Goldman Sachs Says Optical Networking Poised to Drive AI Boom

Wall Street peers are turning their gaze to the fiber highways that move data between thousands of AI processors. A fresh note from Goldman Sachs argues that optical networking will become the next trillion-dollar pillar of AI infrastructure, reshaping which companies win as the AI economy scales. As of mid-2026, investors are pricing in a shift from compute alone to the networks that carry the results of that compute in real time.

The core idea is simple but powerful: AI systems will spike in data movement well beyond the growth in raw processing power. In practice, that means photonics and high-speed interconnects could outpace traditional copper links in both speed and efficiency, opening a fresh capital cycle for optical component suppliers and module makers.

Optical Networks as the Hidden Engine Behind AI Compute

The note from Goldman Sachs lays out a path where AI clusters no longer rely solely on faster chips; they rely on faster, denser data highways. As Nvidia advances its Vera Rubin family and the Rubin Ultra platforms, the value embedded in networking per computing unit is projected to surge dramatically. The research sketches a landscape where per-unit networking costs rise from today’s levels to multimega-dollar ranges, underscoring a shift in where AI spend is actually happening.

In practical terms, the report envisions a multi-year expansion of the market for optical components, including pluggable transceivers and co-packaged optics (CPO). The growth isn’t just about more fiber; it’s about smarter, denser packaging that minimizes latency, power, and space while handling the next generation of AI workloads. This is the kind of trend that tends to redraw supplier lineups and push newer players into the spotlight.

Goldman Sachs Says Optical: The New AI Infrastructure Bet

Goldman’s thesis is anchored on two pillars. First, AI systems will demand dramatically higher bandwidth and longer reach as models scale across hundreds or thousands of accelerators. Second, optical interconnects offer the best path to meeting those demands without crippling power budgets or heat dissipation. In the bank’s framing, the optical layer becomes the bottleneck and thus the key investment hinge for the AI era. Analysts emphasize that the shift could redefine profitability for suppliers tied to optical modules, CPO, and related photonics ecosystems. As one senior tech equity strategist puts it, the market is transitioning to a regime where the value chain from chip to data highway matters as much as the chips themselves.

To frame the opportunity, analysts note that the addressable market for AI networking could scale materially in the coming years. The note highlights a potential leap from a mid-teens billions base to well over a hundred billions range by the end of the decade, driven by the need to connect sprawling AI clusters with ultra-fast, low-latency links. The exact numbers vary by scenario, but the direction is clear: optical technologies could become a large, durable source of growth for suppliers who can scale with data center upgrades and new AI architectures.

In discussing the theme publicly, market watchers have echoed a refrain that is now common in AI supply chains: the speed of data movement is catching up to the speed of computation. With the Rubin Ultra class of systems on the horizon, the optics segment is positioned to ride a prolonged cycle of upgrading and expansion. This aligns with a broader industry trend toward tighter integration of photonics with computing units to sustain the economics of large-scale AI training and inference.

Lumentum May Be the Biggest Winner

Within the optical supply chain, Lumentum Holdings (LITE) stands out as a pure-play beneficiary of a shift to high-bandwidth interconnects. The company, long a cornerstone of fiber-optic modules and photonics for data centers, may see a substantial uplift as demand for pluggable optics and advanced CPO grows alongside AI deployments. Wall Street notes that Lumentum’s product portfolio aligns closely with the type of high-margin, high-volume optics demanded by next-generation AI networks.

Key points underpinning the Lumentum thesis include:

• Rising demand for pluggable optical modules in hyperscale data centers and AI accelerators.
• Stronger margins on high-performance photonics components as data-center capex remains elevated.
• Broad customer exposure across cloud and enterprise platforms, reducing reliance on a single end-market.

Industry observers emphasize that Lumentum’s execution on supply agreements and capacity expansion will be crucial over the next 18 to 24 months. While the optical market remains sensitive to capital cycles and chip demand, the favorable long-term trend is clear: AI infrastructure needs optical lines that can scale with speed and efficiency.

Market Reaction, Risks, and Real-World Implications

Investors greeted the optical thesis with cautious enthusiasm. Tech-focused funds and AI bulls say the optics cycle could unlock a new wave of stock performance for a subset of suppliers, including those with established data-center footprints and robust manufacturing scale. However, skeptics warn that the optical value chain depends on continued capex from hyperscalers and on geopolitical factors that could influence supply lines for advanced photonics components.

As one market observer puts it, the optical upgrade cycle is a classic case of “build it and they will come” if AI workloads keep growing. Still, the timing of demand, pricing discipline among suppliers, and the pace at which data centers migrate to CPO-enabled architectures will determine which players gain sustained advantage. In a market where chips like GPUs are still central, optics offers the crucial second leg of the stool—without which AI performance gains could stall at the software layer.

Given the crosscurrents, investors should watch several factors: capacity expansions, supplier diversification, the pace of AI training fleet upgrades, and the ability of optical vendors to cross-sell into broader data-center ecosystems. With markets remaining volatile in mid-2026, even a strong thesis on optical networks needs a careful risk assessment that accounts for cycle timings and supply-demand balance.

What Investors Should Watch Next

• Capacity growth at leading optical component makers, including transceivers and CPO suppliers.
• Customer concentration and exposure to hyperscalers’ AI upgrade cycles.
• Technological milestones in data-center interconnects, including advancements in packaging density and power efficiency.
• Regulatory and geopolitical developments affecting global supply chains for photonics.

Throughout the summer of 2026, the thesis that goldman sachs says optical is shaping AI’s next major investment cycle has gained traction among investors hunting for the next long-duration growth story in tech. The emphasis is no longer solely on the raw horsepower of GPUs; it now extends to the speed and efficiency of the networks that move the results of AI workloads around the world.

Bottom Line

The optical networking thesis—anchored by the assertion that AI infrastructure will require faster, more capable data highways—appears to be gaining momentum. If the market follows the path Goldman Sachs outlines, optical components and co-packaged optics could emerge as the defining driver of AI-era profits for the late 2020s. Lumentum, in particular, sits squarely at the intersection of demand and margin potential, making it one of the most-watched names as the optical cycle unfolds. While risk remains in the timing and breadth of adoption, the case for optical networking as AI’s growth engine is no longer an afterthought; it has moved to the center of the investing conversation.