Layer 1: Google’s ‘Switzerland of Compute’ — Price-Maker on TPUs, Reseller of NVIDIA

This analysis is part of Google: The Gravitational Center of the AI Market, a deep dive by The Business Engineer.

The silicon layer determines the cost floor for everything above it. Companies dependent on NVIDIA allocation are structurally price-takers. Google is a price-maker on its own silicon while also being a reseller of NVIDIA’s.

The Dual-Track Position

Google is the only hyperscaler with battle-tested custom AI silicon at scale AND a simultaneous NVIDIA partnership (shipping GB300-based A4X Max instances). This “Switzerland of compute — as explored in the economics of AI compute infrastructure — ” positioning means Google Cloud can serve customers who want TPUs for cost efficiency, GPUs for ecosystem compatibility, or both.

Price-Takers vs. Price-Maker

Structurally price-takers: OpenAI — as explored in the intelligence factory race between AI labs — (100% NVIDIA-dependent, no custom silicon), Microsoft (NVIDIA + exploring Maia, supply-constrained), Meta (NVIDIA + custom MTIA, early efforts). All have their cost floor set by an external supplier, with margin compressed from below.

Price-maker + reseller: Google controls its own cost floor on proprietary silicon and offers dual optionality—can undercut OR upsell.

The Validation

10+ years TPU production at scale. 1M+ TPU chips committed by Anthropic alone. GB300 A4X Max NVIDIA partnership live. A unique dual-track position in the entire market.

Read the full analysis on The Business Engineer →