One-Line Summary

Big Tech doubles down on AI efficiency: Google moves on inference chips while Adobe and Siemens ship agents, as Nvidia inks massive supply deals and AWS expands Anthropic’s capacity.

Big Tech

Google Eyes New Chips to Speed Up AI Results, Challenging Nvidia

Google is preparing chips dedicated to inference, the phase where AI models answer queries after training, aiming to build on momentum from deals with Meta and Anthropic, according to Bloomberg. That positions Google to challenge Nvidia in a fast‑growing slice of the semiconductor market tied to running AI software. ¹

Bloomberg reports that leading AI developers, including some Google rivals, are already buying the company’s AI chips, and that the inference push is about speeding up results and meeting surging demand for AI applications. ¹

Legal‑industry coverage of the Bloomberg piece underscores the same theme: as demand rises for quickly processing AI queries, Google’s focus on inference hardware signals a bid to cut latency and cost at scale. ²

On the consumer side, Google is also rolling out new touch‑up tools in the Google Photos app — heal, smooth, under‑eyes, irises, teeth, eyebrows and lips — on Android devices with at least 4 GB RAM and Android 9.0 and up, showing continued product‑level AI upgrades alongside infrastructure moves. ³

Adobe launches AI suite for corporate clients as competition heats up

Adobe is releasing CX Enterprise, a suite of AI agents to help companies automate and personalize digital marketing; shares rise 2.2% in morning trading even as the stock is down about 30% year‑to‑date, Reuters reports. ⁴

Adobe says CX Enterprise — including an agent called Coworker — can coordinate other agents, assemble relevant business data to craft a marketing plan, and execute it. The company is partnering with Amazon, Microsoft, Anthropic, OpenAI and Nvidia so customers can use agents across those platforms. ⁵

Launched at Adobe’s digital marketing conference, the offering is pitched as a governed, enterprise‑ready path to agentic AI amid questions about AI‑native rivals and the durability of legacy software models. ⁵

Industry & Biz

AI startup Thinking Machines clinches capital and a major chip supply deal from Nvidia

Thinking Machines Lab secures a multi‑year partnership with Nvidia that includes a significant investment and at least 1 gigawatt of next‑generation processors (starting with Vera Rubin systems early next year) to train its AI models; the company was founded by former OpenAI CTO Mira Murati. ⁶

Reuters adds that 1 gigawatt of compute — enough to power roughly 750,000 U.S. homes — can cost around $50 billion. Thinking Machines earlier raised about $2 billion at a $12 billion valuation, and Nvidia has also financed the broader ecosystem, including $30 billion in OpenAI and $10 billion in Anthropic. ⁶

Separate analysis points to record capital flowing to AI‑chip challengers — $8.3 billion in 2026 per Dealroom — as startups pitch power‑ and cost‑efficient inference architectures. Nvidia, meanwhile, bought Groq assets for $20 billion in Dec 2025, invested $4 billion in photonics in Mar 2026, and spent more than $18 billion on R&D in the fiscal year ending Jan 2026. ⁷

In parallel, Amazon agrees to invest up to $25 billion in Anthropic, while Anthropic commits to spending more than $100 billion on AWS technologies over 10 years and secures up to 5 gigawatts of capacity; $5 billion arrives now with up to $20 billion later, and Amazon projects roughly $200 billion in 2026 capex, mostly for AI infrastructure. ⁸

New Tools

Siemens launches AI engineering agent to automate PLC coding and industrial workflows

Siemens releases the Eigen Engineering Agent, an AI designed to autonomously execute industrial automation work — including programmable‑logic‑controller (PLC) coding, human‑machine‑interface (HMI) visualization and device configuration — inside its TIA Portal engineering environment. ⁹

Siemens says the agent completes workflows two to five times faster than manual work, with up to 80% higher solution quality and 50% greater engineering efficiency; it validates outputs step‑by‑step and has been piloted with 100+ companies in 19 countries. ⁹

At Hannover Messe, Rockwell Automation demonstrates an AI‑orchestrated workflow that connects digital‑twin simulation, AI‑assisted code generation and validation into one pipeline, reinforcing a shift toward AI‑native engineering. ¹⁰

Treon unveils an AI‑native Maintenance Orchestration Layer to automate ticketing, scheduling and technician dispatch across asset fleets, signaling demand to embed “agentic” automation deeper into operations. ¹¹

What This Means for You

As inference becomes the focal point for chip makers, teams that depend on fast AI responses — customer support, product search, real‑time personalization — gain leverage to expand use without linear headcount growth. This also suggests re‑checking your inference unit economics as new hardware options appear. ²

For marketers and CX leaders, Adobe’s CX Enterprise indicates that AI agents are moving into governed, enterprise deployments that plug into stacks from AWS, Microsoft, Anthropic, OpenAI and Nvidia. That opens room to trial automation for segmentation, offers and journey orchestration without replacing core systems. ⁵

In operations and manufacturing, Siemens’ generally available agent — alongside peer previews at Hannover Messe — points to a practical path to offload repetitive PLC edits, HMI updates and maintenance coordination, refocusing scarce engineers on system architecture and safety review. ⁹

Action Items

1. Test Google Photos’ new touch‑up tools: On an eligible Android device, open Google Photos, select a face in a picture, try heal/smooth/under‑eyes/irises/teeth/eyebrows/lips, and adjust intensity to learn what’s possible.
2. Book an Adobe CX Enterprise demo: If you run marketing or CX, ask your Adobe rep for a pilot and pick one contained workflow (e.g., email segmentation) to compare agent output vs. your current process.
3. Scope two agentic tasks in operations: Meet your automation lead to shortlist 2–3 repetitive tasks (PLC scaffolding, HMI tweaks, device configs) and prepare vendor questions on standards, safety and approvals.
4. Price‑check inference costs: Ask your cloud account team for current per‑1,000‑request pricing on your most‑used model and any discounts for alternative chips; document potential savings for one live feature.
5. Try Claude Design for a quick artifact: Generate a one‑page doc and a slide draft with Claude Design and compare speed, accuracy and brand compliance against your current creative workflow.