Microsoft Keeps AI Capex Unchanged at $190B — First Hyperscaler to Hold the Line

On Wednesday, Microsoft kept its AI capex forecast unchanged at $190B for the calendar year (adjusted to $175B due to an accounting change), making it the first major hyperscaler to hold the line on spending. The stock jumped 8% on the news.
Context: The Industry's Spending Spree
Google, Amazon, Microsoft, and Meta had already planned to spend over $700B this year on AI data centers. Most competitors are still increasing their forecasts:
- Alphabet increased capex by $15B (stock fell sharply).
- Tesla bumped up its projection (stock fell).
- Meta narrowed its range, raising the midpoint by $2.5B.
What's Driving Costs?
Soaring memory chip prices are a major factor. Memory is a large component of data center costs, and higher prices have made it more expensive to build AI capacity. One research finding: higher memory prices explain about 45% of the growth in cloud companies' capex this year.
Microsoft's decision to keep spending flat may imply it's actually pulling back on building new capacity, since rising costs would otherwise push the dollar figure higher. For developers, this could signal a stabilization in AI compute supply — or at least a pause in the relentless expansion.
📖 Read the full source: HN AI Agents
👀 See Also
OpenClaw 2026.9.6 Adds Claude Opus 5.5, GPT-6 Sol and Luna, Grok 4.7, Plus Restart Recovery
OpenClaw 2026.9.6 ships in 2,614 PRs from 351 contributors, adding Claude Opus 5.5, GPT-6 Sol and Luna, and Grok 4.7 models, restart recovery for unfinished conversations, a 30-day Usage report, and live meeting notes.

Claude Code v2.1.139 Adds Agent View, /goal Command, and Major MCP Improvements
Claude Code v2.1.139 introduces a new agent view for session management, a /goal command for multi-turn tasks, expanded hook capabilities, and fixes for MCP server memory issues and terminal corruption.

Opus 4.6 Medium vs Low: Performance Differences and Pricing
Opus 4.6 medium costs approximately 50% more than the low version but addresses significant laziness issues found in the low-powered model. The medium version sits between low and high in performance benchmarks.

Meta's AI Agent Pivot: Zuckerberg Says Progress Not Fast Enough, $145B Spent
Meta CEO Mark Zuckerberg told staff that AI agent development hasn't accelerated as expected. The company spent $145B on AI infrastructure this year and laid off 8,000 employees to reorganize around AI.