The AI Supercycle Isn't Slowing Down -- It Just Shifted. These 2 Stocks Are Riding the Next Wave.

The article argues that AI’s next growth wave is shifting toward physical AI and agentic AI, with Nvidia and Palantir positioned as early leaders. Nvidia said physical AI contributed over $6 billion in revenue last fiscal year and could add "hundreds of billions" long term, while Palantir signed $4.3 billion in Q4 2025 contracts, up 138% year over year, versus 70% revenue growth to $1.4 billion. The piece is fundamentally positive for both stocks but is largely commentary rather than new company-specific disclosure.

Analysis

The market is still pricing AI as a single-wave capex story, but this piece reinforces the real second derivative: the mix is shifting from training spend to deployment spend. That matters because physical AI and agentic AI are much stickier workloads once embedded into workflows and machines, which should extend revenue durability for the incumbents that control the software stack and the physical-inference layer. The winners are likely to be the companies that sit closest to proprietary data, orchestration, and high-frequency inference, while generic application vendors and undifferentiated infrastructure names face margin pressure as the ecosystem commoditizes.

Nvidia’s bigger implication is not just continued accelerator demand, but a broader pull-through effect across networking, memory, power, and robot hardware supply chains. If physical AI scales as described, the bottleneck shifts from model performance to real-world reliability, which tends to favor companies with integrated ecosystems and penalize narrow point solutions. That creates a multi-year tailwind for upstream suppliers, but also raises the risk that customer concentration and in-house silicon efforts eventually cap upside in the outer years.

Palantir’s setup is more asymmetric because agentic AI monetization can expand through software layers with relatively little incremental capital intensity. The key hidden variable is enterprise conversion: if productivity gains are real, budget owners will fund agents out of headcount avoidance rather than new IT budgets, which shortens sales cycles and supports faster adoption than traditional software upgrades. The main risk is valuation multiple compression if growth normalizes before backlog converts, especially if the market decides agentic AI is a feature rather than a standalone category.

Consensus is likely underestimating how much of this is a platform war, not a thematic trend. The near-term trade is still momentum-positive, but over 6-18 months the key question is who owns distribution and data rights once every major software vendor ships an agent layer. That is where the most interesting relative-value opportunities will emerge: long the infrastructure/platform monopolies, short the incremental wrappers that depend on them.

Key Decisions for Investors

• Long NVDA on 6-12 month horizon; use weakness after broader AI-capex rotations to build exposure. Risk/reward remains favorable as physical AI broadens TAM, but trim if evidence emerges that customer in-sourcing or custom ASICs are displacing general-purpose accelerators faster than expected.
• Long PLTR for 12-18 months; consider a call spread structure to express upside while limiting valuation risk. The catalyst is backlog conversion into recurring agent deployments, but the position should be reduced if revenue growth decelerates below the current market-implied trajectory.
• Pair trade: long NVDA / short a basket of non-differentiated AI software beneficiaries over 3-6 months. The thesis is that the market will eventually reward control of inference and infrastructure more than surface-level AI branding.
• Avoid chasing GOOGL, UBER, TSLA, and CAT as direct AI monetization proxies unless there is a clear productization checkpoint. Their optionality is real, but the monetization path is longer and more execution-dependent than the article implies.
• For tactical traders, buy NVDA or PLTR on post-earnings or macro-driven pullbacks rather than strength. The best entry is when AI enthusiasm dips but enterprise and infrastructure order flow remains intact; that usually offers the best 2:1 to 3:1 upside-to-drawdown setup.