English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Google TPU Ironwood vs Nvidia GPUs: How Custom Silicon and Optical Switching Challenge the AI Chip Monopoly

Forum topic · ✨步子哥 · 2026-01-01

Summary

This analysis examines how Google's custom TPU v7 "Ironwood" chips threaten Nvidia's dominance in AI compute as the industry shifts from model training to large-scale inference. Unlike Nvidia's general-purpose GPUs (such as the B200 Blackwell with 20 PFLOPS), TPUs strip out graphics hardware to focus purely on matrix math, delivering better cost and energy efficiency for running models like Gemini at scale. The article highlights Ironwood's key specs (~4,614 TFLOPS FP8, 192 GB HBM3e, 7.37 TB/s bandwidth) and Google's Optical Circuit Switching (OCS) technology, which uses MEMS mirrors to route laser signals between thousands of chips without electrical conversion, enabling dynamic topology reconfiguration. It also explains how Google's JAX framework and XLA compiler erode Nvidia's CUDA software moat by making code hardware-agnostic. The conclusion: Nvidia remains king of training, while Google and other hyperscalers will dominate the inference era, moving the industry toward a multi-polar compute landscape.

This piece, originally published in Chinese on zhichai.net, argues that the AI compute landscape is fracturing as Google's custom TPU silicon challenges Nvidia's long-standing dominance — the so-called "Nvidia tax."

Key points

Two opposing chip philosophies

  • Nvidia (general-purpose): GPUs like the B200 Blackwell are "Swiss Army knives" — two stitched silicon dies delivering 20 PFLOPS, capable of training LLMs, folding proteins, or rendering games. Maximum flexibility at maximum cost and complexity.
  • Google (specialized): TPUs are "laser-guided scalpels" — pure matrix-math circuits with graphics hardware stripped away. They exist to run Gemini faster and cheaper, not to play Cyberpunk 2077.
  • The paradigm shift: from training to inference

    The author frames the industry's move from the Training Era to the Inference Era as the reason TPUs matter now. Training a model is like school; inference is the graduate doing daily work. Once billions of daily requests hit ChatGPT and Gemini, cost and energy efficiency — not raw single-chip power — become decisive.

    TPU v7 "Ironwood" reference specs:

  • Compute: ~4,614 TFLOPS (FP8) per chip
  • Memory: 192 GB HBM3e
  • Bandwidth: 7.37 TB/s
  • Individually, the B200 still wins. But Google packs thousands of Ironwood chips into Pods, where smaller, cooler, denser, more power-efficient TPUs win on economics. If power is the bottleneck and Google uses ~40% less energy for the same work, physics decides the winner.

    Optical Circuit Switching (OCS): the real breakthrough

    Google's Palomar OCS replaces electronic packet switches with MEMS mirrors that physically steer laser beams from one fiber to another in milliseconds:

    1. A data-carrying light beam exits a fiber 2. It hits a mirror smaller than a grain of rice 3. The mirror rotates to redirect the beam 4. The light lands in the target fiber — no optical-to-electrical conversion

    Benefits:

  • Dynamic topology: changing the network shape (rings, meshes) is a mirror rotation, not a re-cabling project
  • No electronic switch bottleneck: bandwidth is limited only by physics
  • Where Nvidia relies on expensive InfiniBand cabling, Google solves interconnect with optics — so a "hive mind" of slower individual chips can outperform a cluster of stronger ones.

    Cracking the CUDA moat

    Nvidia's deepest lock-in is software: CUDA, the de facto language of AI for 15 years. Google's counter is XLA (Accelerated Linear Algebra), a compiler that translates high-level Python code into machine code for TPUs, GPUs, or CPUs, plus the JAX framework, which relies on compiler-driven optimization instead of hand-tuned CUDA kernels. As major AI libraries become hardware-agnostic, switching costs drop from "impossible" to "manageable" — and capital follows the arithmetic: the same model at 60 units of cost on TPU vs 100 on Nvidia, no rewrite required.

    Conclusion: a multi-polar future

    The author does not predict Nvidia's death. Instead:

  • Nvidia stays king of training — frontier research still demands B200-class brute force
  • Google (and AWS, Meta) dominate the inference factory — where AI becomes a utility and efficiency is everything
  • The closing metaphor: Nvidia built the rocket to the moon; Google is laying the railway to the future. "David didn't defeat Goliath — David became another Goliath."

    References cited in the original post

  • Google Cloud Technical Documentation: TPU v7 "Ironwood" Architecture & Performance Metrics (2025)
  • SemiAnalysis: "The Optical Circuit Switching Revolution: How Google broke the datacenter speed limit" (2024)
  • DeepMind Research: "JAX: Composable transformations of Python+NumPy programs" & XLA Compiler Whitepaper
  • Nvidia Corporation: Blackwell B200 Architecture Whitepaper (2024)
  • Financial Times / Bloomberg: "The Shift from Training to Inference: AI's Economic Turning Point" (2025 Analysis)

Tags

#google-tpu#nvidia#ironwood#inference#optical-circuit-switching#jax#xla#ai-chips

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/176415220