Google Launches TPU 8i Max and TPU 8tx for 2027

Key Takeaways

• Google launched TPU 8i Max and TPU 8tx as part of its eighth-generation TPU architecture, with general availability expected in 2027.
• TPU 8tx is the first cloud TPU to offer hardware-enforced confidential computing, using AMD EPYC processors for secure AI workloads.
• TPU 8i Max combines two TPU 8i chips to double compute, SRAM, and HBM bandwidth for large context windows.
• Google Cloud revenue rose 103 percent year over year to $30.7 billion in the third quarter.
• AMD supplies EPYC processors to more than 10 major hyperscalers and AI labs, strengthening its position in AI infrastructure.
• Google also open-sourced tools including TorchTPU and TPU Sync to help developers manage agentic AI workloads.
• Google.org granted Clalit Health Services $2 million to build a voice-based AI platform for healthcare access in Israel.
• Tesla will end Cybertruck production by mid-2027 and retool the Gigafactory Texas line to build Cybercab robotaxis.
• AWS committed $100 million in AI and cloud credits over five years to health organizations in underserved communities.
• Revelion, SK Telecom, and Nvidia signed an MOU to integrate NPU and GPU systems using Nvidia's NV Link Fusion.

Google expands TPU chip family with 8i Max and 8tx

Google introduced two new AI chips, TPU 8i Max and TPU 8tx, as part of its eighth-generation TPU architecture. TPU 8i Max doubles compute, SRAM, and high-bandwidth memory to handle larger AI models and context windows. TPU 8tx is designed for confidential inference, keeping sensitive data protected during processing. The chips target memory and security bottlenecks in AI infrastructure.

AMD EPYC powers Google Cloud TPU 8tx confidential computing

AMD announced that its 5th Gen EPYC processors will host the Google Cloud TPU 8tx system. This system is the first cloud TPU to offer hardware-enforced confidential computing. The partnership combines AMD CPU performance and security features with Google's AI acceleration. It helps enterprises run sensitive AI workloads while keeping data protected during use.

AMD shares rise as Google picks EPYC for secure TPU 8tx

AMD revealed that its fifth-generation EPYC server processors will power Google Cloud's TPU 8tx. This marks the first cloud TPU system with hardware-based confidential computing. The processors handle data preparation while helping protect sensitive AI models and customer data. AMD said its chips are used by over 10 major hyperscalers and AI labs.

Google targets AI bottlenecks with two new TPU chips

Google Cloud announced TPU 8tx and TPU 8ix Max to address growing AI infrastructure demands. TPU 8tx adds confidential computing with security built into the hardware layer, using AMD EPYC hosts. TPU 8ix Max doubles silicon per package and adds more memory to handle larger context windows. Google also introduced TPU Sync to help manage memory and storage for agentic workloads.

AMD gains as Google Cloud uses EPYC for secure AI

Advanced Micro Devices saw a stock rise after Google Cloud chose EPYC processors for secure TPU 8tx AI computing. The processors handle data preparation and workloads while boosting confidential AI computing. This collaboration pairs AMD's chip technology with Google's eighth-generation TPU architecture. It supports AI inference, fine-tuning, and agentic AI workloads.

Google launches dual-chip TPU 8i Max and confidential TPU 8tx

Google added two new eighth-generation TPU systems to its lineup. TPU 8i Max combines two TPU 8i chips in one package to double compute, SRAM and HBM bandwidth. TPU 8tx uses TPU 8t silicon with fifth-generation AMD EPYC hosts for confidential computing. Both systems target secure and high-density AI workloads.

Google gives AMD a major role in new AI chip system

Google chose AMD fifth-generation EPYC server processors for its new TPU 8tx confidential computing system. AMD processors will handle data preparation and supporting tasks alongside Google custom AI accelerators. The setup targets AI inference, model fine-tuning and agentic AI applications. AMD already supplies EPYC chips to more than 10 leading hyperscalers and AI labs.

Google unveils two specialized AI chip systems to challenge Nvidia

Google introduced TPU 8tx and TPU 8i Max as specialized versions of its eighth-generation tensor processing units. TPU 8tx is designed for government and regulated industry customers needing secure data processing. TPU 8i Max combines two TPU 8i chips to handle large context windows like codebases or video. Google Cloud revenue rose 103 percent year over year to 30.7 billion dollars in the third quarter.

AMD lands key role in Google TPU 8tx as chip rivalry heats up

Google's new TPU 8tx uses AMD fifth-generation EPYC processors for confidential computing. This marks a shift because earlier TPU 8t and 8i systems used Google's own Axion CPUs. The system protects sensitive data during AI processing for banks and government agencies. Google expects the new TPU systems to become available in 2027.

AMD stock rises after Google picks its chips for new AI system

AMD shares climbed after Google selected AMD EPYC processors for the TPU 8tx confidential AI system. The EPYC chips will handle data preparation and help protect sensitive AI models and customer data. AMD said more than 10 major hyperscalers and AI labs already use its EPYC processors. The deal builds on a long-running partnership between AMD and Google Cloud.

Google expands TPU lineup with four new AI accelerator systems

Google introduced four eighth-generation TPU systems for different AI workloads. TPU 8t handles scale-up training, TPU 8i focuses on low-latency reasoning, TPU 8tx provides confidential computing, and TPU 8i Max targets massive context windows. The systems offer up to 2.7x better training performance and 80% better inference price-performance. Google also open-sourced tools including TorchTPU and TPU Sync to make TPUs easier for developers. New TPUs are expected to launch in 2027.

Google.org funds Clalit AI project for patient healthcare access

Google.org gave Clalit Health Services a $2 million grant for two years to build a voice-based AI platform. The system uses large language models to help patients access healthcare in regional dialects. It works on standard mobile phones and landlines without needing internet. The platform will do automated symptom intake and triage before appointments. Clalit plans to start with a pilot then expand nationwide and release parts as open-source.

Google funds Clalit voice AI for equitable healthcare access

Google.org granted Clalit $2 million over two years for its Digital Front Door conversational AI assistant. The system uses Google Gemini and large language models on Google Cloud infrastructure. Speech recognition is tuned to local Israeli dialects so patients can book appointments and get triage by phone. The platform works on landlines and basic mobile phones without internet. Clalit plans a pilot rollout then nationwide deployment and will open-source the architecture.

Claude Code missed $800K in contract analysis despite correct answers

Claude Code failed to find $800,000 in procurement savings during a post-merger analysis of Harbor Health and Summit Health Plan. The AI agent gave some correct answers but missed key contract details because documents were unlinked in databases. The team built a five-stage pipeline using DSPy and OCR to process over 40,000 contracts. They found that structured data input prevents Claude Code from rediscovering information. The experience highlights limits of AI agents in contract intelligence without proper data preparation.

Israeli startup Zeroport builds hardware to stop rogue AI agents

Israeli startup Zeroport created a hardware-based security platform to control autonomous AI agents. Founded in 2024 by CEO Joseph Gertz and CTO Lavi Friedman, the system uses custom chips placed between AI systems and enterprise networks. The hardware can physically cut connections when an AI agent tries unauthorized actions. Zeroport believes software-only security is not enough as AI agents become more capable. The company compares its approach to a drawbridge that goes up to block threats.

India may benefit as global AI investment cycle slows

A report from SBI Funds says concerns about AI spending sustainability could redirect investment toward India. AI capital expenditure in the US is projected to average 3.6 percent of GDP between 2025 and 2032. India's market indices rose in October while many emerging markets fell. The country's merchandise exports grew about 23 percent in Q3 2026.

Prisoners' Education Trust launches AI courses for inmates

Prisoners' Education Trust offers three paper-based AI courses for prisoners in England and Wales. The Level 2 Foundations of AI course covers business and marketing basics. Two Level 3 courses focus on business transformation and marketing. Each course takes 140 to 165 hours and runs for nine months.

AI stocks face pressure as spending raises return concerns

Tech stocks are under pressure as investors question AI company returns on spending. Hyperscaler AI capital expenditure is projected at $700 billion this year and over $1 trillion next year. This spending is turning free cash flow negative. Experts advise diversified portfolios for long-term AI value.

Revelion to build AI system combining NPU and GPU tech

Revelion, SK Telecom, and Nvidia signed an MOU to integrate domestic NPU and GPU systems. Nvidia's NV Link Fusion connects Revelion's R200 NPU to Nvidia's platform. GPUs handle question analysis while NPUs generate answers. SK Telecom will test the integrated systems in AI data centers.

AWS commits $100M in AI and cloud credits to health groups

Amazon Web Services committed $100 million in AI and cloud credits over five years to health organizations in underserved communities. The AWS Health Impact Initiative targets groups priced out of AI adoption. Focus areas include health access, women's and children's health research, and nutrition. Recipients can apply through an AWS account team or a public portal.

Tesla ends Cybertruck production in 2027 for Cybercab robotaxis

Tesla will stop making the Cybertruck by mid-2027 after sales dropped 48% to 20,237 units in 2025. The Gigafactory Texas line will be retooled to build Cybercab robotaxis, which are steering-wheel-free two-seaters. Tesla started public Cybercab rides in Austin in September 2026 and grew its fleet from under 40 to over 300 vehicles in one month. This marks the third vehicle Tesla has discontinued as it pivots toward AI and robotics, having also ended the Model S and Model X.

Sources

• Google advances agentic AI custom silicon expanding TPU family with 8tx and 8i Max
• Google Cloud's first AI chip system uses AMD processors to help protect sensitive data as it runs
• AMD surges as Google Cloud taps EPYC for secure TPU 8tx AI
• Google targets AI’s next infrastructure bottlenecks with two new TPUs
• AMD surges as Google Cloud taps EPYC for secure TPU 8tx AI
• Google Unveils TPU 8i Max and TPU 8tx: A Dual-Chip TPU and a Confidential TPU
• Google Hands AMD Major Win in AI Chip Race
• Google unveils 2 specialized AI chip systems in growing Nvidia challenge: Report
• AMD Lands Key Role in Google's New AI Chip System as Nvidia Rivalry Intensifies
• AMD Stock Jumps as Google Taps Its Chips for a New AI System
• Google Releases New TPUs — TPU 8t For Scaled Up Training, TPU 8i For Low Latency Reasoning & TPU 8tx For Confidential Computing
• Google.org awards Clalit $2 million to develop AI system for patient access
• Google.org grants Clalit $2M for voice AI in healthcare
• Claude Code looked right and missed $800K
• Israeli start-up Zeroport seeks to put physical limits on AI agents
• India may gain as global AI investment cycle slows: Report
• Get set for the future with PET’s AI courses Artificial Intelligence (AI) is changing the world of work. But it’s so new, many of us have never tried it.
• AI Stocks Under Pressure: What Investors Need To Know
• K Semiconductor startup Revelion will build an artificial intelligence (AI) infrastructure that uses..
• AWS commits $100M in AI and cloud credits to health groups in underserved communities
• Tesla to End Cybertruck Production in 2027, Pivots Factory to Cybercab