Today's Headlines

  • TSMC projects that AI-era semiconductors worth 1.5 trillion dollars will ripple out to more than 150 trillion dollars in global GDP
  • Surging AI data center power demand is raising US manufacturers' electricity bills and squeezing the Made in America plan
  • Anthropic expands Claude Cowork from desktop-only to mobile and web

What stood out today was that the conversation moved past individual new models toward the foundations beneath them — chips, power, and the place where work happens.

On one side sits a projection that spells out, in orders of magnitude, how large an economy the semiconductor industry underpins. On another, the power demand that AI itself drives is pushing up that same industry's costs and unsettling manufacturers. And on the work side, autonomous agents are leaving the developer's desk and spreading into everyday tasks.

Today's Top Three

TSMC Projects Chips Worth 1.5 Trillion Dollars Rippling to Over 150 Trillion in GDP

TSMC laid out a projection in which AI-era semiconductor demand ultimately ripples out to more than 150 trillion dollars in global GDP.

Speaking at TSMC's 2026 Japan Technology Symposium, Kevin Zhang, Senior Vice President and Co-Chief Operating Officer, said the global semiconductor market will pass 1 trillion dollars this year and grow beyond 1.5 trillion dollars by 2030. Of that 2030 total, HPC and AI would make up a majority at 55%, with mobile at 20% and automotive and IoT at 10% each.

Zhang framed the figure as a foundation on which industries stack. A 1.5 trillion dollar chip market underpins a 4 trillion dollar electronics sector, which in turn enables a 15 trillion dollar IT market, ultimately contributing to more than 150 trillion dollars in global GDP.

The engine behind that cascade, he argued, is AI changing shape. The shift from generative AI to agentic AI and then to physical AI operating in the real world is reshaping the very structure of the chip industry.

AI Data Center Power Demand Squeezes US Manufacturers — and Made in America

In the US, the surging power demand from expanding AI data centers is pushing up manufacturers' electricity bills and becoming a drag on efforts to bring production home.

The weight of it shows up on individual invoices. At Belden Brick, a brickmaker in Ohio, the monthly capacity charge jumped from 1,600 dollars to 12,000 dollars. Traditional manufacturers, less able to pass such increases into their prices, see their competitiveness eroded.

Behind this is a sharp spike in wholesale power prices. The capacity price in PJM, the grid operator covering much of the eastern US, rose roughly elevenfold, from 28.92 dollars per megawatt-day in 2024 to 329.17 dollars, with rising data center demand cited as the main cause.

The result is a contradiction. The Trump administration champions bringing manufacturing back home, yet the power costs pushed up by its AI policy are unsettling the very manufacturers it wants to return.

Claude Cowork Expands to Mobile and Web — Agents Step Beyond the Desk

Anthropic has expanded Claude Cowork, its autonomous work agent, from desktop-only to mobile and web.

Cowork launched in January as a desktop-only app, and this expansion reaches mobile and web first for Max subscribers, with availability for non-Max users expected within a few weeks. Users can start a task at their desk, check progress on a phone while out, and pick up finished output even with the laptop closed.

The actual usage is far from coding alone. In data Anthropic released on roughly 1.2 million sessions, business-process tasks accounted for 33.4% and content creation for 16.4%, while software development came in at just 8.7%. The numbers show the agent's terrain widening from the developer's hands into everyday office work.

TechCrunch frames the expansion as a move from a chatbot rivalry to a fight over control of the workspace itself. OpenAI is pursuing a similar path, and where agents take up residence in daily work is becoming the next front of competition.

More to Note

Models

  • Meta unveiled Muse Image, the first image generation model from Meta Superintelligence Labs, and integrated it into Meta AI. It is free to use in Meta AI and rolls out first in Instagram Stories and WhatsApp DMs. Users can mention a public account to pull that person's public photos into an AI-generated image, with an opt-out setting for those who prefer not to be included. - Introducing Muse Image: Image Generation Built for Your World (Meta Newsroom)
  • Microsoft has begun routing a portion of Excel and Word prompts to its in-house MAI models, reducing both reliance on and cost from third-party AI such as OpenAI and Anthropic. The move builds on the seven new MAI models unveiled at last month's Build and mirrors a broader cost-management trend seen at Amazon, Meta, and others. - Microsoft joins AI cost-cutting trend by relying more on its own models (TechCrunch AI)
  • Google expanded Managed Agents in its developer-facing Gemini API. New capabilities include asynchronous background execution, direct connections to remote MCP servers, custom function calling, and automatic credential refresh — all aimed at making production-grade agent deployments practical. - Expanding Managed Agents in Gemini API: background tasks, remote MCP and more (Google Blog)
  • SoftBank built an in-house gateway called Cloud Proxy to support its "100 agents per employee" initiative. It isolates credentials such as API keys from individual tenants and tracks usage token by token across multiple LLMs, already serving more than 20,000 users across 240 internal systems. - Inside SoftBank's in-house AI gateway powering its 100-agents-per-employee push (ITmedia AI＋)
  • The rise of open source AI is not hurting Anthropic for now, according to one read of the market. Decagon CEO Jesse Zhang argues frontier and open source models occupy different lifecycle stages rather than competing head-on. On Vercel's gateway, DeepSeek leads token volume with over 33%, yet Anthropic still commands more than half of total AI spend even after recent price hikes. - Why the rise of open source AI isn't hurting Anthropic … yet (TechCrunch AI)
  • Anthropic extended free-of-extra-charge access to its flagship model Claude Fable 5 through 3:59 p.m. July 13 Japan time. The extension covers Pro, Max, and Team plans as well as premium seats on seat-based Enterprise plans, letting users apply up to 50% of their weekly usage limit to Fable 5. - Free access to "Claude Fable 5" extended through July 13 (PC Watch)

Research

Policy

  • Discord acknowledged that a bug in its automated moderation system wrongfully banned more than 8,000 users over two months, flagging harmless images such as spreadsheets and chessboards as prohibited content. The flaw stemmed from a design that suspended accounts automatically before human review could catch the error. - Discord admits AI moderation bug wrongfully banned users over harmless images (TechCrunch AI)
  • zkSecurity's AI auditing pipeline, combining Claude Opus 4.6 and GPT-5.3 with custom skills, uncovered seven real bugs in Cloudflare's Circl cryptography library, including two rated Critical, all since fixed upstream. The AI's severity ratings often diverged from Cloudflare's own, and the models struggled to grasp how multiple bugs could compound into larger risks. - AI Meets Cryptography 1: What AI Found in Cloudflare's Circl (zkSecurity)

Business

Other

Source: Selected by the editorial team from the AI news inbox (collected July 8, 2026 — 29 items, 4 primary and 25 secondary).