Pentagon pressures Anthropic while Meta locks in a massive AI chip deal.
$650 Billion, Military Pressure, and a Power Problem
Good morning.
This week, AI stopped feeling abstract. The Pentagon demanded access. Big Tech committed $650 billion. The power grid showed strain. The Fed warned about jobs.
Policy, capital, energy, and labor are colliding at once.
Here’s what you need to understand.
Read Time: 5-mins
TOP STORIES
Pentagon pressures Anthropic over military access
The United States Department of Defense gave Anthropic a deadline to expand lawful military model access or risk losing a $200M contract.
This forces a choice. Can AI companies restrict defense use while taking federal money? The tension between safety policies and national security just moved into the open.
Big Tech prepares a $650B AI buildout
Bridgewater Associates says Alphabet, Amazon, Meta, and Microsoft could invest $650B in AI infrastructure in 2026. That’s up sharply from 2025.
This scale boosts compute capacity fast. It also raises execution risk. If demand slows, returns get pressured. If demand holds, AI accelerates.
Meta locks in AMD to reduce Nvidia dependence
Meta signed a multi-year AI chip agreement worth up to $100B with Advanced Micro Devices.
Shipments begin in 2026. The deal includes potential equity upside for AMD.
This weakens Nvidia’s dominance and reshapes hyperscaler hardware strategy. The AI chip market is no longer single-supplier dependent.
AI expansion hits power grid limits
A Reuters analysis shows turbine shortages, long grid connection queues, and dozens of data centers planning their own power plants.
AI now competes directly with households and industry for electricity. Infrastructure, not algorithms, may slow deployment.
Meme of the week
White House tells Big Tech to build its own power
Donald Trump said tech firms must self-supply electricity for AI data centers.
That shifts cost responsibility away from consumers. It also confirms something bigger. AI infrastructure is now a national economic and political issue.
Fed warns AI may raise unemployment first
Federal Reserve Governor Lisa Cook said AI could increase unemployment in the short term before productivity gains appear.
This is the first clear macro warning. Displacement may precede job creation. Labor markets could feel pressure sooner than expected.
POLL OF THE WEEK
PRODUCT UPDATE
Claude launches enterprise agents
Anthropic rolled out department-specific plugins and integrations for Slack, Gmail, and DocuSign.
This moves Claude from chatbot to workflow automation tool. If reliability holds, internal repetitive tasks become prime targets for replacement.
TOOLS TO WATCH
Grok expands defense access
xAI expanded Grok model access for U.S. defense use.
While some AI labs debate restrictions, others lean in. Military AI partnerships are becoming a strategic positioning move.
Claude Code Security
Anthropic introduced a code vulnerability scanning tool powered by model reasoning.
It reviews repositories and suggests fixes. If accurate, it compresses parts of the cybersecurity workflow and pressures traditional scanning vendors.
NEW DEVELOPMENT
Anthropic alleges large-scale model distillation
Anthropic accused three Chinese AI labs of extracting model capabilities through massive prompt activity using fake accounts.
If confirmed, this exposes how difficult it is to protect frontier models. Competitive advantage may leak faster than expected.
If this helped you understand the week in AI, share it with someone who follows tech closely. See you next Thursday.
And before you go, give your feedback below.