Bernstein Research Interpretation: For every additional 1GW of AI data center, semiconductor equipment needs to purchase an extra 8 billion dollars, and the sector's valuation may be underestimated.

Written by: Rita

Trends Summary

Semiconductor equipment stocks have dropped by 30% from their highs but are still up 80% since the beginning of the year.

Bernstein provides answers using a bottom-up calculation method. For every additional 1GW of AI data center computing power, an extra $8 billion is needed for equipment purchases. Assuming an addition of 50GW per year, cumulative equipment spending from 2027 to 2029 will exceed $700 billion, with annual WFE heading towards $300 billion.

This number carries significant weight. The current market pricing for equipment stocks implies an annual WFE spending of only $120 billion to $150 billion, which is exactly half. If the pace of AI construction remains unchanged, current equipment stocks are not only not expensive, but may actually be very cheap.

Bernstein is generally optimistic about the equipment sector, particularly favoring Applied Materials AMAT. More than half of the incremental wafer demand comes from DRAM and HBM, and Applied Materials has the largest exposure in this field. Applied Materials, Lam Research, KLA, ASML, Tokyo Electron, Kokusai, and Lasertec all outperform the market, while Screen remains flat.

$8 Billion Equipment Bill Behind 1GW of Computing Power

Bernstein breaks it down in detail. A Vera Rubin rig consumes 65 wafers, covering logic, HBM, DRAM, and NAND. There are also servers, CPUs, supporting memory, and storage outside the rig, all contributing to incremental demand.

Overall, for every additional 1GW of annual computing power, 46,000 wafers per month of capacity are needed. Half of this is DRAM and HBM, 20% is NAND, and 10% is advanced logic. This translates to exactly $8 billion in equipment investment.

This figure does not account for the replacement of old equipment. From 2027 to 2029, there will be a batch of old computing power that needs to be updated, meaning actual demand will only be higher.

50GW Scenario: Annual WFE Spending Soars to $300 Billion

By 2030, if an additional 50GW of computing power is added each year, surpassing the 2026 baseline, cumulative WFE spending over three years will exceed $700 billion. Additionally, the baseline demand unrelated to AI is about $120 billion per year, pushing the annualized WFE from $200 billion to nearly $300 billion.

What if the pace is even faster? 75GW or 100GW, $300 billion would just be the starting point.

The market's current expectations are far off. The consensus implies annual WFE spending of only $120 billion to $150 billion, which is more than double the shortfall. If Bernstein's baseline scenario materializes, there is significant room for valuation recovery in equipment stocks.

Calculating: Are Equipment Stocks Expensive Now?

Bernstein performed three calculations to directly compare current valuations.

50GW Scenario. Applied Materials' EPS is projected to rise from the consensus of $18.7 in 2028 to $24.3, a 30% increase. By 2029, it will rise to $30.6, a 60% increase. This corresponds to a decrease in price-to-earnings ratio from 23 times to 15 times, dropping to just 11 times in 2029.

75GW Scenario. The EPS in 2028 goes up to $34.2, an increase of over 80%. By 2029, it reaches $46.7, more than doubling. The price-to-earnings ratio is 11 times in 2028 and 8 times in 2029.

100GW Scenario. The EPS in 2028 increases to $44.7, more than doubling. By 2029, it hits $62.4, more than doubling again. The price-to-earnings ratio is 8 times in 2028 and 6 times in 2029.

Lam Research and KLA show similar elasticity. The conclusion is straightforward. As long as the AI construction continues, current equipment stocks are much cheaper than the market perceives.

Why Applied Materials?

In the incremental wafer demand, DRAM and HBM account for 55%. Applied Materials has the highest exposure in the DRAM and HBM equipment sector, which is the core reason why Bernstein favors it.

The ratings for other targets remain unchanged. Lam Research, KLA, ASML, Tokyo Electron, Kokusai, and Lasertec all outperform the market, while Screen maintains market performance.

Trends Perspective

The market knows that AI needs to build data centers, data centers need chips, and chips need equipment. But no one has seriously calculated how much equipment is needed, corresponding revenue, and the corresponding valuation.

Bernstein has figured it out. 1GW corresponds to an $8 billion equipment cost, 50GW corresponds to $300 billion WFE, and corresponds to equipment stocks with a 15 times price-to-earnings ratio. However, the current market pricing implies $150 billion WFE and a price-to-earnings ratio of over 20 times.

The gap of double is the expectation gap.

Risks are very real: will the construction pace slow down, will equipment capacity keep up, will customers cut orders during a downturn, each is a variable. The volatility of equipment stocks has always been greater than that of semiconductors themselves.

But one thing is very clear. If the narrative of AI computing power construction continues, the valuation of equipment stocks has not peaked yet. Whether the current pullback is a risk or an opportunity ultimately depends on whether investors believe in the speed of AI construction.

Disclaimer

This article is a compilation and interpretation of the third-party brokerage research report (Bernstein, July 20, 2026) by Trend Research. The ratings, target prices, profit forecasts, and related judgments mentioned in the text are the opinions of the brokerage's analysts, representing only the position of their respective institution, and do not represent the views of Trend Research, nor do they constitute any investment advice.

The market carries risks; investment should be cautious. This article should not be used as a basis for buying or selling any securities. Investors should make investment decisions based on their independent judgment.