TSMC is accelerating the adoption of wind power to meet the immense energy demands of manufacturing AI chips, highlighting the challenges of balancing rapid technological growth with local power infrastructure limits in Taiwan.
Jason··2 min read
The AI market is pivoting toward infrastructure and enterprise application-layer dominance, evidenced by Cerebras's planned IPO and Sierra's massive $950M funding round.
Jason··2 min read
AI chip manufacturer Cerebras and geothermal energy startup Fervo Energy are both moving toward blockbuster IPOs, reflecting strong investor interest in core AI infrastructure.
Jason··1 min read
Nvidia-backed SiFive reached a $3.65 billion valuation, as its open-standard RISC-V chip designs gain momentum against traditional architectures in the AI hardware race.
Jason··1 min read
Amazon is investing $50 billion in OpenAI and expanding its custom 'Trainium' AI chip strategy. With major adopters like Apple and Anthropic, Amazon is leveraging vertical integration to strengthen its AI cloud infrastructure.
Jason··2 min read
Despite projecting $1 trillion in AI chip sales at its GTC conference, Nvidia's response from Wall Street has been muted, highlighting investor concerns over an AI bubble.
Jason··2 min read
Nvidia unveiled its next-generation Vera Rubin AI platform at GTC 2026, featuring a complex seven-chip architecture supported by industry leaders like OpenAI and Meta. CEO Jensen Huang projected a $1 trillion market for these chips. Additionally, Nvidia introduced DLSS 5, leveraging generative AI to revolutionize real-time graphics and visual fidelity.
Jason··3 min read
Nvidia unveiled the 'Vera Rubin' platform at GTC 2026, with CEO Jensen Huang projecting $1 trillion in sales. Supported by OpenAI and Meta, the platform offers 10x better inference efficiency. Nvidia also introduced DLSS 5, a generative AI-based graphics technology that marks a 'GPT moment' for the gaming and rendering industry.
Jason··3 min read
Meta has introduced four new custom MTIA AI chips designed to power its recommendation engines and Llama model fine-tuning. This strategic launch aims to reduce the company's multi-billion dollar reliance on Nvidia GPUs and improve data center efficiency. While Meta continues to buy Nvidia hardware for massive training tasks, it expects custom silicon to handle 40% of its inference load by 2027. This move highlights the accelerating trend of 'hyperscalers' becoming major players in semiconductor design.
Jason··2 min read