AI Briefing

Your next phone is about to get more expensive because AI ate all the memory chips. Data centers are now consuming 70% of global memory production, and PC and smartphone makers are warning of 15-20% price hikes with no relief until 2028. Meanwhile, three US states quietly passed AI laws on the same day, MIT figured out how to make AI models shed their own dead weight, and Q1 2026 venture funding hit a staggering $300 billion... with AI swallowing 80% of it.

01IndustryWindows Central ↗

AI Data Centers Are Devouring the World's Memory Chips, and Your Next Phone Will Cost More Because of It

The insatiable demand for high-bandwidth memory (HBM) from hyperscalers like Microsoft, Google, Meta, and Amazon has forced Samsung, SK Hynix, and Micron to pivot production toward higher-margin enterprise chips, creating a global shortage that's now hitting consumers. IDC projects global smartphone shipments will fall 12.9% in 2026, the sharpest annual drop since the 2008 financial crisis, while PC vendors Lenovo, Dell, HP, Acer, and ASUS have confirmed 15-20% price hikes. Up to 70% of all memory chips produced globally in 2026 will be destined for AI data centers, up from 32% five years ago. Intel CEO Lip-Bu Tan says there's no relief until 2028, when new fab capacity finally comes online, meaning this isn't a blip but a structural reshaping of the entire electronics supply chain around AI's appetite.

02PolicyTroutman Pepper ↗

Nebraska, Maryland, and Maine All Pass AI Laws on the Same Day, Targeting Chatbots, Surveillance Pricing, and AI Therapy

Three US states passed AI-focused legislation in the same week, each tackling a different angle of AI governance. Nebraska's LB 525 includes the Conversational AI Safety Act, which regulates how minors interact with AI chatbots and requires services to disclose they're not human when a reasonable person wouldn't know. Maryland's HB 148 bans surveillance pricing, where companies use AI to charge different customers different prices based on personal data. Maine's LD 2082 prohibits anyone from offering therapy or psychotherapy services through AI unless provided by a licensed professional. Together, these bills represent the most concrete state-level AI regulation to date, with each addressing a distinct real-world harm rather than trying to regulate AI broadly.

03ResearchMIT News ↗

MIT's CompreSSM Algorithm Teaches AI Models to Shed Their Own Dead Weight While Still Learning

Researchers at MIT CSAIL developed CompreSSM, an algorithm that prunes unnecessary complexity from AI models during the training process itself rather than after the fact. The technique allows models to find their own most efficient internal structure while simultaneously cutting compute costs, producing leaner and faster models without sacrificing accuracy. This is a meaningful departure from the standard approach of training a massive model first and then compressing it, because CompreSSM integrates pruning into the learning loop. As the industry grapples with the spiraling costs of training frontier models, techniques that reduce computational overhead without quality loss are becoming increasingly strategic.

04FundingCrunchbase News ↗

Q1 2026 Venture Funding Shatters Every Record at $300 Billion, with AI Claiming 80% of All Investment

Global venture funding hit $300 billion in Q1 2026 across 6,000 startups, up over 150% both quarter-over-quarter and year-over-year, with AI accounting for a staggering $242 billion or 80% of the total. Four of the five largest venture rounds ever recorded closed in a single quarter: OpenAI ($122B), Anthropic ($30B), xAI ($20B), and Waymo ($16B). The concentration is unprecedented: foundational AI startups alone received double what they raised in all of 2025. Humanoid robotics emerged as the breakout category, projected to draw over $20 billion in funding this year. The numbers paint a clear picture: venture capital is no longer diversified across tech sectors but has become, for all practical purposes, an AI funding machine.