Amazon & Google chip gains against Nvidia; Anthropic hit with $3B copyright lawsuit; new self-hosted agent.
Top engineers at Anthropic and OpenAI stated yesterday that AI writes 100% of their code, signaling a radical shift in how advanced AI products are built. Separately, music publishers sued Anthropic for $3 billion over alleged copyright infringement, raising critical questions for AI content generation and IP strategy. Meanwhile, Amazon and Google are reportedly eating into Nvidia's AI chip market share, a trend impacting future compute availability and pricing.
Amazon is reportedly negotiating to invest up to $50 billion in OpenAI. This potential investment is part of a larger funding round seeking $100 billion, which could value OpenAI at $830 billion.
Read full story→Google and Amazon are expanding their custom AI chip sales, challenging Nvidia's market share. Google's TPUs generated tens of billions in revenue, including a $21 billion deal to supply Anthropic with chips for non-Google data centers. Amazon's Trainium chips also saw revenue grow 150% quarterly, partly due to its $4 billion Anthropic investment. Nvidia still commands 92% of the market, but these trends signal a growing market for Nvidia alternatives, offering builders more diverse compute options as software support expands.
Major music publishers, including Universal Music Group, filed a $3 billion federal copyright lawsuit against Anthropic. They allege 'flagrant piracy' of over 20,000 copyrighted musical works, including songs and lyrics, used to train Claude AI models without authorization. A ruling could accelerate the adoption of licensed training datasets or push toward compulsory licensing, directly impacting how builders source and pay for AI training data.
Elon Musk's SpaceX, xAI, and Tesla are reportedly in early-stage talks for a potential merger, with scenarios including SpaceX combining with Tesla or xAI. This consolidation aims to align resources and could facilitate xAI's goal of placing data centers in space. Discussions follow recent corporate filings and $2 billion investments from SpaceX and Tesla into xAI.
Apple acquired Israeli AI startup Q.ai for nearly $2 billion, marking its second-largest acquisition. Q.ai's technology includes interpreting whispered speech and clarifying audio in noisy environments using imaging and machine learning. This acquisition could integrate Q.ai's audio AI into products like AirPods and Vision Pro, with the founding team joining Apple.
AlphaGenome is a deep learning model that processes 1 Mb of DNA to predict thousands of human and mouse genomic tracks across 11 modalities, including gene expression and splicing. It outperforms specialized models on 25 of 26 benchmarks for regulatory variant effect prediction. The model uses a two-stage training process (pretraining and distillation) for multimodal variant interpretation at single-base-pair resolution.
As AI shifts from GPU compute to inference, optical networking is now the primary bottleneck. Future AI data centers will require photon-based interconnects like Co-Packaged Optics (CPO) and Silicon Photonics for power efficiency and bandwidth, replacing copper. This transition shifts power dynamics across the supply chain.
A randomized trial found AI assistance slightly speeds coding tasks, though this difference was not statistically significant, but hinders skill mastery, especially debugging. Participants using AI scored 17% lower on a quiz testing new concepts. However, those using AI for conceptual inquiry and comprehension retained more knowledge, suggesting the interaction method matters for learning.
Cloudflare's Moltworker project runs the Moltbot AI agent on Cloudflare's edge, removing the need for local machines. It uses Workers, Sandbox SDK for code execution, R2 for storage, AI Gateway for model access, and Browser Rendering for web automation. This offers a model for deploying AI agents on the edge, providing efficiency, security, and observability.
Tesla's Q4 2025 earnings call revealed a pivot from traditional car manufacturing to 'transportation as a service' and robotics, discontinuing Model S/X and new mass-market vehicles. Electrek argues this decision sacrifices an $80 billion automotive business for unproven ventures, calling it 'automotive suicide' despite declining revenue and deliveries.
Unimpressive LLM answers often stem from generic prompts. This article offers six strategies for ChatGPT, Claude, and Gemini, including defining roles, providing detailed context, and asking the AI to 'think step by step'. Applying these methods results in more effective answers, reframing AI as a sharp colleague.