---
title: "Light instead of copper: why optics will become the main driver of the $154 billion AI market growth"
description: "The AI optical network market will soar to $154 billion: copper yields to light in the race for performance. Goldman Sachs forecasts a 45-fold growth in infrastructure cost per computing unit by 2028. 🚀🔦"
date: 2026-05-29T10:28:12.000Z
lang: en
url: https://xab.info/en/posts/light-instead-of-copper-why-optics-will-become-the-main-driver-of-the-154-billion-ai-market-growth
tags: []
publisher: "XAB.info"
---

# Light instead of copper: why optics will become the main driver of the $154 billion AI market growth

![NVIDIA GB300 AI rack network architecture diagram for 2025, showing the transition to optical connections](https://xab.info/media/2026/05/29/optika-vmesto-medi-rynok-opticheskikh-setey-dlya-ii-vzletit-do-154-mlrd/nvidia-gb300-ai-rack-network-summary-2025.webp)

The era when optical networks were considered merely auxiliary infrastructure for data centers is gone forever. According to the latest forecasts by Goldman Sachs, the market for optical solutions for AI clusters will soar to $154 billion. This surge is driven not just by rising demand, but by a fundamental shift in computing architecture: copper connections have hit physical limits, and now light will become the "circulatory system" of artificial intelligence.

Technological gigantism requires new approaches. Hyperscalers and AI developers are striving to unite thousands of chips into a single computing organism. In this race for performance, a critical factor is not only the power of the GPUs themselves or the volume of HBM memory, but also the speed of data transmission between them. This is where copper loses to optics, and analysts predict a ninefold growth in the market for fiber-optic connections.

### Dominance of Vertical Scaling

Within the structure of this future market, two key directions stand out: vertical (scale-up) and horizontal (scale-out) scaling. However, as calculations show, the lion's share of investments—about $106 billion or 69% of the total volume—will go precisely to vertical scaling. This is due to the need to create ultra-dense connections within the AI clusters themselves.

Special attention is drawn to the CPO (Co-Packaged Optics) technology, which will allow the integration of optical components directly into chips. The CPO market is valued at $91 billion, confirming the thesis that optics is ceasing to be a periphery and is becoming the core of the architecture.

### Exponential Growth of Infrastructure Costs

The transition from current systems, such as NVIDIA GB300 NVL72, to future architectures like Rubin Ultra NVL576, will lead to a colossal change in the cost structure. The cost of network infrastructure per computing unit in the scale-up segment will increase 45-fold. This is due to a sharp increase in the density of optical modules: their number in a single computing unit could grow from 216 to 2,500.

In dollar terms, this means a jump from $315 thousand per rack in current configurations to $9.4 million in next-generation systems. The market for transceivers and optical modules in the scale-up segment will grow 13 times, making this niche one of the most attractive for investors.

### Who will win the race for light?

In conditions where optical density becomes the main growth limiter for AI, the winners are not universal players, but highly specialized manufacturers. Analysts are focusing on companies capable of offering solutions with minimal energy consumption and maximum packaging density.

Among the main beneficiaries are Coherent, Lumentum, and Fabrinet. These companies, specializing in the production of optical components and modules, are at the epicenter of the technological revolution. While broad players like Cisco or Nokia may lose ground, manufacturers of silicon photonics and specialized transceivers will gain access to a colossal Total Addressable Market (TAM), which will grow from $15 billion to $154 billion by 2028.

Optics no longer just connects racks—it defines the limits of what is possible for artificial intelligence. Investors are already fixing this trend, redirecting capital into the sector that will become the foundation of the future digital economy.