AWS GPU Shortage Pushes AI Startups to Cloud Rivals

NVIDIA expanded its NVLink Fusion platform with NVHBM, a custom high-bandwidth memory technology that integrates the memory controller into the HBM base die rather than the XPU die. This design delivers up to 30% greater memory bandwidth, 15% lower HBM power consumption, and frees 25% more compute area on the XPU compared to standard HBM4E. Amazon's Annapurna Labs will be the first to implement NVHBM in its next-generation Trainium4 chips, enabling closer integration between custom AI accelerators and NVIDIA GPUs.

Nvidia projects quarterly revenue of $108 billion in its next earnings report, up from a record $96.2 billion in the most recent quarter. The company's data center business, which generated $89 billion in the latest quarter, continues to drive growth. If realized, Nvidia would join Amazon, Apple, and Alphabet as companies that have exceeded $100 billion in quarterly revenue.

Zhipu's GLM-5.3-Flash model launched on OpenRouter at 7.5 to 25 cents per million tokens (promotional pricing), delivered entirely on Chinese infrastructure. The model scores 57 on Artificial Analysis' intelligence index at roughly nine cents per task, compared to GPT-5.6 Sol at 59 cents and Grok 4.6 at 94 cents, creating significant cost pressure on enterprise AI budgets already strained by unexpected consumption.

Amazon is committing to purchase an additional 2 million Nvidia GPU chips over the next two years, tripling its previous order as it responds to surging demand for AI infrastructure. The expanded partnership extends beyond chip procurement alone. The move underscores the intense competition for GPU capacity among cloud providers building out AI capabilities.