Thursday, 2 April 2026
Google Sells AI Video at $0.05/Sec, Meta Learns to Read Your Brain
Nvidia Bets $2 Billion on Marvell — A New Layer in the AI Infrastructure War
Nvidia has announced a $2 billion strategic investment in Marvell Technology to accelerate AI networking infrastructure and counter AMD’s growing competitive pressure. Large language models don’t just need fast chips — they need fast connections between them. Marvell, now backed by Nvidia’s capital, is positioned to build exactly that. The AI race is now being fought at every level of hardware simultaneously.
AMD’s MI355X Breaks 1M Tokens/Sec at MLPerf — A New Inference Record
AMD’s MI355X GPU has surpassed 1 million tokens per second on the MLPerf industry benchmark — a new record in AI inference performance. The result confirms that the Nvidia vs. AMD competition has expanded beyond pricing into raw capability. Data centers now face a genuine choice: platform loyalty or chasing the latest benchmark winner?
Anthropic Signs AI Safety Partnership with the Australian Government
Anthropic has announced a strategic partnership with the Australian government covering AI safety research and economic impact tracking. The agreement includes continuous monitoring of AI’s effect on the national economy and joint development of safety standards. Australia is becoming an increasingly active player in AI governance — without Europe’s bureaucratic noise, but with quiet strategic precision.
Google Veo 3.1 Lite: AI Video Generation Now Costs $0.05/Second — 50x Cheaper
Google has launched Veo 3.1 Lite, cutting AI video generation pricing to $0.05 per second — a 50x reduction from previous-generation models. In practice, this opens the technology to small creative studios, research teams, and independent journalists who previously couldn’t afford it. The democratization of AI video isn’t happening through simplified tools — it’s happening through price collapse. That distinction matters.
Meta TRIBE v2: AI Now Predicts Your Brain’s Reaction to Video and Audio
Meta’s TRIBE v2 system marks a new phase in neuroscience research: AI can now predict in advance how the human brain will respond to specific video or audio content. The technology opens pathways to neuro-personalized media — content shaped before viewing, based on predicted neurological response. The line between neuroscience and AI is increasingly blurred, and the “attention economy” is acquiring a new dimension entirely.