Z.AI closes the open-source gap but remains token hungry
GLM 5.2 is competitive enough to disprove the widening-gap narrative around Chinese open-weight models, particularly in coding and cybersecurity. Its weak point is economics: heav…
OpenRouter becomes the front door to the fragmented AI market
OpenRouter captures value across a fragmented model ecosystem by serving as the distribution layer for hundreds of models, gaining exposure to AI usage without bearing frontier-mo…
Meta's open-source strategy fails to commoditize frontier AI suppliers
Meta's Llama strategy has not eliminated dependence on closed frontier models: the company is now buying so much Gemini capacity that Google's limits have delayed internal project…
Frontier AI gives CrowdStrike a shrinking defensive head start
CrowdStrike is using frontier cyber models months before comparable capabilities reach open-source attackers, giving customers a window to harden systems. The narrowing model gap…
Frontier AI gives Palo Alto a shrinking defensive head start
Palo Alto Networks is using frontier cyber models months before comparable capabilities reach open-source attackers, giving customers a window to harden systems. The narrowing mod…
Gemini demand overwhelms Google's capacity despite record AI capex
Google cannot fully satisfy Meta's Gemini demand even after roughly $200 billion of capital expenditure, demonstrating strong pricing power and an infrastructure bottleneck consis…
Micron captures AI profits as memory scarcity drives historic pricing
Limited high-bandwidth-memory capacity is transferring economics from loss-making AI model providers to Micron. DRAM prices rose more than 60% sequentially and NAND more than 80%,…