Tenstorrent Blackhole
Tenstorrent
15
Open RISC-V AI hardware
- Version
- Blackhole
- Cost
- from $999/unit
- Model
- Tensix cores
Google TPU Ironwood
Google · #2 most active of 18 in Chips & Compute
v7 TPU pods scaling to 9,216 chips, available through Google Cloud and used for Gemini.
50
Trust score · Mixed
3 data points checked · 2026-09-11
Current version
Ironwood (7th generation)
Entry cost
from $5.4/unit
Changes / 30d
10
The short answers · verified September 16, 2026
- What is the latest version of Google TPU Ironwood?
- The current shipped version of Google TPU Ironwood is Ironwood (7th generation), as of September 16, 2026.
- How much does Google TPU Ironwood cost?
- Google TPU Ironwood starts at $5.4 per month on its cheapest paid tier, as of September 16, 2026.
- What AI model does Google TPU Ironwood use?
- Google TPU Ironwood runs primarily on Gemini.
Source: Google — cloud.google.com · Reusable under CC BY 4.0 — cite Tomorrow
Capabilities
- 9,216-chip pods
- inference optimised
- liquid cooled
- JAX + PyTorch/XLA
- Large-scale training
- Reasoning and inference
- 9,216 liquid-cooled chips per pod
- 42.5 ExaFlops performance
- AI Hypercomputer integration
- JAX support
- PyTorch support
- vLLM support
- 42.5 ExaFlops peak compute
- 4X better performance per chip over Trillium
- AI Hypercomputer architecture
- Liquid-cooled chips
- JAX native support
- PyTorch native support
- Native JAX & PyTorch support
- vLLM engine support
- OpenXLA compiler integration
- Engineered for large-scale training, reasoning, and inference
- Native support for JAX and PyTorch
- AI training
- AI inference
- ML Training
- ML Inference
- GKE support
- large-scale training
- reasoning
- inference
- liquid-cooled chips
- TPU7x
- 9216 chips per pod
- 42.5 ExaFlops
- Reasoning
- Inference
- 9216 liquid-cooled chips per pod
- AI Hypercomputer
- OpenXLA compiler
- Frontier training
- Large-scale inference
- Multi-step reasoning
- OpenXLA support
- liquid-cooled
- PyTorch
- JAX
- vLLM
- OpenXLA
- MaxText
- Tunix
- GKE
- Liquid-cooled
- ML training
- JAX Support
- PyTorch Support
- vLLM Support
- MaxText blueprint
- Tunix library
- Training
- JAX supportvLLM support
- Google Kubernetes Engine (GKE)
- Tensor Processing Units
- training
- 9,216 chips per pod
Pricing
On-Demand (us-central1 Iowa)
Per chip-hour
$12
DWS Flex-start price (us-central1 Iowa)
Per chip-hour
$6
DWS Calendar Mode price (us-central1 Iowa)
Per chip-hour
$8.4
1-year Commitment (us-central1 Iowa)
Per chip-hour / monthly billing equivalent
$8.4
3-year Commitment (us-central1 Iowa)
Per chip-hour / monthly billing equivalent
$5.4
On-Demand (europe-west2 London)
Per chip-hour
$13.2
DWS Flex-start price (europe-west2 London)
Per chip-hour
$6
DWS Calendar Mode price (europe-west2 London)
Per chip-hour
$8.4
Entry price over time
Geek mode
Models under the hood
- disclosed
Gemini
Google
- disclosed
Ironwood
Google
- inferred
Ironwood (TPU v7)
Google
- inferred
Ironwood TPU
Google
- inferred
Ironwood TPU (7th Generation)
Google
- disclosed
TPU v7
Google
- inferred
TPU7x
Google
- disclosed
TPU7x (Ironwood)
Google
- disclosed
XLA compiler
Google
Context window
—
Public API
yes
Multi-model routing
no
Stack signals
- JAX
- PyTorch
- vLLM
- OpenXLA
- MaxText
- Tunix
- GKE
Shared model stack
Other tracked products running on the same foundation models — a quick read on how much of the catalog moves when one of these models changes.
- GeminiGoogle · 13 products
- IronwoodGoogle · 1 product
- Ironwood (TPU v7)Google · 1 product
- Ironwood TPUGoogle · 1 product
- Ironwood TPU (7th Generation)Google · 1 product
- TPU v7Google · 1 product
- TPU7xGoogle · 1 product
- TPU7x (Ironwood)Google · 1 product
green disclosed · amber inferred · grey unattributed. Aliases are folded into one model; provider concentration counts only models with an established vendor.
Reported scale
Reference data. Each figure is whatever the source actually said — weekly users, downloads, revenue run-rate — with its own definition and date. These are not comparable between tools and are never used to rank anything.
No public usage figure on record for this product.
Public market scorecard
Alphabet investor relationsGoogle trades as NASDAQ: GOOGL. Figures below are as reported for Q2 2026 on Jul 23, 2026; market levels are the close on Aug 6, 2026 and are a reference marker, not a live quote.
Q2 2026 revenue
$103B
+15% YoY
- Google Cloud
- ~$16B
- +30% YoY
- Gemini app users
- ~950M MAU
- company stated
Also in Chips & Compute
All Chips & ComputeCerebras WSE-3
Cerebras
8
Wafer-scale inference and training
- Version
- WSE-3
- Cost
- Free tier / developer tier pricing available (contact/partner integrations)
- Model
- Codex-Spark
NVIDIA Blackwell
NVIDIA
8
GB200/B200 rack-scale AI systems
- Version
- Blackwell
- Cost
- Pricing is not publicly listed (enterprise hardware architecture/system)
- Model
- Blackwell
5
Cloud-native training silicon
- Version
- Trainium2
- Cost
- Pricing not published; available via AWS EC2 instance pricing.
- Model
- Neuron compiler
Groq LPU
Groq
4
Deterministic low-latency inference
- Version
- LPU
- Cost
- Pricing not published. Contact sales or check account console for usage-based rates.
- Model
- canopylabs/orpheus-arabic-saudi
Huawei Ascend 910C
HuaweiChina
4
China domestic AI accelerator
- Version
- Ascend 910C
- Cost
- Pricing is not publicly available (enterprise hardware)
- Model
- Ascend 910C
Elsewhere on the same models
Products in other segments that name one of Google TPU Ironwood's models in their stack.
Box AI
Box
Content-layer AI over enterprise document repositories.
- Version
- 2025
- Cost
- Bundled from Enterprise Plus
- Model
- Gemini
Gemini in LookerGoogle
Conversational analytics over the Looker semantic model.
- Version
- 2025
- Cost
- Bundled with Looker
- Model
- Gemini
LearnLM
GoogleUnited States
9
Pedagogy-tuned model family
- Version
- Gemini 2.5 integration
- Cost
- Integrated into Gemini starting with the 2.5 model series. Pricing is governed by Gemini API plans.
- Model
- Gemini
MedLM
Google Cloud
4
Medical model family on Vertex
- Version
- MedLM
- Cost
- Pricing is not detailed in the provided text for MedLM; contact sales or refer to Google Cloud Vertex AI pricing.
- Model
- Gemini
Change history
- version
Google TPU Ironwood moved to Ironwood (7th generation)
7th-generation (Ironwood)Ironwood (7th generation)
source - version
- version
- capability
- version
Google TPU Ironwood moved to Ironwood (7th generation)
7th-generation (Ironwood)Ironwood (7th generation)
source - version
Google TPU Ironwood moved to 7th-generation (Ironwood)
7th generation (Ironwood)7th-generation (Ironwood)
source - version
Google TPU Ironwood moved to 7th generation (Ironwood)
Ironwood (7th generation)7th generation (Ironwood)
source - version
Google TPU Ironwood moved to Ironwood (7th generation)
7th generation (Ironwood)Ironwood (7th generation)
source - version
Google TPU Ironwood moved to 7th generation (Ironwood)
Ironwood (7th Gen)7th generation (Ironwood)
source - capability
- version
- capability
- version
Google TPU Ironwood moved to 7th generation (Ironwood)
7th generation (TPU7x)7th generation (Ironwood)
source - capability
New capabilities: JAX supportvLLM support, Google Kubernetes Engine (GKE)
60 tracked62 tracked · +JAX supportvLLM support, Google Kubernetes Engine (GKE)
source - version
Google TPU Ironwood moved to 7th generation (TPU7x)
7th generation (Ironwood)7th generation (TPU7x)
source - capability
- version
Google TPU Ironwood moved to 7th generation (Ironwood)
Ironwood (7th generation)7th generation (Ironwood)
source - model
Google TPU Ironwood added Ironwood TPU (7th Generation) to its model stack
Gemini, Ironwood, Ironwood (TPU v7), Ironwood TPU, TPU v7, TPU7x, TPU7x (Ironwood), XLA compilerGemini, Ironwood, Ironwood (TPU v7), Ironwood TPU, Ironwood TPU (7th Generation), TPU v7, TPU7x, TPU7x (Ironwood), XLA compiler
source - capability
New capabilities: MaxText blueprint, Tunix library
57 tracked59 tracked · +MaxText blueprint, Tunix library
source - version
Google TPU Ironwood moved to Ironwood (7th generation)
7th-generation (TPU7x)Ironwood (7th generation)
source - version
Google TPU Ironwood moved to 7th-generation (TPU7x)
Ironwood (7th generation)7th-generation (TPU7x)
source - model
Google TPU Ironwood added Ironwood (TPU v7) to its model stack
Gemini, Ironwood, Ironwood TPU, TPU v7, TPU7x, TPU7x (Ironwood), XLA compilerGemini, Ironwood, Ironwood (TPU v7), Ironwood TPU, TPU v7, TPU7x, TPU7x (Ironwood), XLA compiler
source - version
Google TPU Ironwood moved to Ironwood (7th generation)
TPU7x (Ironwood)Ironwood (7th generation)
source - version
- capability
New capabilities: JAX Support, PyTorch Support, vLLM Support
54 tracked57 tracked · +JAX Support, PyTorch Support, vLLM Support
source - version
- version
- version
Google TPU Ironwood moved to 7th generation (TPU7x)
7th generation (Ironwood)7th generation (TPU7x)
source - version
Google TPU Ironwood moved to 7th generation (Ironwood)
7th-generation (TPU7x)7th generation (Ironwood)
source - version
Google TPU Ironwood moved to 7th-generation (TPU7x)
7th generation (Ironwood)7th-generation (TPU7x)
source - capability
- version
Google TPU Ironwood moved to 7th generation (Ironwood)
7th-generation TPU (Ironwood)7th generation (Ironwood)
source - version
Google TPU Ironwood moved to 7th-generation TPU (Ironwood)
7th-generation (TPU7x)7th-generation TPU (Ironwood)
source - version
- capability
New capabilities: PyTorch, JAX, vLLM
45 tracked53 tracked · +PyTorch, JAX, vLLM, OpenXLA, MaxText, Tunix, GKE, Liquid-cooled
source - version
- capability
- version
- capability
- capability
New capabilities: Frontier training, Large-scale inference, Multi-step reasoning
40 tracked43 tracked · +Frontier training, Large-scale inference, Multi-step reasoning
source
Google TPU Ironwood compared
Straight head-to-head pages against the busiest products in Chips & Compute.
How to cite this page
Free to cite and reuse under CC BY 4.0. Permalink: https://tomorrow.aliensquad.ai/tools/google-tpu
APA
Tomorrow. (2026). Google TPU Ironwood — version, pricing and model stack [Data set entry]. AlienSquad. Retrieved 2026-09-17, from https://tomorrow.aliensquad.ai/tools/google-tpu
BibTeX
@misc{tomorrow-tools-google-tpu,
author = {{Tomorrow}},
title = {Google TPU Ironwood — version, pricing and model stack},
year = {2026},
publisher = {AlienSquad},
howpublished = {\url{https://tomorrow.aliensquad.ai/tools/google-tpu}},
note = {Accessed: 2026-09-17}
}