Nvidia

Thesis

• The business: An AI data center systems provider and a key HPC acceleration enabler.
• Moat: Own the architecture + CUDA, system-level vertical integration
  • scale-up networking, pre-purchased supply, software/developer installed base, financial capacity.
  • Leaking moat: scale-out networking (to Ethernet) and CUDA (to AMD, Broadcom, for customers with their own kernel engineers).
• Growth drivers: data center compute volume and content per rack, network attach, annual architecture cadence
• Margin drivers: mix, inventory write down, memory price.
• Risks:
  • Compounding risk (revenue/inventory/investments) as Nvidia increasingly finances the demand for its own product.
  • Asymmetric commitment risk: Customers on cancellable purchase orders, while Nvidia holds non-cancellable commitments.

Broadcom vs Nvidia

Reports

• NVIDIA Business Overview
• NVIDIA — Bull & Bear Memo
• NVDA Coverage File
• Broadcom vs Nvidia Peer Duel

Reference

Nvidia (NVDA) Part 2 of 2 Nvidia's long-term bull case: AI is going to be ubiquitous and requires GPUs for training and inference. Nvidia makes GPUs with the most complete software stack and developer buy-in. Its AI platform is integrated with all the hyperscale clouds, runs behind corporate firewalls, and tailored to various industries. The company is foundational to our AI future.