3 Before 9: March 10, 2026

Two significant US federal AI deadlines fall today. The Department of Commerce must publish its evaluation of state AI laws that the Trump administration considers "onerous" and inconsistent with its deregulatory agenda. Simultaneously, the Federal Trade Commission must issue a policy statement clarifying when state laws requiring AI models to alter their outputs are preempted by the FTC Act. Both reports stem from an executive order signed in December 2025 titled "Ensuring a National Policy Framework for Artificial Intelligence." States that don't roll back laws flagged as burdensome risk losing access to up to $21 billion in unallocated BEAD broadband infrastructure funds. Colorado, California, and New York frameworks are all in the crosshairs. The DOJ's AI Litigation Task Force, established in January, is waiting on today's reports to begin filing challenges in federal court.

After missing mid-February, Lunar New Year, the Two Sessions window, and a succession of community-predicted dates, DeepSeek V4 has arrived in early March 2026 - now confirmed by multiple industry sources. The model is DeepSeek's first natively multimodal release, processing and generating text, images, and video from a single architecture rather than through bolted-on adapters. Core specs: roughly one trillion total parameters on a Mixture-of-Experts architecture with approximately 32 billion active per inference pass, a one million token context window powered by the Engram Conditional Memory system, and open weights under an Apache 2.0 licence. Critically, the model was optimised for Huawei Ascend and Cambricon chips, with DeepSeek explicitly withholding early access from Nvidia - a deliberate signal of China's parallel AI hardware push. Independent benchmark verification is still underway, but leaked internal figures point to around 90% on HumanEval and above 80% on SWE-bench Verified.

Meta and AMD announced a multiyear chip agreement worth up to $100 billion, structured around six gigawatts of AMD Instinct computing capacity with the first gigawatt deployment beginning in the second half of 2026. As part of the deal, AMD issued Meta a performance-based warrant for up to 160 million AMD shares - roughly a 10% stake - vesting against milestones and conditional on AMD's share price reaching $600, about three times its current level. Meta will receive AMD MI540 GPUs and next-generation CPUs to power AI inference workloads across its platforms. This follows a separate expanded Nvidia deal Meta struck weeks earlier, and comes as Meta has committed to spending up to $135 billion on AI infrastructure in 2026 alone as part of Mark Zuckerberg's push toward what he has called "personal superintelligence."