Top Stories
Nvidia GTC 2026 Begins Tomorrow
Nvidia’s biggest event of the year kicks off March 16 in San Jose, with CEO Jensen Huang’s keynote expected to unveil significant announcements across chips, software, and AI applications.
Key expectations include updates on the Vera Rubin microarchitecture (Blackwell’s successor), details on Nvidia’s rumored laptop CPUs (N1 and N1X) targeting Windows on Arm gaming, and the debut of NemoClaw, an open-source platform for deploying AI agents across enterprise systems.
The company has been pitching NemoClaw to Salesforce, Cisco, Google, Adobe, and CrowdStrike as it positions itself for the agentic AI wave. Huang will also cover Vera Ultra (slated for late 2027) and the Feynman GPU architecture planned for 2028.
Source: NVIDIA GTC 2026 blog
Meta Planning 16,000 Layoffs to Fund AI Infrastructure
Meta is weighing layoffs affecting up to 20% of its workforce (roughly 16,000 employees) to offset costs from a planned $600 billion AI infrastructure investment through 2028.
The cuts aim to prepare for “greater efficiency brought about by AI-assisted workers,” according to Reuters. This follows Meta’s pattern of using AI as justification for workforce reductions, joining Block and Atlassian in the trend.
Source: CNBC
March 2026 Tech Layoffs Hit 45,000
The technology sector has shed 45,000 jobs in March alone, with over 9,200 attributed directly to AI and automation. Seattle leads with 16,590 affected employees, followed by San Francisco (9,395) and Menlo Park (1,500).
Notable cuts include Atlassian (10% workforce, ~1,600 jobs) and Oracle (potentially 20,000-30,000 positions, or 12-18% of global staff). The pattern is clear: companies are citing AI efficiency gains while restructuring to fund AI investments - often the same AI they claim is making workers redundant.
Source: TechNode Global
Morgan Stanley: AI Breakthrough Coming First Half 2026
Morgan Stanley’s “Intelligence Factory” model predicts a significant AI breakthrough in the first half of this year, driven by unprecedented compute accumulation at major US AI labs. OpenAI’s GPT-5.4 “Thinking” model reportedly scores 83.0% on the GDPVal benchmark, at or above human expert level on economically valuable tasks.
However, infrastructure is struggling to keep pace. Morgan Stanley projects a net US power shortfall of 9-18 gigawatts through 2028, a 12-25% deficit. Developers are converting Bitcoin mining operations, firing up natural gas turbines, and deploying fuel cells to stay ahead.
Source: Fortune
Quick Hits
-
DeepSeek V4 imminent: The trillion-parameter MoE model with 1M-token context window and native multimodal capabilities is expected within days, optimized for Huawei Ascend chips rather than Nvidia GPUs. AwesomeAgents
-
Apple M5 Pro/Max ship: New MacBook Pros deliver up to 4x AI performance vs M4, with Neural Accelerators in every GPU core enabling LLMs to run entirely on-device. Apple Newsroom
-
Block’s AI layoff skepticism: Jack Dorsey cut 40% of Block’s workforce (4,000 jobs) citing AI efficiency, but insiders and analysts question whether AI is driving cuts or just providing cover for cost-cutting. Bloomberg
-
NemoClaw goes open source: Nvidia’s upcoming AI agent platform will allow companies to deploy agents across enterprise systems, with partnerships being negotiated with major software vendors. CNBC
-
Qwen 3.5 small models: Alibaba launched compact LLMs positioned as “more intelligence, less compute,” targeting edge deployments and lightweight agents. LLM Stats
Worth Watching
The AI jobs narrative is getting complicated. As companies simultaneously claim AI makes them more efficient while laying off workers to fund AI investments, expect pushback. The Block controversy - where insiders suggest AI is cover for standard cost-cutting—may be the tip of the iceberg.
Nvidia’s GTC keynote Monday will set the tone for AI infrastructure in 2026. Watch for NemoClaw details, Vera Rubin architecture specs, and any surprise consumer hardware announcements.
DeepSeek V4’s timing is calculated, launching during China’s Two Sessions parliamentary meetings. If performance claims hold up, it represents significant progress on non-Nvidia hardware, with implications for chip export controls.