Computing Power

Nvidia CEO Jensen Huang asserts that substantial tech investments in AI infrastructure are justified and sustainable. He links these capital expenditures to projected cash flow increases, driven by AI’s transformative capabilities. Major tech companies are expected to spend $660 billion on AI infrastructure this year, with a significant portion for Nvidia chips. Huang calls this the largest infrastructure buildout in history, with demand for computing power “sky high” as AI unlocks new revenue and operational efficiencies.

Cerebras Systems inked a $10 billion deal with OpenAI for 750 megawatts of computing power through 2028. This landmark agreement diversifies Cerebras’ client base beyond G42 and aims to provide OpenAI with a dedicated, low-latency inference solution for faster AI interactions. The partnership stems from prior technical collaborations and positions Cerebras as a contender against established players like Nvidia in the AI chip market.

Michael Dell acknowledges booming AI server demand while anticipating a potential saturation point. Fueled by AI’s evolution, Dell’s server and networking business surged, with AI servers integrating Nvidia chips used by CoreWeave and xAI. Dell raised revenue and EPS growth expectations, projecting $20B in fiscal 2026 AI server shipments. Power consumption is a key challenge, as highlighted by OpenAI’s energy-intensive data center plans, requiring focus on energy-efficient solutions and infrastructure development.

Rumors circulated about DeepSeek’s R2 model launching in August after the model itself hinted at a release window. However, sources close to DeepSeek denied these claims. While internal testing is underway, CEO Liang Wenfeng is reportedly unconvinced by the model’s performance. Furthermore, Nvidia chip restrictions may have impacted DeepSeek’s computing power, potentially leading to resource strain and a suboptimal user experience upon release.