Race to AGI Daily Digest - Tuesday, December 30, 2025

TLDR

Meta’s multi‑billion Manus deal shows agents are now strategic M&A targets, not side projects.

The India AI Impact Summit is turning Delhi into a neutral meeting ground for Anthropic, NVIDIA, Salesforce, OpenAI and state actors.

India AI Impact Summit recap ->

China’s MIIT is baking AI and 800G into new internet exchange centers, hard-coding model workloads into national infrastructure.

China MIIT AI+800G directive ->

ACCA scrapping remote exams is an early example of professional bodies treating AI cheating as an operational risk, not an edge case.

ACCA remote-exam decision ->

US–China tension over NVIDIA’s H200 exports is crystallizing into a narrative where AI chips are treated as controlled strategic assets.

AI chip export narrative ->

Stocks tied to narrative-heavy AI stories like Tesla, Alibaba, and Palantir fell, while Baidu and Intel gained, reinforcing a slow shift toward infrastructure bets.

Browse AI-exposed companies ->

The Full Story

Building on Monday’s Nvidia–Groq licensing move, the arms race is sliding up the stack from chips to agents. Meta’s multi‑billion purchase of Manus, an AI agents startup, is basically an acqui-hire for autonomy. These aren’t just chatbots; Manus has been focused on software that can plan, call tools, and act across long workflows. That’s the same playbook we saw in hardware yesterday: if you can’t out-iterate the specialists, buy them Meta–Manus agents deal ->. Now zoom out. The India AI Impact Summit pulled in leaders from Anthropic, NVIDIA, Salesforce, and OpenAI, plus delegations from the US and China India AI Impact Summit recap ->. Yesterday we talked about licensing and acqui-hires in hardware; today it’s clear that whole countries are doing the same thing at the national level, trading market access and data for capital and compute. On the infrastructure side, China’s MIIT ordered new internet exchange centers to bake in AI workloads and 800G networking from day one China MIIT AI+800G directive ->. Pair that with the U.S. decision to allow NVIDIA’s H200 exports to China with a 25% fee, and you get a hardware race that’s now explicitly about national security AI chip export narrative ->. The frontier-safety-as-operations storyline is moving too. ACCA scrapping remote exams because of AI-powered cheating is not a research paper; it’s an ops incident ACCA remote-exam decision ->. That lines up neatly with OpenAI’s new risk management push, where safety is run like a business function instead of a whitepaper topic OpenAI risk management storyline ->. If you want a vibe check, look at Hacker News. One top post is literally titled “LLMs Are Not Fun,” another complains that Meta’s ad tools are quietly swapping out best-performing creatives for AI-generated ones. Users are saying: the tools are powerful, but they’re also starting to feel weird. Markets are still sorting winners and losers. Tesla, Alibaba, and Palantir sold off sharply, while Baidu and Intel ticked up. That’s a small but consistent signal: investors keep nudging money toward infrastructure and regional power centers, while they wait to see which AI narratives actually convert into durable products.

Want more?

Explore 100 AI companies →Track emerging AI trends →View AI deals graph →

Dec 29, 2025

Get This Delivered Daily

Join thousands of AI professionals who start their day with Race to AGI.