Semiconductor industry analysis: global market, value chain and India's chip opportunity

Every phone, car, server and rice-cooker controller runs on chips, and the companies that make them are no longer a quiet corner of the hardware trade. In 2026 the industry is on course to sell more in one year than in the two before it combined.

That growth is real, but it is not spread evenly. A few firms earn most of the profit, memory prices explain most of the jump, and one island makes most of the advanced chips. For an Indian founder or investor, the useful question is which layer of this industry they can occupy, and what it pays.

01Who this semiconductor report is written for

We wrote it for three readers. The first is a founder building a chip design company in Bengaluru or Hyderabad, who needs to know how the industry pays its suppliers and its designers. The second is an MSME owner asking whether the new plants in Gujarat and Assam need what their factory makes.

The third is an investor deciding which chip businesses, listed or private, deserve capital at 2026 prices. Global figures are in US dollars, because the industry reports in dollars. Indian figures are in rupees. Everything is as of October 2026, and forecasts are labelled as forecasts.

Five earlier pieces on this site answer narrower questions, and we link to them rather than repeat them. This report is the map they sit on.

03Why do chip sales rise and fall in cycles?

Chips have always been a boom-and-bust business. The chart shows it: sales fell 8% in 2023, after the pandemic buying spree, then recovered over two years. Fabs take years to build, so capacity ordered during a shortage tends to arrive in a glut.

Memory swings hardest. It is sold close to spot prices, so its revenue tracks this quarter's balance of supply and demand. WSTS data show memory sales fell 29% in 2023, then rose 79% in 2024 and 39% in 2025.

In 2026 the swing is extreme. WSTS now calculates memory growth at 302% for the year, against 42% for logic chips, the processors and accelerators. Analog chips, which manage power and signals in cars and factories, grow 11%. Sensors grow 2%.

A market growing 108% while sensors grow 2% is not one market. It is a memory price spike sitting on top of an industry growing at an ordinary pace.

A market growing 108% while sensors grow 2% is not one market.

That matters for anyone reading a chip company's results this year. Revenue from a price spike reverses when new memory capacity arrives, as it did in 2023. Revenue from a product a customer cannot buy elsewhere tends to stay. Our piece on why AI is driving the chip boom works through the memory squeeze.

05Which business models make chips, and how does each earn?

Four business models divide the work. Most confusion about the industry comes from treating them as one.

The split dates from 1987, when TSMC was founded to make chips for other companies. Before that, a chip company had to own a fab. After it, a team with a good design and foundry access could compete without one.

Each model carries its risk in a different place. An IDM carries the cost of empty fabs, because idle equipment still depreciates. A fabless firm carries design risk: a chip that fails means paying for another tape-out, the step where final design files go to the foundry.

A foundry carries technology risk, since a process that falls a generation behind loses its best customers. An OSAT carries price pressure from customers who can move volume elsewhere.

Two smaller models sit beside these. Licensors such as Arm sell circuit designs and earn a royalty on every chip that uses them. Design-service firms bill engineering hours, which describes much of the chip design work done in India. Our companion piece on the fabless model explains how designers earn, and why the split won.

07Why does the industry depend on so few firms and places?

Concentration is the industry's defining risk. TrendForce estimates TSMC held a 72.5% share of the foundry market in the second quarter of 2026. Samsung was a distant second at 5.9%, and China's SMIC third at 5.4%.

At the leading edge the grip is tighter still. Industry estimates put Taiwan's share of the most advanced chip output at around 90%. About 77% of TSMC's wafer revenue in that quarter came from 7nm and finer processes, generations only a few fabs in the world can run.

The tools and inputs are concentrated too. ASML is the only supplier of EUV machines. SK hynix, Samsung and Micron make nearly all of the high bandwidth memory that AI accelerators use. SEMI expects equipment spending of US$165.9 billion in 2026, with China, Taiwan and South Korea the largest buyers.

Politics now shapes the market as much as demand does. US export controls restrict sales of advanced chips and chipmaking tools to China. Nvidia's guidance for the quarter to October 2026 assumes no data centre compute revenue from China at all.

None of this changes quickly. A competing leading-edge fab takes the better part of a decade and tens of billions of dollars. That protects the incumbents, and it means a disruption in Taiwan would have no fast substitute.

09Where does India stand in the chip value chain?

India is a large buyer of chips and a small maker of them. IESA, the industry association, puts the Indian market at US$52 billion in 2024-25 and projects US$103.4 billion by 2030. Nearly all of those chips are made abroad, since the first Indian plants began shipping only in 2026.

The government's answer is the India Semiconductor Mission, launched in December 2021 with ₹76,000 crore. It pays up to half the project cost of approved fabs and assembly plants, and states add their own support. By May 2026 it had approved twelve units with investment of about ₹1.64 lakh crore.

The Cabinet cleared a second phase, ISM 2.0, on 15 July 2026, with an outlay of about ₹1.27 lakh crore. It extends support to equipment, materials, design and talent. Reports say support for new silicon fabs falls to about 40% of project cost.

What do the twelve approved plants add up to?

Three things stand out. More than half the money, ₹91,000 crore, sits in one wafer fab that will not run commercially until about mid-2028, starting at older processes such as 90nm. Every unit in production does assembly and test, the stage India chose to enter first. And six of the twelve are in Gujarat.

The plant-by-plant story is in our piece on the Sanand cluster. What the mission has bought so far, and what ISM 2.0 changes, is in our note on the India Semiconductor Mission.

Can India's design talent become Indian-owned chips?

India's strongest card is its engineers. The IT ministry told Parliament in March 2026 that the country has about 20% of the world's chip design workforce. In February 2026 Qualcomm completed a 2nm chip design using its Indian teams.

Most of that talent works for foreign firms, which own what it designs. The Design Linked Incentive scheme pays part of an Indian startup's design cost to change that. By August 2026 the ministry said DLI-backed firms had completed more than 30 tape-outs and raised over US$100 million in venture funding.

Some of that work now runs on real silicon. Aheesa's VIHAAN broadband chip worked on its first manufacturing run, and Netrasemi has tested a 12nm vision chip. These are engineering milestones; production orders come next.

Against the global industry, US$100 million is one mid-sized funding round. That is the honest measure of India's chip product companies today: proven in engineering, early in revenue. Our assessment of India as a chip hub goes layer by layer through what the country owns and what it rents.

11What should Indian founders, suppliers and investors do?

How a chip design startup should choose its market

• Sell a chip, not hours. Design services pay salaries; owned chips build a company's value. If services fund the business, keep a separate product team with its own roadmap.
• Start on a mature process. Energy meters, surveillance cameras, motor controllers and power electronics do not need 3nm. Older processes cost far less to prototype and have spare foundry capacity.
• Win Indian buyers first. Indian electronics makers, utilities and government buyers are easier to reach than global ones. DLI's second incentive pays a share of net sales on deployed designs.

The questions to settle before a first tape-out, from foundry choice to the cost of a second run, are in the fabless model piece linked above. Our grants and schemes team helps founders time DLI claims against their spending.

Where component makers and MSMEs fit in

• Supply the plants. Twelve units need cleanroom construction, ultra-pure water, specialty gases and chemicals, precision parts, tool maintenance and test fixtures. Much of it is imported today, and ISM 2.0 now extends support to equipment and materials makers.
• Expect a long sales cycle. A chip plant buys only from approved suppliers, and approval takes months. Fund the first approval from your existing business, not from new debt.
• Protect your bill of materials. If you buy chips for meters, inverters, phones or auto parts, 2026 memory prices have already raised your costs. Qualify a second source, and add a price review clause to customer contracts.
• Price in slow payment. Large plants and their contractors pay on long cycles. Build that working capital into the quote before you sign, not after.

Manufacturers with hardware, electronics or precision engineering skills are the closest fit. Our deep-tech practice works with them on where to enter and how to fund it.

How investors should judge a chip company

Standard multiples mislead here, because earnings swing with the cycle and young companies have none. We start with five questions.

1. Which layer is it in? Design, foundry, assembly, equipment and materials each deserve a different multiple and a different view of risk.
2. Is the margin capability or price? Split revenue growth into volume, product mix and price. Price gains on commodity memory usually reverse when new capacity arrives.
3. What are earnings worth mid-cycle? Value memory and commodity chip makers on average earnings across a cycle, never on a 2026 peak.
4. How concentrated are the customers? A supplier to one cloud buyer, or one chip plant, carries that buyer's spending cycle as its own.
5. What happens without the subsidy? For an Indian plant, model the economics at ISM 2.0 support levels, not phase one's.

A proper due diligence process tests each answer against contracts, order books and receivables, not the pitch deck. For founders raising against those tests, our VC and PE advisory team works on the round itself.

13Frequently asked questions

How much is the semiconductor industry worth?

Global chip sales were US$791.7 billion in 2025, according to SIA. WSTS calculates 2026 at about US$1.66 trillion, after a first half worth US$702 billion. Sales for the first eight months of 2026 passed US$1 trillion.

What is the semiconductor value chain?

It has four stages: design, wafer fabrication, assembly and test, and the equipment and materials that feed them. Different companies usually own each stage. Profit concentrates in design platforms, leading-edge fabrication and lithography tools.

What is the future of the semiconductor industry?

The WSTS calculation puts 2027 at about US$2.1 trillion, driven by AI data centres and memory. We expect growth to continue but memory prices to correct, so revenue will be bumpier than the trend line suggests.

How do you invest in the semiconductor industry?

Pick the layer first, since each has different margins and cycles. In listed markets that means choosing among designers, foundries, equipment makers and memory makers. In Indian private markets it means design startups, assembly plants and their suppliers, judged on silicon, customers and runway.

Is the semiconductor industry growing?

Yes, and fast: sales more than doubled in the first half of 2026. But the growth is uneven. Memory is up about fourfold while sensors are almost flat, so the headline overstates what most chip companies earn.