The Second Derivative of AI: 3 Non-Chip Stocks Positioned to Capitalize on the Infrastructure Buildout

The central investment question in AI is no longer about who makes the fastest chip. It is about where value will migrate as the initial hardware frenzy subsides. The structural drivers point to a clear answer: the next wave of winners will be the companies building the essential infrastructure and data platforms that are difficult to displace. This is a shift from a monopoly on compute to a gatekeeping role on data and connectivity.

The scale of the opportunity underscores this fragmentation. Bank of America projects the total addressable market for AI data center systems will exceed $1.2 trillion by 2030. Within that colossal pie, AI accelerators are forecast to capture around $900 billion. That leaves a massive $300 billion slice for the supporting technologies that make the compute work. This isn't just a side bet; it's the primary bottleneck. As AI clusters expand, the physical constraints of scaling compute are shifting from raw GPU supply to power access and network interconnectivity. The evidence is stark: data center energy demand is set to surge, with some data centers unable to come online due to a lack of grid access. This creates a new set of winners.

The thesis is one of the "second derivative." While Nvidia and Broadcom are the obvious architects, the unsung heroes are the providers of the critical bits and bolts. For every $100 spent on AI hardware, $15 to $20 goes into networking, and a third flows directly into interconnects. This is where the value migration is happening. Companies like Credo and Astera Labs are benefiting from being the only providers in niche technologies like active electrical cables and high-speed PCIe 6.0 connectivity. Their gross margins, while not in Nvidia's "different galaxy" territory, are currently in the 35%-40% range, a testament to their essential role.

This shift is not without friction. The interconnect players face brutal pricing compression, with transceiver prices falling from $1 per gigabit to $0.50 in just months. Their growth trajectory will never match the accelerator companies. Yet, their stickiness and necessity create a durable, if less glamorous, business model. The real gatekeepers, however, are the software platforms that control the data itself. Companies like Salesforce and Snowflake are repositioning from perceived AI losers to winners because their platforms serve as the "system of record" for customer data. Once data is locked into a platform like Snowflake, it becomes difficult to move, creating a powerful network effect. Salesforce's recent acquisition of Informatica and its AI agent platform, Agentforce, are strategic moves to solidify this gatekeeping role.

The bottom line is a structural reallocation of capital. The initial phase rewarded chipmakers for their monopoly on compute. The monetization phase will reward those who control the flow of data and the infrastructure that connects it. The $1.2 trillion market is not a monolith; it is a complex ecosystem where the most valuable assets may be the ones that are hardest to replicate.

The Three Enablers: Software, Data, and AI Agents

The new AI economy is not built on silicon alone. It is being constructed on three critical software layers: voice interfaces, integrated workflow platforms, and secure data clouds. These are the enablers that will power the next wave of productivity gains and digital transformation. Their business models are distinct, but their growth mechanics are converging on a single, powerful theme: becoming the essential infrastructure for AI agents.

SoundHound AI is betting on voice as the primary interface for the next generation of AI. Its strategy is a direct evolution from its core AI voice technology into a full-fledged agentic platform, accelerated by the acquisition of Amelia. The growth story is compelling: the company's revenue has more than doubled through the first nine months of 2025. This rapid scaling is paired with a clear path to profitability, as the company expects to soon generate positive EBITDA. The P&L impact is clear-top-line acceleration is translating into improving operating margins. Yet the market's verdict is skeptical. Despite the operational progress, the stock trades down -46.07% YTD and has seen a rolling annual return of -52.86%. This disconnect suggests investors are pricing in execution risk or questioning the scalability of its niche, even as the underlying platform gains traction in auto and restaurant voice systems.

By contrast, Salesforce is leveraging its entrenched position as a system of record to become the central workflow layer for AI agents. Its Agentforce solution is not a standalone product but an integrated suite across Slack and Tableau, embedding AI directly into daily operations. The growth metric is staggering: Agentforce's annual recurring revenue (ARR) surged 330% last quarter to $540 million. This explosive uptake validates the model of using an existing, trusted platform as the launchpad for AI. The valuation reflects this strategic advantage. Salesforce trades at a modest forward price-to-sales (P/S) ratio of below 5.5 times, a multiple that implies the market sees the AI growth as a powerful add-on to a durable, cash-generative core business. The P&L impact is a powerful combination of top-line acceleration and a low-cost, high-margin foundation.

Snowflake occupies the third pillar: the secure data source. Its platform is designed to be sticky, with a net revenue retention rate of 125%, meaning existing customers are spending more over time. This stickiness is now being monetized for AI. Its Snowflake Intelligence solution, which allows customers to build AI agents that access data securely, has reached a $100 million AI revenue run rate. The growth is broad-based, with revenue jumping 29% last quarter. The P&L impact here is a classic SaaS story: high gross margins (65.96%) on top of a growing, sticky revenue base. The company's stock performance, up 44.99% YTD, reflects the market's recognition of this dual engine of data and AI.

The bottom line is a clear division of labor. SoundHound owns the voice interface, Salesforce owns the workflow, and Snowflake owns the data. Each is executing a distinct but complementary growth playbook, with their financial metrics telling a story of scaling, margin improvement, and market validation. For investors, the question is not which one will win, but which layer offers the most durable and profitable position in the AI agent stack.

Risks & Guardrails: The Constraints on the Second Derivative Thesis

The investment case for AI software is a second derivative play, betting that the massive capital buildout in AI infrastructure will eventually translate into profitable software platforms. This thesis is not without significant friction. The primary risk is execution. These companies must successfully monetize their AI platforms without significant cost overruns or competitive erosion. SoundHound AI, for instance, is expected to soon generate positive EBITDA, but its path to profitability is a critical test. Similarly, Salesforce's Agentforce solution has seen strong uptake, but its growth must be sustained and profitable to justify the current valuation.

The competitive landscape is a formidable guardrail. These software companies face intense pressure from two directions. First, the hyperscalers-Amazon, Microsoft, Google, and Meta-are building their own AI agent platforms directly into their cloud ecosystems. This gives them a massive advantage in scale and integration. Second, the chipmakers themselves are vertically integrating into software, blurring the lines between hardware and platform. This dual-front competition threatens to compress margins and erode the moats that software companies rely on.

The macro environment adds another layer of vulnerability. The thesis is predicated on continued growth in AI spending, but this cycle is inherently cyclical. The market's rotation into productivity beneficiaries is a guardrail against pure speculation, but it also means these stocks are vulnerable to a macro or sector rotation away from growth. The severe underperformance of the names in this thesis is a stark warning. SoundHound AI is down -46.07% YTD and has a rolling annual return of -52.86%. Salesforce, while showing a recent bounce, is still down -20.41% YTD and carries a rolling annual return of -22.26%. This persistent weakness suggests the market is pricing in significant execution risk and competitive pressure.

The bottom line is that the second derivative thesis is a high-wire act. It requires these companies to win a complex battle against entrenched cloud giants and vertically integrated hardware makers, all while navigating the inevitable ebbs and flows of a capital-intensive cycle. The current price action reflects deep skepticism about their ability to execute. For the breakout to materialize, they must not only demonstrate technical prowess but also prove they can convert that into durable, high-margin business models in a brutally competitive landscape.

Valuation, Scenarios, and Catalysts for 2026

The valuation story for AI software in 2026 is one of stark divergence and potential re-rating. While the sector as a whole has been a laggard, the current price action creates a potential entry point for the second-derivative thesis. SoundHound AI (SOUN) trades at a discount to its 52-week high of $23.50, with its stock down 46.07% year-to-date. Salesforce (CRM) is down 20.41% YTD, and Snowflake (SNOW) is down 9.97% over the last 20 days. This underperformance, juxtaposed with their fundamental positioning, sets up a classic contrarian opportunity. The bullish scenario hinges on 2026 being the year AI monetization accelerates, validating the need for specialized data and software platforms, while the bear case sees continued chip dominance and margin compression in secondary technologies.

The catalysts that could validate this thesis are concrete and imminent. The first is the resolution of a critical physical constraint: power grid capacity. As Microsoft's Satya Nadella noted, the biggest issue we are now having is not a compute glut, but it's power. This energy bottleneck is currently limiting data center builds and could slow the planned pace of deployment. A breakthrough in securing long-term energy capacity or a shift in regulatory policy would be a major positive for the entire AI infrastructure stack, including the software that runs on it.

The second, and more direct, catalyst is the market's first major AI software IPOs. The prediction is clear: OpenAI and Anthropic likely filing S-1s by the end of 2026. These debuts would set a valuation benchmark for pure-play AI software, potentially triggering a re-rating of existing public companies. For SoundHound, Salesforce, and Snowflake, this would be a powerful signal that the market is ready to pay for AI-native growth, not just infrastructure.

The bottom line is a test of conviction against a backdrop of extreme sentiment divergence. The technical setup for Bitcoin, with its defined breakout levels, offers a parallel narrative for momentum transfer. If silver's momentum is a leading signal of a shift in capital flows, then the first major AI software IPOs could be the lagging indicator that confirms the rotation into productivity beneficiaries. For investors, the strategy is to identify the specific catalysts that will bridge the valuation gap and trigger the re-rating. The next major move in AI software is not a forecast; it is a function of observable events in the IPO calendar and the energy policy landscape.