A year ago, suggesting that Dell, Nokia, Cisco, Corning, or Western Digital would become hot AI trades would have been met with skepticism. The market was fixated on GPUs, optical modules, and the most direct layers of compute expansion. Old-guard tech companies were stamped with labels like “slow growth,” “stale narrative,” and “valuation multiple with no spring.” Yet recently, these seemingly unexciting names have staged a remarkable run, forcing a rethink.
The explanation is straightforward: as AI moves from model launches to concrete data‑center build‑out, the market naturally gravitates toward firms with real delivery capability and infrastructure know‑how. It is not nostalgia; it is the recognition that AI is not a single GPU problem but a whole‑system engineering challenge. Models need data centers, inference demands servers, networking, storage and power, and enterprises require complete IT architecture and deployment—hence the starting point for repricing the old guard.
Old Labels, New Roles
In the past, Dell meant PCs and traditional servers, HPE meant enterprise hardware, Nokia meant 5G equipment, Cisco meant legacy networking, Corning meant glass and fiber, and Western Digital and Seagate meant cyclical hard‑drive stocks. Those labels are now being rewritten: an AI data center requires full‑rack servers, liquid cooling, storage, switches, fiber connectivity, data management, power infrastructure and enterprise‑grade delivery. The larger the cluster, the greater the demands on system integration, network throughput, and operational reliability.
Dell is the clearest case. In its latest quarter it posted revenue of $43.8 billion, $24.4 billion in AI orders, and recognized $16.1 billion in AI server revenue. The company raised its full‑year AI server revenue forecast to $60 billion and its total revenue guidance midpoint to $167 billion. The market now views Dell as a kind of “AI factory general contractor,” because its value lies not in making GPUs but in the supply chain, integration and delivery that gets chips into working systems. HPE follows the same logic: Q2 revenue of $10.68 billion rose 40% year‑over‑year, cloud and AI‑related revenue hit $7.71 billion, and the Juniper acquisition added networking capabilities, transforming it into an “AI network + enterprise infrastructure” platform.
Connectivity and Storage: The Hidden Champions
Compute does not live in isolation. Inside the data center, high‑speed interconnection is critical; between data centers, fiber links are essential; and as AI moves to the edge, stronger telecom infrastructure is required. This has brought Corning, Nokia and Cisco back into focus. Corning reported Q1 2026 core sales of $4.35 billion, up 18% year‑over‑year, with optical communications revenue surging 36% to $1.846 billion, driven by Gen AI product demand and long‑term agreements with hyperscale customers. Nokia, meanwhile, is extending its story from 5G to AI‑RAN and 6G—Nvidia announced a $1 billion investment to jointly advance AI‑native wireless networks. As long as AI applications spread to phones, robots and AR/VR, telecom‑infrastructure firms will have a growing narrative. Cisco’s Q3 FY2026 revenue of $15.8 billion rose 12%, and data‑center switching orders jumped more than 40% year‑over‑year. In high‑performance AI clusters, networking is no longer a supporting act but a key determinant of compute efficiency.
Storage is another part of the AI repricing. Western Digital and Seagate are not benefiting from a “disk revival” tale, but from the data explosion that makes high‑capacity, cost‑efficient drives indispensable for cold data, training sets, video archives and logs. As clusters grow denser, storage capacity and cost efficiency become unavoidable infrastructure requirements.
The Limits of Repricing: Not Every Old Story Is New
A broad AI repricing does not mean every legacy company deserves a free pass. Three tests matter: whether revenue is sustainably driven by AI and converts into orders; whether those orders come from long‑term capex of data centers, cloud providers, or enterprise AI deployments; and whether profit quality actually improves, rather than just reflecting a short‑term inventory restock. If revenue grows but gross margins are squeezed, the re‑rating will be limited. The market is not buying a new story; it is asking whether old assets, coupled with new demand, can generate new profits.
In short, AI will not turn every traditional tech company into a growth stock. It will only select those occupying critical infrastructure nodes and capable of turning AI demand into orders, revenue and earnings.
From Valuation Repair to Structural Repricing
The AI rally today is no longer just about “who has the strongest model” or “who has the most GPUs.” As data centers multiply, server companies get repriced; as clusters grow more complex, networking companies get repriced; as fiber connectivity demand balloons, materials companies get repriced; and as AI data keeps exploding, storage companies get repriced. This is why legacy tech firms are back in sight—not because they have suddenly rejuvenated, but because the AI era once again needs their infrastructure. Crucially, this repricing will not be evenly distributed across all old‑guard names. Only those that genuinely enter the capex cycle of data centers and enterprise deployment can move from mere valuation repair to a genuine logic‑driven re‑rating.