Artificial intelligence spending is entering a new phase. For the past three years, investors focused almost exclusively on GPUs because they powered every major AI breakthrough. That story isn’t over, but another bottleneck is emerging.\u00a0 <\/p>\n
Moving data between thousands of processors is becoming just as important as the processors themselves. Goldman Sachs argues that networking — not computing — is the next major investment cycle in AI infrastructure. If that thesis proves correct, companies supplying the optical components that move data at light speed could enjoy years of demand growth.\u00a0 <\/p>\n Among them, Lumentum Holdings (NASDAQ:LITE<\/a>) stands out as one of the purest and potentially highest-upside opportunities. <\/p>\n AI Needs Faster Highways, Not Just Faster Cars <\/p>\n In its April 2026 optical networking report, Goldman Sachs argues that AI clusters are rapidly outgrowing traditional electrical connections. As Nvidia (NASDAQ:NVDA<\/a>) transitions from Blackwell to Vera Rubin and eventually Rubin Ultra systems, networking content per computing unit is projected to expand from roughly $315,000 in today’s GB300 NVL72 architecture to approximately $9.4 million in Rubin Ultra NVL576 systems. <\/p>\n That represents a nearly 29-fold increase in networking value per computing unit. <\/p>\n Even more striking, Goldman estimates the overall addressable market for AI networking will grow from roughly $15 billion during the GB300 cycle to approximately $154 billion by 2028. Optical technologies – including pluggable transceivers and co-packaged optics (CPO) — become the biggest beneficiaries because copper connections increasingly struggle with higher bandwidths, longer distances, and rising power consumption. <\/p>\n Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks \u2014 and Lumentum didn’t make the cut. Grab the names FREE today<\/a>. <\/p>\n To put that into perspective, Goldman projects the market for pluggable optical modules alone could increase about 10-fold, while the broader optical components market expands roughly 13-fold as optics migrate beyond traditional rack-to-rack networking into shorter-reach connections inside AI systems themselves. <\/p>\n Story Continues <\/p>\n 24\/7 Wall St. <\/p>\n Stop betting solely on chips while data highways choke. The next AI gold rush isn’t in processing, but in light-speed networking infrastructure. \u00a9 24\/7 Wall St. <\/p>\n Why Lumentum Could Deliver Outsized Returns <\/p>\n Unlike Nvidia, whose business spans GPUs, networking, software, and complete AI systems, Lumentum is almost entirely leveraged to optical communications. That concentration creates more risk, but also more upside. <\/p>\n Lumentum already supplies lasers, optical engines, and photonic components used in high-speed data center transceivers. Those products become increasingly valuable as hyperscalers upgrade from 800G networking to 1.6-terabit and eventually 3.2-terabit optical connections. <\/p>\n Goldman also expects co-packaged optics to account for roughly $91 billion, or 59%, of total AI networking spending under its base-case 2028 scenario. While widespread CPO adoption still carries execution risk, Lumentum participates in both today’s dominant pluggable market and tomorrow’s emerging optical engine market. <\/p>\n That gives investors exposure regardless of whether the industry moves gradually or rapidly toward CPO. <\/p>\n Compared with diversified semiconductor companies like Broadcom (NASDAQ:AVGO<\/a>) or Marvell Technology (NASDAQ:MRVL<\/a>), Lumentum offers much greater sensitivity to rising optical demand because communications hardware accounts for such a large share of its business. If AI networking spending accelerates anywhere near Goldman’s projections, Lumentum’s revenue growth could outpace that of larger, more diversified peers. <\/p>\n Risks Remain, but the Tailwinds Are Powerful <\/p>\n Granted, Goldman acknowledges its assumptions are optimistic. The report assumes approximately 29% CPO penetration by Rubin Ultra’s rollout, while several independent industry forecasts envision slower adoption extending toward 2030. Manufacturing complexity, packaging yields, serviceability concerns, and hyperscaler qualification timelines could all delay deployment. <\/p>\n There’s also the broader AI spending question. Goldman bases its projections on continued aggressive hyperscale investment. If capital expenditures slow or AI infrastructure spending becomes more disciplined, networking demand would likely moderate as well. <\/p>\n That said, the broader trend appears difficult to dismiss. Every major AI roadmap from Nvidia points toward larger clusters, faster interconnects, and greater optical content. Even if CPO adoption slips, Goldman still expects pluggable optics to experience substantial volume growth. <\/p>\n Key Takeaway <\/p>\n In short, Goldman Sachs isn’t suggesting GPUs are becoming less important. Instead, it is arguing that networking is becoming equally essential because AI systems cannot scale if data cannot move quickly enough between processors. <\/p>\n For investors, that creates an opportunity beyond the familiar AI leaders. Lumentum may not be the safest AI investment — that would likely be Nvidia — but it could be one of the most leveraged to the optical networking buildout Goldman forecasts. If AI networking spending expands from $15 billion to $154 billion over the next several years, few companies offer as direct an investment on that secular trend.\u00a0 <\/p>\n The risks are real, particularly around CPO adoption timing, but the potential reward is equally compelling for investors willing to accept the volatility. <\/p>\n Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks \u2014 and Lumentum didn’t make the cut. Grab the names FREE today<\/a>. <\/p>\n<\/a> luchschenF \/ Shutterstock.com <\/p>\n