2025-11-28_tpuv7-google-takes-a-swing-at-the::TPU7-0020

① SA Source

Context Before

Now that Google has gotten their act together on TPU and is selling them externally for people to put in their own datacenters, what are the implications on Nvidia’s business? Does Nvidia finally have a legitimate competitor that will put its market share and margins at threat? Behind the paywall, we will share our thoughts on what this means for Nvidia as well as reveal more about the TPU roadmap.

While Ironwood is a proper competitor to Blackwell, Nvidia once again hits back with Vera Rubin. Vera Rubin will deliver huge performance uplifts across compute, memory and network with TPU v8 seeing much smaller improvements.

Evidence

The 8th generation TPUs will be available in 2027 and will be competing against Nvidia Vera Rubin. There will be 2 versions of TPU v8 but this will be different from the “P” (full) and “E” (lite) SKUs that featured in the then 4th and 5th generations. Rather, it is a dual track with one SKU co-designed with Broadcom (TPU 8AX codenamed “Sunfish”) and one SKU co-designed with MediaTek (TPU 8X codenamed “Zebrafish”).

Context After

TPU 8AX is very similar to Ironwood, even staying on the same N3E logic node with a similar layout with 2 compute die, 1 I/O chiplet, and 8 stacks of HBM3E. The memory is upgraded to 12-high stacks from 8-high and with the highest pin speeds of 9.6Gbps from SK Hynix giving it over a ~30% boost in memory bandwidth vs TPU v7.

TPU v8X only has a single compute die, one I/O die and 6 stacks of HBM3E 12-high. The compute die for TPU v8X will be on N3P, not N3E. The strategy behind Google choosing to partner with MediaTek is to reduce the margin that they pay to their silicon design partners or specifically Broadcom. Broadcom charges Google for the whole System in Package which they stack a healthy margin on. This includes HBM.

② Atomic Claim

TPUv8 預計 2027 年推出雙軌版本:Broadcom co-design 的 8AX/Sunfish 與 MediaTek co-design 的 8X/Zebrafish。

  • Epistemic Mode: EXPECTED
  • Mapping Status: PARTIAL

③ Semantic Frame

{
  "frame_type": "NARY_RELATION",
  "participants": [
    {
      "node": {
        "id": "02_companies/GOOG",
        "label": "Google"
      },
      "role": "vendor"
    },
    {
      "node": {
        "id": "02_companies/AVGO",
        "label": "Broadcom"
      },
      "role": "partner"
    },
    {
      "node": {
        "id": "02_companies/MediaTek",
        "label": "MediaTek"
      },
      "role": "partner"
    },
    {
      "node": {
        "id": "04_knowledge_base/Vera Rubin",
        "label": "Vera Rubin"
      },
      "role": "competitor"
    }
  ],
  "qualifiers": {
    "condition_text": "8AX Sunfish and 8X Zebrafish",
    "numeric_mentions": [
      "2027",
      "2"
    ],
    "temporal_mentions": []
  },
  "relation_type": "TPUV8_DUAL_TRACK"
}

④ Canonical Entity Mapping

RoleSurface LabelCanonical Target
vendorGoogleGOOG
partnerBroadcomAVGO
partnerMediaTekMediaTek
competitorVera Rubin04_knowledge_base/Vera Rubin

⑤ Human Review

請在 Properties 逐項確認:

  • 原文 → Atomic Claim 是否忠實
  • Atomic Claim → Semantic Frame 是否忠實
  • Canonical Entity mapping 是否正確
  • Epistemic mode 是否保留原文語氣
  • 最後選擇 review_action

Review state

Markdown 內文不是正式 approval。只有 Apply bridge 寫入的 Decision Ledger event 才是正式決策。