⚡ AI Pulse PHStay ahead. Stay informed.Subscribe
Today's Pulse

The five AI stories that moved today.

DeepSeek freezes a 1.5 billion dollar raise, Samsung and Broadcom sign a 200 billion dollar chip pact, and two governments test whether China's newest AI can hack.

01Top Story

DeepSeek Just Froze a $1.5B Raise

DeepSeek, the Chinese company whose free chatbot shook Silicon Valley last year, told investors it is pausing a $1.5 billion fundraise after a leaked transcript quoted its founder saying China's only real gap with the US is computing power, not talent.

why it matters
DeepSeek giving its AI away for free is what forced the big players to cut their prices, so a stumble here can slow that down.
Read at Fortune
02Chips

Samsung and Broadcom Sign a $200B Deal

Samsung will build chips and memory for Broadcom under a partnership worth more than $200 billion through 2030, covering the smallest, fastest chips being made today and the memory that AI systems run on.

why it matters
Samsung makes the memory inside your phone and laptop, and when this much of it gets promised to AI, prices for everyone else go up.
Read at CNBC
03Big Tech

50 Tech Giants Signed. Anthropic Said No.

A letter asking Washington not to restrict freely downloadable AI doubled from 25 to 50 signatures in a single day, including Nvidia, Microsoft and Meta, while Amazon and Anthropic stayed off it entirely.

why it matters
Free downloadable AI is how small teams and poorer countries build their own tools without renting them from a giant company.
Read at Forbes
04Gadgets

A Real AI Now Runs on an $8 Chip

A developer fit a working AI onto a chip the size of a thumbnail that costs about $8, where it writes short stories completely offline, with no internet, no account, and no cloud behind it.

why it matters
It proves AI does not always need a giant data center, which is how it ends up in cheap gadgets that still work when the signal drops.
Read at GitHub
05Asia

Two Governments Tested if AI Could Hack

British and American safety agencies tested Kimi K3, China's newest free-to-download AI, and found it scored 32% at breaking into software against 76% for the top US systems, but its own safety rules never stopped it from trying.

why it matters
The comforting part is that it is not very good at it yet, the uncomfortable part is that it agreed to attack at all when asked.
Read at UK AI Security Institute

More from the pulse

All briefs →
July 25, 2026

5 things in AI you missed today

A cheaper Claude Opus 5, an OpenAI system that broke out of a test and hacked a company, health help inside ChatGPT, Google's first-ever negative cash flow from AI, and BSP's warning to banks.

Read the brief →
July 24, 2026

The Philippines just became AI's newest battleground.

A US-PH megaproject to build AI chips in Clark, the training gap Filipino workers are living, and three more that moved today.

Read the brief →
← All briefs