NVIDIA (NVDA) presents a compelling long-term opportunity with potential for a 40% run toward a $7.5 trillion market cap over the next three years as the foundational hardware provider for global AI compute.

Recent market overreactions in software create an attractive buying entry for Salesforce (CRM), which is protected from AI disruption by its deeply embedded enterprise data governance, security, and complex system-of-record moats.

UiPath (PATH) is well-positioned as an AI orchestration play, serving as the necessary structured middleware that enables enterprises to reliably connect generative AI models to core business processes.

Investors should seek exposure to the open-source AI infrastructure theme, as up to 90% of enterprise workloads are expected to shift toward cost-effective, decentralized models to prevent vendor lock-in and ensure data sovereignty.

UiPath Inc. (PATH)

• UiPath currently generates approximately $1.6 billion in revenue, growing at around 14% year-over-year.
• The company has transitioned from pure Robotic Process Automation (RPA) to an orchestration and automation platform, recently named a leader in Gartner's Business Orchestration and Automation Technologies Magic Quadrant.
• AI models operate probabilistically, meaning they are prone to compounding errors over multi-step tasks. UiPath aims to provide the deterministic "rails" and structured "map of work" required for enterprise reliability.
• The company is deploying "cartographer agents" that interview enterprise subject matter experts to document internal workflows, and coding agents to generate auditable software.
• The bull case centers on enterprises needing an underlying orchestration layer to harness generative AI safely, while the bear case is that frontier models achieve near-zero token costs and full autonomous human-level reasoning.

Takeaways

• UiPath represents a play on enterprise AI orchestration, acting as the deterministic middleware that connects probabilistic large language models (LLMs) to mission-critical business processes.
• While market sentiment has penalized mid-growth SaaS companies, UiPath's strategy to turn enterprise workflows into structured IP provides a defensible position against raw model providers.

NVIDIA Corporation (NVDA)

• Discussion centered around NVDA potentially expanding from current valuation levels toward a $7.5 trillion+ market capitalization over the next three years, with expectations of a potential 40% run.
• NVIDIA's long-term dominance is heavily tied to the proliferation of the open-source AI ecosystem. If closed frontier labs form a duopoly, they are likely to manufacture custom silicon (ASICs), bypassing merchant chip providers.
• The acquisition and support of open-source platforms like Hugging Face serve as a strategic moat to keep model development decentralized and hardware-hungry across a wide range of enterprises.
• While infrastructure buildouts may not be overbuilt on a 10-year timeline, there is a distinct risk that data center and chip supply could be overbuilt for a shorter 3-year window.

Takeaways

• NVDA remains the premier picks-and-shovels asset for AI compute, though sustained multiple expansion depends on a flourishing open-source model ecosystem rather than hyper-concentrated frontier AI labs.

Salesforce, Inc. (CRM)

• System of record platforms like Salesforce face limited immediate risk from "vibe-coding" or basic AI-generated internal tooling.
• Prototype generation with AI is simple, but enterprise production requires complex security, database schemas, connectors, governance, and audit trails that are expensive and difficult to replicate.
• Software as a category experienced an overreaction during recent market pullbacks ("SaaSpocalypse"), creating attractive entry points for high-quality core platforms.

Takeaways

• Deeply embedded enterprise systems of record like CRM have structural moats in data governance and security that protect them against displacement by casual AI app-builders.

Fireworks AI (Private)

• Fireworks AI provides infrastructure and inference for open-source AI models, benefiting from enterprises seeking model optionality and lower compute costs.
• Enterprise operational workloads are expected to shift heavily (up to 90%) toward cost-efficient and open-source models rather than expensive frontier models for daily tasks.
• Growth in this sector is driven by European and regulated industries demanding on-premise deployments and data sovereignty.
• Securing vast compute capacity is the critical operational bottleneck that infrastructure providers must solve to scale enterprise inference.

Takeaways

• Open-source model serving and fine-tuning platforms present strong secular growth opportunities as enterprises diversify to prevent closed-vendor lock-in and retain proprietary IP.