Sent to 13 readers
Race to AGI Daily Digest - Monday, December 15, 2025
Share:
TLDR
China’s 16x surge in cloud large-model usage shows real AI demand behind the hype.
Broadcom and other AI infra names sold off hard as investors push for disciplined AI spending.
Chiiki Shimbun’s patented ‘psychological digital twin’ tech hints at a new, more invasive wave of AI-driven ads.
From militant groups testing AI to a Global South G20 op-ed, safety and fairness are becoming front-line issues, not side conversations.
Big cloud players are steering fresh AI capital into India, even as markets question AI ROI elsewhere.
The Full Story
Last week we talked about the AI hangover: after a year of massive GPU build-outs, investors were finally asking if all that spend would pay off. We wondered whether markets would start punishing vague AI stories and reward teams that show real usage and revenue instead.
Today, that tension is front and center. China’s CAICT says public-cloud large-model usage jumped 16x as the country’s AI core industry passes 1.2 trillion yuan full breakdown here ->. That’s real demand, with companies like ByteDance -> leaning hard into large models. At the same time, the market just took a chunk out of the AI infra names: Broadcom is down double digits, with AMD, Oracle, Intel, and TSM all sliding 4–5%. For a theme we’ve been tracking in “Investors Demand Discipline in AI Spending,” this looks like the first sharp reality check track that narrative here ->.
On the tech side, the weird is getting weirder. Japanese local media firm Chiiki Shimbun just won a patent on a generative-AI ‘psychological digital twin’ for ad targeting—basically a simulated you for selling you things patent story ->. That lands on the same day we get hard numbers on AI’s impact on work in Spain, plus a Global South G20 op-ed pushing for fairer, safer AI rules.
Safety is no longer an abstract lab topic either. OpenAI is flagging that militant groups are already experimenting with AI for propaganda and cyber operations, even as everyday users deal with outages and scary bugs—like Anthropic’s models going down and a Claude CLI that nuked someone’s home directory.
And in the background, the money is quietly shifting. Amazon and Microsoft are plowing undisclosed sums into India-focused cloud and AI infrastructure programs, a sign that the AI build-out is going more global even as investors demand cleaner stories on returns see the deal map ->.
So this week, we’ll be watching three things: whether real usage data can calm jittery AI investors, how far companies push into behavioral modeling like these ‘digital twins’, and how fast AI safety and fairness debates move from position papers into actual rules and product changes.
Get This Delivered Daily
Join thousands of AI professionals who start their day with Race to AGI.