Market Call: Nicholas Mersch's outlook on Technology Stocks (May 13, 2026)

Key Concepts

• AI Infrastructure: The physical and logical foundation (hardware, power, data centers) required to support AI development.
• Hyperscalers: Large cloud providers (AWS, Azure, Google Cloud) that invest heavily in AI infrastructure.
• Semiconductor/Hardware Scarcity: The current market dynamic where hardware components are in short supply, giving them significant pricing power.
• Agentic AI: The latest evolution of AI that disrupts workflows by acting autonomously, exponentially increasing compute and token usage.
• Neo-Clouds: Alternative cloud providers (e.g., CoreWeave, Nebius) that specialize in high-performance compute (HPC) and GPU rentals.
• Vertical Integration: Owning the entire stack—from silicon and models to distribution—as seen in companies like Google.
• Power Delivery/Optimization: The critical need to manage energy consumption at the grid, data center, and chip levels.

---

1. The Investment Thesis: "Cash the Checks, Don't Write Them"

Nick Mersch argues for a shift in investment strategy. While investors have historically favored "Hyperscalers" (the ones writing the checks for AI infrastructure), the current opportunity lies with the semiconductor and infrastructure players (the ones cashing the checks).

• Capex Surge: Market expectations for 2026 capital expenditure (capex) have ballooned from $350 billion to over $1 trillion, with the vast majority flowing into the hardware and infrastructure layer.
• Scarcity vs. Abundance: Mersch advises investors to "invest in scarcity and avoid abundance." Software is becoming abundant and easily disrupted, while hardware components remain scarce and highly valuable.

2. The Semiconductor and Hardware Landscape

• TSMC (Taiwan Semiconductor Manufacturing Co.): Described as the "foundational layer" and a "bottleneck monopoly." They manufacture chips for NVIDIA, AMD, Apple, Google, and Microsoft. With gross margins of 67% and operating margins of 58%, they possess immense pricing power.
• ARM Holdings: Owns the "blueprint" (instruction set) for chips rather than manufacturing them. As cloud companies design their own AI server chips to reduce reliance on NVIDIA/Intel, they are almost universally choosing ARM architecture.
• Micron (MU): A "scarcity trade" within the AI space. Micron produces High Bandwidth Memory (HBM), which is sold out through 2026. Mersch argues this is a structural shift rather than a cyclical one, as AI agents require significantly more memory than traditional computing.
• Lumentum: Focuses on "scale-out" technology—the fiber optic lasers that allow thousands of GPUs to communicate. NVIDIA has invested $2 billion in Lumentum, signaling its critical role in the ecosystem.
• Vicor: Specializes in power conversion modules that sit on circuit boards. As AI chips draw up to 1000W, efficient power delivery is a major bottleneck. Vicor is a key supplier for Cerebras, a company building "wafer-scale" AI chips.

3. Software and "The Penalty Box"

Mersch maintains a cautious stance on traditional software companies, noting they are currently in the "penalty box."

• Disruption: AI is making software "abundant." Systems that were once considered "systems of record" (hard to rip and replace) are now being challenged by AI-driven alternatives that can be deployed in days.
• Contract Churn: As contracts come up for renewal, companies are opting for shorter-term agreements, signaling potential revenue instability for legacy software firms.
• Specific Mentions:
  • OpenText: Labeled as "too hard" due to its position in the center of AI disruption and leadership changes.
  • Adobe: Facing pressure as generative AI models (like those from OpenAI) democratize image and asset creation, threatening Adobe’s core creative vertical.
  • Constellation Software: Despite strong historical performance, Mersch warns that their "organic revenue growth" is in decay, making them vulnerable to AI-driven competition.
  • Palantir: An exception in the software space. Mersch owns it because of its "forward-deployed engineers"—a go-to-market strategy where they embed staff to help clients implement AI.

4. Data Centers and Energy

• Neo-Clouds (CoreWeave vs. Nebius): These companies act as "intelligence farms." Mersch prefers Nebius over CoreWeave due to better balance sheet management, though both are benefiting from the massive demand for GPU compute.
• Emcor: An electrical and mechanical contractor essential for building data centers. Mersch highlights the labor shortage in "data center alley," where specialized tradespeople are in high demand.
• Digital Realty: A data center REIT. While stable, Mersch views it as a "landlord" play that lacks the high-growth "acceleration" metrics of the Neo-Cloud providers.

5. Emerging Technologies: Quantum Computing

• Xanadu Quantum Technology: A Toronto-based company using photonics (light) to solve compute problems. Mersch describes this as a "binary event" or a "call option"—it is currently in R&D, but if it achieves commercial viability, it will represent a "before and after" moment for the entire computing industry.

Synthesis and Conclusion

The core takeaway is that the AI revolution is a hardware-first event. The massive influx of capital is currently being captured by the companies that build, power, and cool the physical infrastructure of AI. Investors should prioritize companies that own the "bottlenecks" (TSMC, Micron, Lumentum) and those that provide the physical build-out services (Emcor). Conversely, traditional software companies face a "prove-me" period where they must demonstrate that their products are not easily replaced by autonomous AI agents. Mersch emphasizes that in this environment, distribution and vertical integration (as seen in Google or Palantir) are the ultimate competitive advantages.