Google TPU：即将到来的日期

最近

• 半导体2026/11/08@ChipDesk

  Google Announces TPU 8i for Low-Latency AI Inference

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2026/04/22已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 6 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉200天前

  Dated Google^[2][3] announcement; commercialization milestones are recorded separately in the sourced timeline.

  Google announces its inference-focused eighth-generation TPU, combining 288 GB of HBM, 384 MB of on-chip SRAM and Boardfly networking across up to 1,152 chips. Availability is still listed as coming soon.

  000
• 半导体2026/11/08@ChipDesk

  Google Announces TPU 8t for Frontier AI Training

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2026/04/22已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 6 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉200天前

  Dated Google^[1][3] announcement; commercialization milestones are recorded separately in the sourced timeline.

  Google announces the training-focused half of its eighth TPU generation: 9,600-chip superpods, native FP4, 216 GB of HBM per chip and Axion hosts. The current product page lists it as coming soon.

  000
• 半导体2026/11/08@ChipDesk

  Google Unveils Ironwood, Its Seventh-Generation TPU

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2025/04/09已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 6 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉1.6年前

  Dated Google^[2][3] announcement; commercialization milestones are recorded separately in the sourced timeline.

  Google announces Ironwood for large-scale inference, later documented as TPU7x for both training and inference, with 192 GiB of HBM per chip and pods of up to 9,216 chips.

  000
• 半导体2026/11/08@ChipDesk

  Google Unveils Trillium, Its Sixth-Generation TPU

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2024/05/14已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 7 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉2.5年前

  Dated Google^[1][2][3][4] announcement; commercialization milestones are recorded separately in the sourced timeline.

  Google announces Trillium (TPU v6e), with 4.7 times the peak compute per chip of TPU v5e, double its HBM capacity and a claimed improvement of over 67% in energy efficiency.

  000
• 半导体2026/11/08@ChipDesk

  Google Announces TPU v5p and the AI Hypercomputer

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2023/12/06已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 5 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉2.9年前

  Dated Google^[1][2][3] announcement; commercialization milestones are recorded separately in the sourced timeline.

  Google announces its performance-focused fifth-generation TPU, scaling to 8,960-chip pods with 96 GiB of physical HBM per chip, alongside its integrated AI Hypercomputer architecture.

  000
• 半导体2026/11/08@ChipDesk

  Google Introduces TPU v5e for Cost-Efficient AI Training and Inference

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2023/08/29已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 5 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉3.2年前

  Dated Google^[1][2][3] announcement; commercialization milestones are recorded separately in the sourced timeline.

  Google launches TPU v5e in preview, targeting medium- and large-scale AI training and inference with 256-chip pods and workload-dependent improvements in performance per dollar.

  000
• 半导体2026/11/08@ChipDesk

  Google Announces TPU v4, Scaling AI Pods Beyond One Exaflop

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2021/05/18已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 5 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉5.5年前

  Dated Google^[2][3] announcement; commercialization milestones are recorded separately in the sourced timeline.

  Google unveils TPU v4 at I/O 2021. Its 4,096-chip pods scale beyond one exaflop, while each chip provides 275 BF16 teraflops and 32 GiB of HBM2.

  000
• 半导体2026/11/08@ChipDesk

  Google Introduces TPU v3 With Liquid-Cooled AI Supercomputing

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2018/05/08已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 5 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉8.5年前

  The I/O 2018 reveal date is reported by 9to5Google^[5] on May 8, 2018; Google^[1][2] release notes separately date beta and general availability.

  Google introduces its third TPU generation at I/O 2018. Cloud documentation specifies 32 GiB of HBM2 and 123 BF16 teraflops per chip, scaling to 1,024-chip pods.

  000
• 半导体2026/11/08@ChipDesk

  Google Announces TPU v2 and Cloud TPU Training Pods

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2017/05/17已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 5 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉9.5年前

  Dated Google^[1][2][3] announcement; commercialization milestones are recorded separately in the sourced timeline.

  Google adds neural-network training to its custom TPU platform, announcing Cloud TPU devices with up to 180 teraflops and interconnected pods for larger workloads.

  000
• 半导体2026/11/08@ChipDesk

  Google Reveals TPU v1, Its First Custom AI Inference Chip

  Googleplex, Mountain View, California, United States (company headquarters)Google

  

  开始于 2016/05/18已确认日期确定性: 已确认已公布为确定日期来源对日期的表述有多确定：已确认、已排期、预估、已推迟或未核实。显示所有“已确认”的图钉可信度 95%总体可信度: 95%该图钉的来源和参考资料对其日期的支持程度。包括来源在内的 13 份资料对图钉开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半显示所有可信度不低于 75% 的图钉10.5年前

  Dated Google^{[1][2][4][5][7][8][9][11]} announcement; commercialization milestones are recorded separately in the sourced timeline.

  Google discloses its first Tensor Processing Unit after more than a year of internal use, a custom TensorFlow accelerator already powering Search, Street View and AlphaGo.

  000