AI & Hardware desk · AI & ML, Hardware

Maya Iverson

Covers AI models and the chips underneath them, from the model cards and spec sheets.

Maya covers the technology layer of AI and the silicon underneath it. She's the byline you want when a model gets released, a benchmark gets gamed, or a chip vendor changes the math on inference economics.

Articles by Maya Iverson

Micron Beat Q3 Hard, Guided Q4 to a $50B Midpoint. The HBM Cycle Is Real.
Hardware & Devices

Micron Beat Q3 Hard, Guided Q4 to a $50B Midpoint. The HBM Cycle Is Real.

Micron reported Q3 revenue of $41.46 billion (vs $35.84B consensus) and EPS of $25.11 (vs $20.39). The Q4 guide came in at $49-51 billion versus the $43.2B street estimate.
Daily Roundup, June 16: Capital, Talent, and Regulation All Moved at Once
Big Tech

Daily Roundup, June 16: Capital, Talent, and Regulation All Moved at Once

Eight stories in three buckets: Anthropic's flagship got shut down by Commerce, SpaceX raised the largest IPO in history and spent $60B on Cursor in five days, and Google's two most decorated AI researchers walked.
Cursor Was a Top-Tier Anthropic Customer. SpaceX Just Bought It.
AI & ML

Cursor Was a Top-Tier Anthropic Customer. SpaceX Just Bought It.

Cursor, the AI coding tool SpaceX just acquired for $60 billion, ran the bulk of its inference on Claude. The integration roadmap under Musk decides what happens to that revenue line.
Google's AI Brain Drain: Shazeer to OpenAI, Jumper to Anthropic
AI & ML

Google's AI Brain Drain: Shazeer to OpenAI, Jumper to Anthropic

Noam Shazeer, the Google VP behind the original transformer paper, is leaving for OpenAI. DeepMind VP John Jumper, the AlphaFold Nobel winner, is leaving for Anthropic. Google stock fell on the news.
Microsoft's Seven In-House AI Models Just Became Risk Management
AI & ML

Microsoft's Seven In-House AI Models Just Became Risk Management

Microsoft launched seven in-house AI models at Build 2026 to reduce OpenAI and Anthropic reliance. The Fable 5 takedown turned that hedge into a procurement requirement.
Apple Shipped Siri AI by Routing It Through Google Gemini
AI & ML

Apple Shipped Siri AI by Routing It Through Google Gemini

At WWDC 2026, Apple revealed that AFM Cloud Pro (the frontier tier of its Foundation Models) runs on Google's Gemini, plus a developer protocol that swaps Apple, Google, and Anthropic models behind one API.
Marvell Jumped 21% After Jensen Called It the Next Trillion-Dollar Company
Hardware & Devices

Marvell Jumped 21% After Jensen Called It the Next Trillion-Dollar Company

Marvell shares popped 21% after Nvidia CEO Jensen Huang singled it out onstage at Computex Taipei. The custom-silicon connection between the two explains the move.
Nvidia's New Vera CPU Targets Agent Workloads at 88 Cores and 1.2 TB/s
Hardware & Devices

Nvidia's New Vera CPU Targets Agent Workloads at 88 Cores and 1.2 TB/s

Jensen Huang unveiled Vera at GTC Taipei on June 1. It is Nvidia's first CPU purpose-built for AI agent workloads, with 88 custom Arm cores and 1.2 TB/s of memory bandwidth.
Lenovo Hit a Record $21.6B in Q4. AI Server Revenue Jumped 84%.
Hardware & Devices

Lenovo Hit a Record $21.6B in Q4. AI Server Revenue Jumped 84%.

Lenovo posted record Q4 revenue of $21.6 billion, AI revenue grew 84%, and the stock has more than doubled in May. AI hardware demand is spreading past the hyperscalers.
Micron Hit a $1 Trillion Market Cap. Memory Is the AI Bottleneck Now.
Hardware & Devices

Micron Hit a $1 Trillion Market Cap. Memory Is the AI Bottleneck Now.

Micron shares jumped about 18% to push its market cap to $1 trillion, as HBM demand for AI accelerators reprices the memory business.
Mistral Is Building a French Data Center and Defending AI in Defense
AI & ML

Mistral Is Building a French Data Center and Defending AI in Defense

Mistral CEO Arthur Mensch backed AI for defense and announced a 10MW data center in Les Ulis, France, framing both as European sovereignty plays.
Blackwell, Explained: Nvidia's Current-Generation AI GPU Architecture
Hardware & Devices

Blackwell, Explained: Nvidia's Current-Generation AI GPU Architecture

Blackwell is the GPU architecture behind Nvidia's B200, GB200, and Blackwell 300 chips that now drive most of its data-center revenue. Here is what is in it and why it matters.