跳到主要内容Skip to main content

← 返回硬件算力← Back to Compute

加速卡Accelerators未公布单价No public price

Google TPU 8i Max(超大上下文)Google TPU 8i Max (massive context)

未公布单价No public price未公布单价No public price · 官方页未给公开单价;这类产品通常按整机、云实例或询价成交,不拿经销商价冒充官方价。No public unit price on the official page; these usually sell as systems, cloud instances or on request, so dealer quotes are not presented as official.

回答「长上下文 agent 把 KV cache 和状态撑爆时,TPU 侧有什么解法」。8i Max 不是一颗新芯片,而是把两颗低延迟 8i 封进同一封装:算力、SRAM、HBM 带宽相对 8i 都翻倍,同时在原机架 footprint 内塞进两倍硅片,提高每机架算力密度。
不主张:翻倍是相对 TPU 8i 的厂商口径;不是单芯片,软件与并行策略要按双芯片封装处理;没有公开价格、功耗或第三方长上下文实测;不能直接推导「比 GPU 更适合长上下文」。
Answers what TPU offers when long-context agents blow past KV-cache and state budgets. 8i Max is not a new chip: it combines two low-latency 8i dies in one package, doubling compute, SRAM and HBM bandwidth versus 8i while fitting twice the silicon into the existing rack footprint for higher compute density per rack.
It does not claim: the doubling is a vendor figure versus TPU 8i; it is not a single chip, so software and parallelism must treat it as a dual-die package; no public pricing, power or third-party long-context measurement; it does not directly prove TPUs are better than GPUs for long context.

关键参数Key specs

定位 / role
超大上下文 + 最高单机架算力密度
结构 / package
两颗 TPU 8i 封在同一封装
算力 / compute
2× TPU 8i(Google 口径)
SRAM
2× TPU 8i(Google 口径)
HBM 带宽 / bandwidth
2× TPU 8i(Google 口径)
机架密度 / rack
相同 footprint 内容纳 2× 硅片
同代 / family
TPU 8t、TPU 8i、TPU 8tx、TPU 8i Max
状态 / status
2026-11-09 发布

官方原文怎么说What the source says

TPU 8i Max for massive context windows and the highest compute density per rack
1 https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/tpu-8i-max-and-8tx/
we took two low-latency TPU 8i chips and combined them onto a single package.
2 https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/tpu-8i-max-and-8tx/
delivering double the compute, double the SRAM and double the HBM bandwidth of TPU 8i
3 https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/tpu-8i-max-and-8tx/
because TPU 8i Max fits twice the silicon within the existing rack footprint, data centers can deploy significantly more computing power per square foot and per watt.
4 https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/tpu-8i-max-and-8tx/

资料与许可Source and licence

资料时间Source checked
2026-11-20
许可 / 类型Licence / kind
vendor-claim
许可说明Licence evidence
https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/tpu-8i-max-and-8tx/
类目Category
加速卡Accelerators
状态Status
已发布 / 未出货Announced
适合Use cases
在线服务Serving
价格口径Price basis
未公布单价No public price · not-published

同类目还有More in this category

下一步Next step

去官网看原页Open the official page →

← 返回硬件算力← Back to Compute