1. AI Coding Company Cognition Raises $1B, Valuation Hits $26B
Cognition, the developer of AI coding agent Devin, has closed over $1 billion in funding, more than doubling its valuation to approximately $26 billion in under nine months. The company reports an annualized revenue run rate of $492 million, marking one of the largest funding rounds in the AI coding agent space. This massive investment underscores the capital frenzy around AI coding agents, even as their real-world value remains hotly debated.
AI Pulse View: The valuation leap from $12B to $26B reflects the capital frenzy in the AI agent space. Notably, Cognition’s $492M annualized revenue versus its $26B valuation implies a 50x+ revenue multiple — meaning the market is pricing in future potential rather than current performance. When expectations are this high, delivery capability becomes the ultimate test.
Source: TechCrunch | 2026-05-27 Link: https://techcrunch.com/2026/05/27/ai-coding-startup-cognition-raises-1b-at-25b-pre-money-valuation/
2. Meta Launches Cross-Platform Subscriptions Covering Instagram, Facebook, and WhatsApp
Meta has officially rolled out its “Meta One” subscription brand globally, spanning paid plans for Instagram, Facebook, and WhatsApp, while testing additional offerings focused on AI features, creators, and businesses. This marks Meta’s strategic shift from a purely advertising-driven model toward a diversified revenue structure.
AI Pulse View: Meta’s integration of AI features into paid subscriptions explores a direct monetization path for AI capabilities. As users accustomed to free, ad-supported services are asked to pay for AI enhancements, it becomes a massive market experiment. This may signal a future where AI features transition from “universal tools” to “paid privileges.”
Source: TechCrunch | 2026-05-27 Link: https://techcrunch.com/2026/05/27/meta-officially-launches-instagram-facebook-and-whatsapp-subscriptions-with-more-to-come-including-ai-plans/
3. Snowflake Signs $6B Deal with AWS for AI Chips
Snowflake has inked a five-year, $6 billion agreement with Amazon to secure AWS AI chips, putting renewed competitive pressure on Nvidia. This deal highlights how cloud providers’ self-developed AI chips are reshaping the infrastructure landscape.
AI Pulse View: Snowflake’s choice of AWS self-developed chips over Nvidia GPUs is another signal of the de-Nvidia-ification trend in the AI chip market. As AWS Trainium and similar chips approach GPU-level cost-effectiveness, large customers gain more bargaining power. This competition will ultimately drive down AI computing costs across the industry.
Source: TechCrunch | 2026-05-27 Link: https://techcrunch.com/2026/05/27/in-more-good-news-for-amazon-snowflake-signs-6b-deal-with-aws-for-ai-cpu-chips/
4. Microsoft MAI-Image-2.5 Rivals Google on Image Generation Benchmarks
Microsoft’s MAI-Image-2.5 model ranks third on Arena’s text-to-image leaderboard, on par with Google’s Nano Banana 2, though still behind OpenAI’s Image-2 overall. The model shows notable improvements over its predecessor, particularly in text rendering within images and commercial visual generation.
AI Pulse View: The image generation competition is shifting from “who makes pretty pictures” to “who produces usable commercial assets.” MAI-Image-2.5’s progress in text rendering means AI image generation is crossing a critical threshold for commercial deployment — accurately rendering brand copy and logos.
Source: The Decoder | 2026-05-27 Link: https://the-decoder.com/microsofts-mai-image-2-5-pulls-even-with-googles-nano-banana-2-on-benchmarks/
5. Nvidia’s Annual Taiwan Spending Surges from $15B to $150B
Driven by explosive AI chip demand, Nvidia’s annual spending on Taiwan-based suppliers (including TSMC) has skyrocketed from $15 billion to approximately $150 billion — a tenfold increase. This underscores the AI hardware supply chain’s deep dependence on Taiwan’s semiconductor ecosystem.
AI Pulse View: A 10x spending increase in a decade reflects AI compute demand far outpacing what Moore’s Law alone can deliver. It also means global AI supply chain risk is highly concentrated in Taiwan — geopolitical factors have become a non-negligible systemic risk variable for the AI industry.
Source: The Decoder | 2026-05-27 Link: https://the-decoder.com/the-ai-boom-drove-nvidias-yearly-taiwan-spending-from-15-billion-to-150-billion/
6. arXiv Bans Submitters of AI-Generated Hallucination Papers for One Year
Preprint server arXiv announced it will impose a one-year submission ban on authors who submit AI-generated “hallucination” papers. The new policy aims to combat the growing influx of low-quality AI-generated content flooding academic platforms.
AI Pulse View: Academic publishing is becoming a frontline for AI abuse. arXiv’s ban represents the academic community’s first systematic counterattack against AI content pollution, but it raises a thorny question: how to distinguish “legitimate AI-assisted writing” from “AI ghostwriting producing academic garbage”? This boundary will define the new standard for academic integrity.
Source: Ars Technica | 2026-05-27 Link: https://arstechnica.com/science/2026/05/preprint-server-arxiv-will-ban-submitters-of-ai-generated-hallucinations/
7. AI Science Assistants Succeed in Drug Retargeting Tasks
Two AI-based science assistant tools have successfully completed drug retargeting tasks — one tool not only generated hypotheses but also performed data analysis. This marks the evolution of AI in drug discovery from “assisted brainstorming” to “end-to-end research execution.”
AI Pulse View: AI’s leap from “proposing hypotheses” to “verifying hypotheses” is one of the most significant directions in scientific AI. When AI can not only imagine new drug targets but also analyze experimental data to validate them, the cost and time compression in drug development becomes truly meaningful.
Source: Ars Technica | 2026-05-27 Link: https://arstechnica.com/science/2026/05/two-ai-based-science-assistants-succeed-with-drug-retargeting-tasks/
8. MiniMax Teases M3 Model: Sparse Attention Mechanism Delivers 15.6x Long-Context Speed Boost
Chinese AI company MiniMax has teased its upcoming M3 model, featuring a novel sparse attention mechanism that achieves a 15.6x speed improvement in long-context inference scenarios. This signals that Chinese AI companies continue to push forward with foundational architecture innovation.
AI Pulse View: Sparse attention mechanisms are a key pathway to breaking through Transformer computational bottlenecks. MiniMax’s progress in this direction reflects Chinese AI companies shifting from “chasing snapshots” to competing on “architecture-level innovation.” A 15.6x long-context speed boost, if validated in real-world scenarios, would have substantial impact on AI agents and document-processing applications.
Source: VentureBeat | 2026-05-27 Link: https://venturebeat.com/technology/minimax-teases-upcoming-m3-model-with-new-sparse-attention-mechanism-and-15-6x-response-speed-boost
Other Updates
- Samsung reaches AI profit-sharing agreement, employees to receive ~$400K bonuses each, benefiting from the memory chip boom driven by AI demand
- Pope Leo quotes Gandhi calling to “disarm AI” — the first time a religious leader has made such a strong public statement on AI safety
- Author caught fabricating book quotes with AI continues to defend AI-assisted writing, sparking debate on the boundaries of AI-augmented authorship
- Nvidia releases Polar framework for token-faithful rollouts in GRPO training, supporting optimization for coding agents like Codex, Claude Code, and Qwen Code