Research Archive
AI Factories: Reframing Infrastructure from Cost Center to Profit Center
All AI factories are data centers, but not all data centers are AI factories. For much of the last two decades, enterprise infrastructure strategy has been grounded in cost control. Data centers were built to support core IT functions—storage, compute, and networking—with an emphasis on efficiency, uptime, and total cost of ownership (TCO). These environments…
Google Surges Ahead on GenAI Momentum and Resilient Capex Strategy Amid Macro Uncertainty
Key Takeaways on Google’s 1Q25 Earnings Sustained AI Momentum Across Products and Infrastructure: Gemini 2.2 is now embedded across 15 products with 500M+ users, while AI Overviews in Search surpassed 1.5B MAUs. Visual and voice-first inputs like Google Lens and Circle to Search are driving deeper engagement and monetization parity with traditional search. Robust…
Google Cloud Next 2025: Ironwood TPU, Agent Toolkits, and Google’s vertical advantage
Key Takeaways: Google Cloud unveiled Ironwood (TPUv7), its first inference-optimized TPU, offering massive leaps in compute (5x), HBM capacity (6x), and power efficiency (2x) over the previous generation. Performance is estimated to be within 5% of an Nvidia B200, despite TPUv7’s single primary compute die design. New Agent Development Kit (ADK) and Agent2Agent (A2A) protocol…
We Do Not Have Enough Compute
We are in a huge AI compute shortage right now. At this very moment! When you hear most people talk about this, it’s generally in regards to model scaling and compute scaling for training. Not enough compute so labs can’t scale up the scaling laws. Right now is different, it’s the Ghibilification of AI, there’s…
A Deeper Look at Grace – NVIDIA’s Custom Arm-based Super Chip
Executive Summary Custom CPU development is essential for maximizing the efficiency and performance of accelerated computing architectures. NVIDIA’s strategic approach to CPU design, specifically its purpose-built Arm-based CPUs, delivers significant competitive advantages by addressing the limitations of general-purpose CPUs in GPU-accelerated environments. This report analyzes how this strategy delivers competitive advantages through improved performance, power…
Nvidia Management on GPU vs. Custom ASIC Debate
The GPU vs. Custom ASIC debate is easily one of the most talked about dynamics regarding AI compute infrastructure. I have dug deeply into the debate in this article here (Understanding the Hyper Scaler Custom ASIC Strategy). There is no doubt this is a central question NVIDIA executives get from investors, so it was helpful…