|---|---|---|
| 1990s | Telecom / Internet | ~10 years | Bust, then internet era |
| 2000s | Oil & Gas | ~15 years | Overcapacity, price collapse |
| 2020s | AI / Data Centres | ~2–3 years | Unproven — still writing |
The telecom boom also seemed unstoppable. The companies building the fibre cables and switching equipment made extraordinary money — until suddenly, they did not. The infrastructure they built eventually powered the modern internet. But most of the companies that funded it went bankrupt first. Speed does not reduce that risk. It concentrates it.
India: A More Nuanced Story Than the Market Is Telling
Closer to home, the picture for Indian equity investors has become more interesting — and more differentiated — than the broad market commentary suggests.
The headline number first: the Nifty 50, India's benchmark index of the country's 50 largest companies, is currently trading at a trailing price-to-earnings (P/E) ratio of approximately 20.6 times. The 10-year historical average sits at around 23.4 times. That means India's large-cap benchmark is trading roughly 12% below its own historical mean. Not bubble territory. Not even "fair value." By one important measure, large-cap India is the cheapest it has been relative to its own history in years.
That is a story the noise of daily market commentary frequently drowns out.
Where the Danger Actually Lives
The calm in large caps masks turbulence elsewhere. Small and mid-cap stocks tell a very different story.
The Nifty Smallcap 50 is trading at a P/E of around 32.8 times — approximately 18% above its 7-year median. The Nifty Smallcap 250 sits at 34.4 times — again, roughly 19% above its long-term median. The mid-cap indices are even more extended, trading 26-30% above their long-term average earnings multiples.
These are not stretched valuations. They are priced-for-perfection valuations — the kind where everything needs to go right, and nothing is permitted to go wrong. In markets, that is precisely when something goes wrong.
India Market Valuation Snapshot — August 2026
| Segment | Current P/E | vs Historical Average | Assessment |
|---|---|---|---|
| Nifty 50 (Large Cap) | ~20.6x | 12% below 10-yr avg | Attractively priced |
| Nifty Midcap 100 | ~28–29x | 26–30% above average | Elevated — caution needed |
| Nifty Smallcap 50 | ~32.8x | 18% above 7-yr median | Expensive — high risk |
| Nifty IT | ~18.3x | 33% below 7-yr median | Most undervalued sector |
| Financials | In range | Below historical average | Value opportunity |
| Auto / Healthcare | In range | Below historical average | Value opportunity |
That IT number deserves a second look. Nifty IT — the sector that directly builds and sells software — is trading at a 33% discount to its own 7-year median P/E. At the same time that global AI hardware is capturing the market's imagination and multiple, the companies that are likely to eventually monetise AI through software and services are sitting at some of the cheapest valuations they have traded at in years. The irony is almost elegant.
Return Expectations: Do the Math First
A question investors often skip: what should a long-term Indian equity investment actually return from here?
The honest answer is anchored in a simple relationship. Over long periods, corporate revenue broadly tracks the growth of the overall economy — India's nominal GDP (the total size of the economy in rupee terms, including inflation). That number has historically run at 10% to 12% per year, and there is no particular reason to expect it to break dramatically from that range.
The constraint is margins. Indian corporate profit margins are at or near their highest levels in recent history. When margins are already stretched, they cannot expand much further. And margin expansion is how companies grow earnings faster than revenue. Without it, earnings growth — the real driver of long-term stock prices — is capped at roughly the same pace as the economy: 10% to 12% annually.
That is a respectable return. It is not 18%. It is not 25%. Anyone constructing an investment plan around those kinds of numbers should revisit their assumptions.
The Borrowed-Money Problem
There is one more risk worth naming plainly. India's Margin Trading Facility (MTF) book — the total outstanding loans that stock brokers have extended to investors to buy shares on credit — hit an all-time high of ₹1.27 lakh crore in May 2026. That is a 65% year-on-year increase. Three years ago, this number was under ₹25,000 crore. It has grown fivefold.
Add the enormous notional volumes in Futures and Options (F&O) — derivative contracts used to amplify bets on price movements — and the picture of a market with substantial embedded leverage becomes clear. Leverage does not create bear markets. But it is a powerful amplifier. If markets fall for any reason — a global growth scare, a monetary policy shock, a disappointment in AI earnings — that leverage accelerates the descent. The spring stores energy on the way down, not just on the way up.
This is not a prediction. It is a structure worth understanding before you need to.
The One Number That Actually Looks Encouraging
Not everything in India's current economic story is a warning. Private-sector capital investment — for years the missing piece in India's growth narrative — is finally, genuinely showing signs of life.
Bank credit to industry grew at 12.1% year-on-year in January 2026 — a significant acceleration from the low single-digit growth rates of earlier years. Capital goods output (a direct measure of companies buying equipment to expand production) surged 16% year-on-year in April 2026, the strongest category within India's industrial production data. India's largest cement manufacturers have collectively committed $13.5 billion in new plant investment for FY26-FY28, adding capacity that only makes economic sense if demand is expected to grow.
Private investment is the engine that sustains economic cycles after government spending moderates. When it picks up, the multiplier effects work through employment, wages, and consumption. These are early signals, not a completed recovery. But they are real, and they matter.
What It All Adds Up To
The story of global markets in 2026 is ultimately a story about two kinds of uncertainty running in parallel.
Globally, an extraordinary amount of money is being spent, at unprecedented speed, on infrastructure whose revenue case is unproven. History suggests transformative technologies eventually justify their build-out costs — but the timeline and the survivors are never who the initial excitement predicted. The companies physically supplying the hardware have captured the valuation. The companies that will monetise the intelligence it enables are sitting, quietly, at historical discounts.
In India, the market is more internally divided than headlines suggest. Large caps are cheap by their own history. Small and mid-caps are not. A handful of sectors — IT most strikingly — have been left behind by the rally even as the businesses themselves remain sound. And private investment is beginning to add momentum to an economy that needs it.
The investor who understands these distinctions — who knows which part of the market they own, what they paid, and what a realistic return looks like — is positioned very differently from the one riding the momentum. The gap between those two investors tends to widen, not narrow, when markets get complicated.
They are getting complicated.
Start Investing — Platforms We Trust
Using the links below supports VilfinTV at no extra cost to you.
| Platform | Best For | Link |
|---|---|---|
| Kuvera | Direct MF investing, 0% commission — 2000+ funds | Open Account (code: 1T6BH) |
| Zerodha | India's largest broker — stocks, ETFs, MF via Coin | Open Account |
| Dhan | India + US stocks via GIFT City | Open Account |
| IndiaBonds | Fixed income — bonds up to ~14% p.a. | Open Account |
| Wint Wealth | Curated bonds and debt instruments | Open Account (code: 3AC7AF) |
This article is for informational and educational purposes only and does not constitute investment advice. Equity markets carry inherent risk and past performance does not guarantee future returns. Consult a SEBI-registered financial advisor before making investment decisions.