The "proportion" of AI chips continues to increase, Goldman Sachs is optimistic about the growth of optical modules, focusing on the convergence of valuation differences between first and second-tier manufacturers

With the rapid development of AI technology, the "ratio" of AI chips to optical modules continues to rise, becoming an important engine driving the growth of the optical module industry.

According to news from the Chasing Wind Trading Desk, Goldman Sachs analysts Jin Guo and Allen Chang stated in their latest research report that the ongoing trend of increasing the ratio of AI chips to optical transceivers will significantly enhance the long-term resilience of the industry.

Goldman Sachs has raised its sales forecast for 800G optical transceiver modules for 2025 and 2026 to 19.9 million and 33.5 million units, respectively, an increase of 10% and 58%, and expects the market size to grow by 60% and 52% in dollar terms from 2025 to 2026.

At the same time, Goldman Sachs has raised its net profit forecasts for covered optical module manufacturers for 2025-2027 by as much as 42%, with target price increases ranging from 23% to 82%, demonstrating strong confidence in the industry's prospects.

Behind this optimistic forecast is the significant increase in the ratio of AI chips (such as GPUs and ASICs) to optical modules. The ratio has risen from 1:3 for the H100 to 1:4.5 for the B300, and even reached 1:8 in certain ASIC architectures, directly driving the demand for high-speed optical modules in data centers.

The report points out that optical module-related stocks have seen a significant rebound since the low in April, and current valuations remain attractive, especially as the valuation gap among leading manufacturers is expected to further narrow.

Strong Support for Optical Module Demand Driven by Ratio Increase

The increase in the ratio of AI chips to optical modules is the core driving force behind industry growth.

The report shows that based on reference designs released by manufacturers, the ratio of GPU/ASICs to optical transceivers is showing a clear upward trend.

For example, the ratio for the H100 and GB200 is about 1:3 (each GPU requires three 800G optical transceiver modules), while the B300 has increased to 1:4.5, and certain ASIC training clusters even reach 1:8. This change stems from the higher bandwidth requirements of new chips and the upgrade of network architecture from 400G to 800G and even 1.6T.

The report emphasizes that this trend means that even if the sales growth of AI chips slows, the demand for optical modules will continue to grow due to the increase in the ratio, providing stronger cyclical resilience for the industry.

In terms of specific data, Goldman Sachs expects that sales of 800G optical transceiver modules will reach 19.9 million and 33.5 million units in 2025 and 2026, respectively, an upward revision of 10% and 58% from previous forecasts; sales of 1.6T modules are expected to reach 7 million units in 2026, a slight increase of 2% from old forecasts.

Regarding market size, the report predicts that the total market value of optical modules will reach approximately $12.73 billion and $19.37 billion in 2025 and 2026, respectively, representing year-on-year growth of 60% and 52%, indicating a strong growth trajectory.

Leading Manufacturers May Converge in Valuation, Second-Tier Manufacturers Welcome Demand Overflow Opportunities

The report indicates that over the past two years, the market's focus on the optical module industry has gradually shifted from the growth intensity of 2025 to the sustainability of 2026 and 2027. Even though the market has recently priced in strong expectations for 2026, investors still tend to value optical module stocks at historically low price-to-earnings ratios (10-15 times) to hedge against cyclical risks However, Goldman Sachs believes that the "cross-cycle" growth potential brought by the increase in allocation ratios may support higher valuation levels. Taking Zhongji Innolight and Eoptolink as examples, both currently have price-to-earnings ratios of 15 times and 12 times for 2026, respectively, which are below one standard deviation of the historical average and at a low point.

The report further points out two major investment themes going forward, namely the convergence of valuations among first-tier manufacturers and the potential benefits for second-tier manufacturers from demand spillover.

The report predicts that Zhongji Innolight and Eoptolink will have similar profit scales and growth rates from 2025 to 2027, but Eoptolink's current price-to-earnings ratio for 2026 is about 20% lower than that of Zhongji Innolight, with a market value only 75%-80% of the latter. Goldman Sachs expects this valuation gap to further converge after the second quarter earnings reports are released, especially if Eoptolink's profits continue to approach or exceed those of Zhongji Innolight, or if its 1.6T product progresses smoothly.

Secondly, the surge in 800G demand has led to capacity constraints among first-tier manufacturers, and some orders may spill over to second-tier manufacturers.

The report points out that HGTECH, as a potential beneficiary, is expected to enter the mass production phase in the second half of 2025, leveraging its advantage of over 100,000 units of monthly production capacity at its Thailand base and the progress of 800G product testing with American customers.

Goldman Sachs estimates that if HGTECH successfully secures orders from American customers, its net profit for 2026 could be adjusted upward by 5%-24% compared to the baseline forecast, although this contribution is not currently included in the forecast. The upcoming second quarter earnings report will be a key observation point