Technical Intelligence Report Date: 2026-01-07


🤼‍♂️ Market & Competitors

[2026-01-07] Steel, Sensors and Silicon: How Caterpillar Is Bringing Edge AI to the Jobsite

Source: NVIDIA Blog

Key takeaway relevant to AMD:

NVIDIA is aggressively capturing the “Physical AI” and industrial edge market by integrating its Jetson Thor hardware with local LLM serving. This creates a high barrier to entry for AMD’s embedded (Versal/Kria) solutions, as NVIDIA is providing a full-stack turnkey solution (ASR/TTS + LLM + Digital Twins) that operates entirely offline.

Summary: Caterpillar has debuted a “Cat AI Assistant” integrated into a Cat 306 CR Mini Excavator, powered by the NVIDIA Jetson Thor edge AI platform. The system enables operators to control machinery and troubleshoot issues using natural language in real-time without requiring a cloud connection.

Details:

  • Hardware Platform: Utilizes NVIDIA Jetson Thor, a dedicated edge AI platform designed for functional safety and high-performance inference in robotics.
  • Language Model: Employs Qwen3 4B, a compact Large Language Model (LLM) optimized for intent parsing and low-latency response generation.
  • Inference Stack: Served locally via vLLM, ensuring high throughput and low latency for on-device interactions.
  • Speech Stack: Uses NVIDIA Riva with Parakeet Automatic Speech Recognition (ASR) and Magpie Text-to-Speech (TTS) models.
  • Safety Features: Implements “E-Ceiling” technology, allowing operators to set physical boundaries (overhead or underground) via voice commands to prevent machine damage.
  • Digital Twins: Caterpillar is utilizing NVIDIA Omniverse and OpenUSD to create digital twins of manufacturing sites to simulate material flow and line changes before physical implementation.
  • Data Integration: Connects to the Caterpillar Helios data platform to provide machine-specific context to the AI assistant.

🤼‍♂️ Market & Competitors

[2026-01-07] From Warehouse to Wallet: New State of AI in Retail and CPG Survey Uncovers How AI Is Rewiring Supply Chains and Customer Experiences

Source: NVIDIA Blog

Key takeaway relevant to AMD:

A significant 79% of retail/CPG respondents prioritize open-source models and software. This trend represents a major opportunity for AMD’s ROCm ecosystem to gain ground if they can offer superior performance-per-dollar for open-source LLMs compared to NVIDIA’s proprietary locked-in stacks.

Summary: NVIDIA’s third annual survey of the Retail and Consumer Packaged Goods (CPG) industry reveals a massive shift from AI experimentation to full-scale production. The report highlights the rise of “Agentic AI” and “Physical AI” as the next frontiers for operational efficiency and revenue growth.

Details:

  • Adoption Metrics: 91% of surveyed companies are actively using or assessing AI; 90% plan to increase AI budgets in 2026 (half of those by 10% or more).
  • Economic Impact: 89% reported increased revenue due to AI, with 30% seeing gains of over 10%. 95% reported cost decreases.
  • Open Source Dominance: 79% of respondents cited open-source models as “moderately to extremely important,” highlighting a shift away from vendor lock-in and a desire for data sovereignty.
  • Agentic AI: 47% are assessing or using “AI Agents”—autonomous systems that handle inventory rebalancing, dynamic pricing, and vendor negotiations. 20% already have these agents active.
  • Physical AI & Robotics: 17% are evaluating Physical AI (in-store robotics) to manage inventory, pricing accuracy, and presentation quality.
  • Supply Chain Resilience: 51% use AI primarily for operational throughput to combat geopolitical instability and labor constraints.
  • Technical Shift: The industry is moving from “vanity projects” to “P&L-focused” deployments, specifically targeting inventory optimization at the individual store level rather than regional levels.