Nvidia’s exposure to Anthropic and OpenAI is beneficial for the company, says Melius' Ben Reitzes
Summary
- Nvidia's relative valuation compared to Mag-7 peers is historically compelling, despite concerns over Hyperscaler spending.
- Nvidia is effectively hedged against a Hyperscaler capex slowdown through direct exposure to model builders (Anthropic, OpenAI) and the rise of Sovereign AI.
- The "Total Addressable Market" (TAM) for AI is not just IT spending but "Labor" itself; trillions of dollars in labor costs will shift to software and tokens, directly benefiting Nvidia.
- A structural crisis is looming for traditional SaaS companies: The shift from "Per-Seat" pricing to "Consumption/Token" pricing will cause significant earnings misses ("missing a few quarters").
- The current rally in software stocks is characterized as a "dead cat bounce," though critical chip-design tools (Cadence, Synopsys) remain safe.
Trade Ideas
"Relative valuation versus its peers... is really pretty compelling... They have exposure to Anthropic now and OpenAI as well, if they happen to win and the hyperscalers lose, they will get a lot of that exposure." The market fears a slowdown in Hyperscaler (AMZN, MSFT, GOOGL) capex. However, Nvidia has diversified demand. If value shifts from the Cloud Providers to the Model Builders (Anthropic/OpenAI) or Governments (Sovereign AI), Nvidia still supplies the chips. Furthermore, as AI Agents replace human labor, the economic value transfers to "compute" (Nvidia's product), expanding the TAM beyond traditional IT budgets. LONG. Nvidia is the ultimate beneficiary of the "Labor-to-Compute" capital rotation. Hyperscaler spending cuts occur faster than Sovereign/Model-builder demand ramps up.
"He's saying the tools won't go away. Like Cadence and Synopsis tools won't go away." While general B2B SaaS is at risk from AI agents ("AI eating software"), the tools required to design the chips that power AI (EDA software) are indispensable. They are the "picks and shovels" for the hardware layer and are immune to the "seat-to-token" disruption facing general enterprise software. LONG. These are the safe havens within the software sector. Cyclical downturn in semiconductor R&D spending.
Ben Reitzes
Analyst, Melius Research
"Software moving to tokens means we're moving to a consumption model, not a seat model. Are you telling me that all these SaaS guys are going to make that transition... without missing a few quarters? Are you kidding me?" The current rebound in software stocks is a "dead cat bounce." The fundamental business model of B2B software is breaking. As AI agents do the work, companies will stop paying for "seats" (per employee licenses) and pay for "tokens" (compute). This transition will destroy the predictable recurring revenue models of legacy SaaS firms, leading to earnings volatility and multiple compression. SHORT / AVOID. The transition economics are hostile to incumbents relying on seat-based pricing. SaaS companies successfully pivot to consumption models faster than expected.
