Alphabet Unveils TPU 8t, TPU 8i with Triple Performance and $750M Fund
Alphabet unveiled TPU 8t and TPU 8i chips at Cloud Next, scaling TPU 8t to 9,600 units for threefold training power and boosting TPU 8i inference SRAM threefold in 1,152-TPU pods. It unveiled a $750 million fund, processed 16 billion API tokens and recorded 40 percent growth in Gemini Enterprise.
1. TPU 8t and TPU 8i Launch
Alphabet introduced its eighth-generation Tensor Processing Units at Cloud Next. TPU 8t is optimized for model training, scaling up to 9,600 chips in a single superpod with two petabytes of shared memory and delivering three times the processing power of its predecessor, while TPU 8i is tailored for inference, connecting 1,152 TPUs in a pod with triple on-chip SRAM to support low-latency, high-throughput AI agents.
2. $750 Million Ecosystem Fund
Google Cloud committed a $750 million partner fund to its 120,000-member ecosystem, supplying resources for AI value assessments, prototyping, agent development and upskilling initiatives across consulting firms, systems integrators and software partners.
3. API Throughput and Enterprise Growth
The platform now processes 16 billion API tokens and noted that 75 percent of customers use Google’s AI products. It also recorded 40 percent quarterly growth in paid Gemini Enterprise users, underscoring accelerating adoption of its AI agent infrastructure.