Google Cloud AI infrastructure supply-chain and manufacturing operations vice president Rehan Sheikh said at Foxconn Technology Day that Taiwan remains a core hub for Google’s custom chips and systems innovation. The company’s TPU 8t and TPU 8i are designed for training and inference respectively, while Google is moving toward releasing two inference and training chips per year. Sheikh also said Google aims to reduce the chip-to-service deployment cycle from 12–18 months or more to about six months.
Alphabet shares fell more than 13% in intraday trading, erasing over $500 billion in market value and setting a record for the company’s single-day loss, according to a GeekPark report. Third-quarter revenue excluding partner share revenue was $111.2 billion, below analyst estimates, while search and other advertising revenue grew at its slowest rate in five quarters. At 9:35 a.m. New York time, the shares were down 13% at $281.37, with $511 billion in market value erased. Google Cloud was a bright spot, reporting $30.7 billion in revenue and $11.1 billion in profit, above analyst forecasts.
Google Developers Blog says DevFest 2026 will run from October 1 through December 31, hosted by Google Developer Groups, with sessions and workshops covering Google AI technologies, Google Cloud, Android and Web. Google expects the community-led series to include more than 800 events across 115 countries and nearly one million attendees, although those attendance and scale figures are vendor-reported.
According to a 9to5Google newsletter, Alphabet reported 25% revenue growth and $26 billion in net income for Q3, but analyst disappointment sent shares down as much as 11% in after-hours trading. Cloud revenue doubled year over year to $30 billion as the business attracted frontier AI labs, while advertising revenue missed expectations. The newsletter also said Alphabet was expanding advertising in AI Mode to 10 new markets.
TMTPost’s Edge AI Daily reports divergent quarterly results for Microsoft and Alphabet, alongside major AI infrastructure financing, Oracle’s planned deployment of Google TPU 8t and 8i, and NVIDIA’s open-weight Nemotron 3.5 Transcribe model. It also covers Anthropic’s S-1 filing, OpenAI’s diversified chip strategy, Meta’s equity raise, and Gemini 4 Carbon, while framing AI investment scrutiny around capital spending, free cash flow and AI-related revenue.
Anthropic outlines a “control, track, and prove” framework for organizations using Claude Code on Google Cloud. Recommended tools include consumption-based billing, dedicated projects, billing budgets and alerts, preview spend caps, quotas, provisioned throughput, BigQuery exports, OpenTelemetry telemetry and Cloud Monitoring dashboards. Anthropic says token consumption varies with task complexity, conversation length and model pricing, and that Claude Code typically uses more tokens per task than chat because it performs multistep work such as reading files, running tools and reasoning across a task.
TechNews reports that NVIDIA appointed former Google Cloud executive Alison Wagonfeld as its first chief marketing officer and Microsoft veteran Nick Parker as sales chief. Several longtime executives also retired or handed over responsibilities, with Shanker Trivedi’s portfolio divided among younger managers as NVIDIA adopts a flatter structure. Finance, legal, operations and engineering remained comparatively stable.
Google Developers Blog introduced HappyDelivery, a functional prototype that implements an earlier blueprint for conversational commerce using LangGraph, Gemini, FastAPI, and Google Cloud. The prototype uses a three-layer decoupled design for message ingress, session and cart persistence, and cognition, with specialized agents, deterministic financial tools, and an automated verification suite. Google also reported multi-agent latency of 2–4 seconds and emphasized that keeping calculations in deterministic tools is necessary for financial accuracy.
Sierra announced that its AI-agent platform is now available through Google Cloud Marketplace, allowing customers to contract through existing Google Cloud relationships with unified billing and Private Marketplace governance. Sierra says qualifying Marketplace purchases can count toward existing Google Cloud commitments, while its platform charges for delivered agent outcomes rather than consumed tokens.
An Edge AI Daily report said AWS ended the second quarter of 2026 with $496 billion in remaining performance obligations and $42.2 billion in quarterly revenue, up 37% year over year. Amazon CEO Andy Jassy said the company’s 2026 capital-spending guidance had been raised to $220 billion but still would not meet all demand, with a capacity gap expected in 2027 and demand described as “shocking” for 2028. The report also said Amazon added 3.9 GW of data-center power capacity in 2025 and plans to double total power capacity by the end of 2027.
Google announced Agent Compute Engine (ACE), a Google Cloud service that gives each AI agent its own hardware-isolated microVM for running code, calling tools, controlling browsers and completing multistep tasks. The environments can suspend when idle, release resources and later restore file and memory state. Google said a test involving 200,000 agents showed pause and resume taking less than 500 milliseconds. ACE initially supports Gemini, the Gemini platform, Antigravity and Google’s internal model-evaluation and reinforcement-learning workflows. Developers will be able to use ACE through Google Cloud, controlling environments’ images, identities, network access and lifecycles while Google manages underlying resource scheduling and scaling.
Claude Blog and Google AI on DEV report that Claude Models on Google Cloud (Anthropic managed) is generally available through US and Global endpoints. According to Anthropic, the offering provides the full Claude Platform feature set, with new features and betas available on the same day they launch natively, while using existing Google Cloud IAM credentials, one invoice and applicable Google Cloud commitments. The Google-managed option remains available with Google as data processor, regional data residency and native integrations such as Model Armor and Agent Engine.
A commentary republished by TMTPost and written by Naojiti asks who actually funds Anthropic, citing Reuters figures that its 2025 revenue approached $4.6 billion, roughly 12 times the previous year. More than 80% of that came from usage-based API and token billing rather than subscriptions, and about 47% of sales flowed through Amazon and Google cloud platforms, the piece says, while 2025 operating losses exceeded $8 billion, including $7.3 billion in compute and infrastructure spending. The article also flags that talk of a $2 trillion valuation and a record IPO sits among unverifiable rumours, and argues that the loudest Chinese-language discussion — account resale and access workarounds — is not Anthropic's core business, with the US market supplying nearly two-thirds of sales.
Google announced that streaming support for Google Cloud API Gateway is available in Public Preview. The feature supports SSE, chunked transfer encoding, HTTP response streaming, WebSockets and gRPC streaming while retaining existing authentication, API key enforcement, routing and observability policies. Google says gateways using model routing can stream responses from Gemini, Claude and OpenAI gpt-oss models, with streams able to remain open for up to one hour.
Google Cloud introduced Project Chrysalis, a research reference architecture for governed, self-evolving AI agents that can diagnose outages, synthesize mitigation code, assess interventions, retain operational lessons and propose changes to an agent hierarchy. Built on the Gemini platform and Agent Development Kit, it uses Model Armor, Cloud KMS, Binary Authorization, Open Policy Agent and human approval gates to constrain autonomous actions. In automated Black Friday cache-stampede demo simulations, Google reported that time to first mitigation fell from 15 to 45 minutes to 1.8 seconds, HTTP 5xx error rates fell to 0.0%, and P99 latency dropped from 519.8 milliseconds to 18.2 milliseconds. The report did not provide independent production validation.
Google Cloud and Nvidia announced on October 21 that A5X instances running Nvidia's Vera Rubin NVL72 system will open next month, with plans to extend the offering to the European market, according to a 36Kr brief citing Jiemian. The two companies position the infrastructure as compute for AI agent inference and reinforcement learning workloads. The report did not disclose pricing, specific regions or configuration details.
OpenAI and Google Cloud announced a partnership to distribute OpenAI frontier models and Codex agentic workflows through Model Garden, with purchases transacted through Google Cloud Marketplace. Google Cloud customers with existing spend commitments would be able to apply those commitments, while OpenAI models are also planned for Google Antigravity and Gemini Agent. OpenAI says general availability is planned for the coming weeks, while secondary reports describe a private preview for selected enterprise customers and disagree on the exact model lineup and affected products.
Google Cloud and NVIDIA announced that A5X instances built on NVIDIA Vera Rubin NVL72 will reach general availability next month, with planned expansion of availability to Europe. The instances combine NVIDIA ConnectX-9 SuperNICs, Google Cloud's Virgo Network and fourth-generation liquid cooling, and Google Cloud plans to scale its Vera Rubin-based infrastructure to as many as 960,000 Rubin GPUs across multiple sites. According to NVIDIA's announcement, London-based Ineffable Intelligence, founded by AlphaGo creator David Silver, was the first customer to access A5X and is using it for large-scale reinforcement learning.
NVIDIA announced Confidential Bare Metal, a new deployment option within NVIDIA Confidential Computing that it says lets frontier model builders run sensitive workloads on dedicated third-party infrastructure while retaining control of their models and data. NVIDIA says the option adds platform security controls that protect data in transit, measure platform integrity and enforce hardware-level access restrictions, and it now sits alongside Confidential Containers and Confidential Virtual Machines. Support extends to Grace Blackwell NVL72, HGX Rubin NVL8 and DGX Rubin NVL8 systems and will integrate with NVIDIA DSX OS, while Cohere plans to add it to Cohere Model Vault and Google Cloud is extending support to A4X Max instances powered by GB300 NVL72 systems, according to the company.
At GTC Berlin on October 21, NVIDIA said European AI clouds and infrastructure partners are adopting its DSX AI Factory Platform, a full-stack reference architecture that co-designs compute, networking, power and cooling as one system. NVIDIA says DSX reference designs using closed-loop liquid cooling and dry coolers can bring facility cooling water use close to zero, with potential savings of 54 million liters a day for a 2-gigawatt deployment, equivalent to the daily water use of 180,000 households. The company also says DSX MaxLPS software can use up to 30% less power for the same token output, with CoreWeave, Firmus, Google Cloud, IREN, Lambda, Nebius, Nscale and Sharon AI planning to use it, and it reported partner evaluations such as 49% higher inference throughput for Nscale within a 264-kilowatt budget, while giving no general availability dates for several DSX software tools still in developer preview.
NVIDIA announced the OpenShell Reference Architecture at GTC Berlin, a full-stack design for building, containing, observing and enforcing safe AI agents on enterprise and edge infrastructure, and said it is now available. It combines OpenShell secure runtime software with Vera Rubin systems for inference, Vera CPUs for agent harnesses and tool use, and BlueField-4 DPUs running NVIDIA Sentry for in-silicon enforcement, extending to RTX Spark, DGX Spark and DGX Station at the edge. NVIDIA says Vera Rubin NVL72 delivers 30x higher throughput per megawatt on agentic workloads and that Sentry stops an agent leaving its boundary in milliseconds while raising agent capacity by up to 20%; those figures are vendor claims not independently measured. Anthropic, AWS, Cisco, CrowdStrike, Dell, Google Cloud, HPE, IBM, Meta, Microsoft, OpenAI, Palo Alto Networks, SAP and ServiceNow are among companies NVIDIA lists as working with the architecture, and Sentry is planned for availability early next year, with DOCA libraries available now.
Google Cloud announced a version of Gemini for government services, built on a customized Gemini Agent and developed with federal and state agencies including the Missouri state government and the U.S. Department of Health and Human Services, IT之家 AI reported. The offering has five parts: Skills that teach agents laws, regulations and procedures; connectors to existing systems; partner-built agents; a governance control platform; and a partner ecosystem. It can access Oracle ERP and HR systems, Microsoft 365 and Google Workspace without agencies migrating their data, according to the report. Google says the product meets the highest FedRAMP level, keeps data in a private cloud environment within the United States and will not be used to train its foundation models without permission — claims reported from the vendor, not independently verified.
Google CEO Sundar Pichai said the company is shifting some compute from Google Cloud toward frontier AI research, in some cases sacrificing near-term cloud revenue. He also said customers seeking 2027 capacity face extreme difficulty because data centers take two to three years to deploy while demand for intelligent-agent workloads is rising; Google rented external compute last year to meet some customer needs.
Nomura published a global industry research report arguing that market worries about slowing AI revenue and tight financing are overreactions, and that the global memory shortage will last at least until 2028, TechNews reports. The report projects global data-center power consumption rising from 60GW in 2025 to 115GW in 2028, with AI's share climbing from 17% to 35%, and estimates combined OpenAI and Anthropic revenue grew 320% to 507% year over year in the first three quarters of 2026. Nomura also estimates the combined net cash of nine semiconductor giants including Samsung, SK Hynix, Micron, TSMC and Nvidia rose from $27 billion in Q1 2024 to $282 billion in Q2 2026, while seven data-center operators swing to an estimated $300–400 billion net debt by 2028; it maintained Buy ratings on Samsung Electronics and SK Hynix.
Anthropic publicly filed an S-1 registration statement with the SEC and plans to list Class A shares on Nasdaq under the ticker ANT. The filing reports first-half 2026 revenue of $16.286 billion, usage-based revenue of $13.064 billion, and a first quarterly operating profit of $974 million in the second quarter. A secondary analysis estimated compute commitments at about $1.01 trillion on an upper-bound basis, including contracts still under negotiation and potential additions. TechCrunch reported, citing an S-1 leaked to The Wall Street Journal, that Anthropic has had difficulty securing sufficient capacity from AWS, Google, and Microsoft, which increasingly view the Claude maker as a competitor. Anthropic did not immediately respond to TechCrunch’s request for comment.
In an AWS guest post co-written with Eon, Eon described using Amazon Bedrock AgentCore Gateway to replace per-tool credentials and local Model Context Protocol (MCP) processes with one authenticated gateway. The system gives developers and autonomous agents unified access to investigation tools across AWS, GCP, and Azure. Cedar policies separate human and agent permissions, while IAM cross-account role assumption connects the gateway to runtimes in multiple AWS accounts. Eon reported that developer onboarding fell from roughly two days to about two minutes; the figures come from its internal deployment.
Chief Telecom (是方電訊) announced on Oct 20 that it has broken ground on an AI data center in the Huwei area of the Central Taiwan Science Park, with total investment of about NT$3.3 billion and operations targeted for the second half of 2028. The company says the facility is designed to Tier 4 standards with 16MW of total power capacity, support for both air and liquid cooling, an estimated PUE of 1.4, and diesel generators able to run at full load for more than 24 hours. Management says about 50% of capacity is already signed, mainly with large international cloud service providers and smart-finance customers, and is targeting profitability in 2031; all figures are company-provided. The site is intended to work with Chief Telecom's Taipei data centers and a partner renewable-energy data center in Ishikari, Japan, for cross-region backup and compute scheduling.
An analysis by 正见TrueView, republished by 36Kr and TMTPost, says Alibaba, Tencent and Baidu spent a combined 131.852 billion yuan on capital expenditure in Q2 2026, up 114% year on year. It argues that large-scale agent deployment is pushing cloud providers to rewrite stacks across chips, scheduling, storage, runtime and inference platforms, and that the three companies' combined free cash flow was a net outflow of 66.424 billion yuan in the quarter. It notes Alibaba Cloud CEO Wu Yongming's Sept. 22 announcement of a target for more than 20GW of datacenter capacity by 2032; Goldman Sachs estimated current capacity at about 5-6GW and projected related capex. Gartner separately warned that more than 40% of Agentic AI projects could be canceled by end-2027. The piece also cites Morgan Stanley's estimate that Chinese cloud providers' ROIC is in a 13%-20% range.
Google and Meta’s PyTorch team released TorchTPU as an open-source native backend for running PyTorch workloads on Google Cloud TPUs. It supports standard PyTorch APIs, eager-mode debugging and torch.compile optimization, with integrations spanning Hugging Face Transformers and Diffusers, vLLM, SGLang, llm-d, TorchTitan and LiteRT. Google says its training setup was validated on up to 8,000 Ironwood TPUs and achieved as much as 52% model FLOPS utilization with Qwen3-8B; these figures are vendor claims.
What problem it solves
Go applications need client libraries for interacting with Google Cloud Platform APIs. The project provides packages for those services and supports common authentication flows for local development and Google Cloud Platform environments.
Key features
Google Cloud API client packages — The repository contains Go packages for Google Cloud Platform services, including storage,…
What problem it solves
Developers and enterprise teams need a terminal-based agent that can understand and modify code, automate operational tasks, and integrate additional capabilities into development workflows. Gemini CLI also needs clear access requirements and authentication options for organizations using Gemini Code Assist or supported Google Cloud services.
Key features
Code understanding …
What problem it solves
Accessing Google Workspace APIs from the command line can require repetitive request construction, manual pagination, and custom tooling for agents. The project addresses this by generating its command surface from Google Discovery Service and producing structured JSON output across Workspace services.
Key features
Dynamic Workspace command surface — The CLI reads Google's D…
What problem it solves
The repository provides reusable Agent Skills for working with Google products and technologies. It covers areas including Google Cloud, AI/ML, infrastructure, databases and analytics, application development, security, identity, and advertising.
Key features
Broad Google technology coverage — Skills are organized across product and workflow areas, including Google Cloud, AI…