2026's Undervalued AI Plays: The Networking, Agents, and Healthcare Inflection

The AI investment thesis for 2026 must pivot from the GPU-centric frenzy of the past to a new industrial reality. The build-out has matured from a "GPU Gold Rush" into a massive, interconnected infrastructure project-the "AI nervous system." The primary bottleneck has shifted from raw compute power to the high-speed networking and optical components required to link it all. For investors, this structural shift creates a new set of scalable, high-margin opportunities in specialized providers.

The scale of this new build-out is staggering. The total addressable market for AI data center systems is projected to reach over $1.2 trillion by 2030. This isn't just about more chips; it's about the entire ecosystem that makes them work. The catalyst is the sheer size of hyperscaler clusters, which are now scaling beyond 100,000 compute nodes. At this scale, the speed and efficiency of data movement become the limiting factor, not the processing power of the individual chips. This has triggered a "Connectivity Wall," where the cost and power consumption of networking equipment rival the cost of the GPUs themselves.

This shift is already driving a capital rotation. Companies providing the "plumbing" for these networks are seeing their multiples expand as hyperscalers seek open, scalable solutions to avoid vendor lock-in. The winners are those with deep technical moats in critical subsystems. Broadcom has emerged as the "Essential Architect," with its Tomahawk 6 switching silicon capturing over 80% of the market and a staggering $73 billion backlog for AI-related content. In optics, specialists like Coherent and Lumentum are critical for deploying the 800G and 1.6T modules that form the backbone of these clusters.

The bottom line is a clear market inflection. The era of buying the AI "brain" is giving way to investing in the "nervous system" and the specialized agents and infrastructure that will run on it. The $1.2 trillion TAM by 2030 is a massive, scalable opportunity, but it's being captured by a new generation of industrial providers. For 2026, the focus must be on the high-speed networking, AI agents, and healthcare infrastructure that are now the critical bottlenecks, not the compute that defined the last cycle.

The Growth Investor's Playbook: Undervalued Winners in the Bottlenecks

The AI investment thesis has evolved. The initial "GPU Gold Rush" is giving way to a structural build-out of the "AI nervous system," where the primary bottleneck is no longer compute power but the high-speed networking and optical components required to connect it. For growth investors focused on market penetration and scalability, this shift reveals a new generation of essential, high-margin winners. The playbook is clear: target companies with dominant market positions in these critical infrastructure layers, backed by massive and growing total addressable markets.

The undisputed architect of this new fabric is Broadcom (AVGO). The company has captured over 80% of the high-end Ethernet switching market with its Tomahawk 6 chips, becoming the indispensable "Essential Architect" for hyperscalers building million-GPU clusters. This dominance is reflected in its staggering financials, with AI-related revenue now accounting for over 30% of total sales and a $73 billion backlog. This isn't just a cyclical surge; it's a structural capture of a critical, recurring revenue stream as the industry standardizes on open Ethernet. For a growth investor, Broadcom represents the foundational, scalable layer upon which the entire AI ecosystem is being built.

From architecture to deployment, Arista Networks (ANET) is the primary vehicle. The company's software-driven switches, built on Broadcom's silicon, are the leading platform for deploying this new networking fabric. Arista's visibility is exceptional, with management targeting $2.75 billion in AI networking revenue for 2026. More importantly, the company's total addressable market is projected to exceed $100 billion. This combination of near-term revenue acceleration and a vast, multi-year TAM provides a clear path for sustained growth, making Arista a pure-play beneficiary of the connectivity wall.

The light-speed data transfer required for these clusters is handled by optical component specialists. Coherent (COHR) and Lumentum (LITE) are critical enablers as the industry moves from 800G to 1.6T modules. Coherent, in particular, has secured record bookings for 1.6T optical transceivers, a mandatory requirement for next-generation AI racks. Their role is not commoditized; they are essential suppliers of the high-bandwidth, low-power components that make these massive clusters viable. Their growth is directly tied to the scaling of AI infrastructure, offering a scalable, high-margin opportunity.

Beyond the hardware, the software layer for AI agents is a burgeoning market. Salesforce (CRM) is positioning itself as a key beneficiary through its Agentforce platform. The company's AI revenue is surging, with annual recurring revenue (ARR) for its AI products jumping 330% last quarter to $540 million. At a forward price-to-sales ratio below 5.5, this explosive growth is trading at a valuation that still leaves significant room for market penetration. For growth investors, Salesforce represents a scalable play on the deployment of AI across business functions.

Finally, the healthcare sector presents a massive, underpenetrated market for AI. The AI in healthcare market is valued at $18.0 billion in 2026 and is projected to reach $80.7 billion by 2036. This explosive growth is driven by a fundamental global shortage of health workers, projected to reach 11 million by 2030. AI tools that augment clinical workflows, drug discovery, and administrative tasks are not a luxury but a necessity, creating a powerful tailwind for companies that can capture this scalable market.

The risk/reward here is compelling. These are not speculative bets on unproven technology; they are investments in the essential, high-margin components of a multi-trillion-dollar industrial build-out. The market is still pricing in the transition, leaving room for these scalable winners to compound value as they capture their respective bottlenecks.

The Roadmap to 2026: Key Milestones and Catalysts

The investment case for AI infrastructure hardware hinges on a clear, near-term catalyst: the industry's rapid transition from 800G to 1.6T networking. As hyperscalers deploy AI clusters with over 100,000 compute nodes, the demand for high-speed, high-bandwidth interconnects has surged. This shift is already underway, with companies like Coherent reporting record bookings for 1.6T optical transceivers and Broadcom securing $73 billion in total AI-related backlog. The industry's pivot to 1.6T silicon photonics-based interconnects is the primary driver for 2026, creating a massive, immediate growth opportunity for suppliers of networking and optical components.

Yet this growth is not without friction. A major risk is execution and supply chain constraints. Converting production lines for new HBM and optical components takes months, creating potential bottlenecks. This is particularly acute for memory, where SK Hynix controls over half the global HBM market and its ability to scale new generations is a key constraint. The risk is that demand for these critical components will outstrip supply during the ramp-up, limiting the ability of hyperscalers to fully deploy their AI clusters and, by extension, the revenue growth for suppliers.

The primary metric to watch is quarterly revenue growth and backlog visibility for networking and optical component suppliers. For companies like Broadcom and Coherent, the trajectory of their AI-related backlogs-Broadcom's $73 billion total AI backlog and Coherent's record bookings extending more than 12 months-will confirm the scalability of their TAM capture. Strong, sequential growth in these figures will signal that the industry's transition is proceeding on schedule and that suppliers are successfully converting orders into revenue.

For the AI agent layer, the key watchpoint is enterprise adoption beyond pilot phases. While nearly 90% of organizations are using AI, the McKinsey survey reveals a critical gap: only 23% of respondents report their organizations are scaling an agentic AI system across their enterprises. This figure is the leading indicator for the next phase of AI investment. A significant acceleration in this percentage would signal that the technology is moving from experimentation to core business transformation, validating the long-term TAM for AI agent platforms and the underlying infrastructure they require.

The bottom line is a two-track investment thesis. The near-term catalyst is the 1.6T networking transition, which is already generating massive backlogs. The primary risk is supply chain execution, which could throttle growth. The key metrics are backlog growth for networking suppliers and the enterprise scaling rate for AI agents. Success on both fronts will determine whether 2026 delivers on the promise of the AI infrastructure build-out.