Nvidia shares edged up approximately 1.3% to $202.50 in premarket trading on Thursday, buoyed by Google's announcement of its new in-house TPU 8 series AI chips and a reaffirmed commitment to offer Nvidia's upcoming Vera Rubin NVL72 server platform. The move underscores the ongoing tension between cloud giants developing custom silicon and their continued reliance on Nvidia's dominant hardware.
Alphabet, Google's parent company, held firm on its 2026 capital expenditure plan of $175 billion to $185 billion, signaling that cloud AI budgets remain robust even as competitors introduce their own custom chips. Google Cloud CEO Thomas Kurian declared that “the experimental phase is behind us,” emphasizing the shift toward monetizing AI agents—software capable of autonomous planning and execution—rather than mere demonstrations.
Google's new tensor processing units, the TPU 8t for large-model training and the TPU 8i for inference, are designed to power AI agents with improved efficiency. According to Google, the TPU 8i delivers an 80% improvement in inference performance per dollar over its predecessor. Nvidia's Vera Rubin NVL72 platform remains a cornerstone for Google's infrastructure, highlighting that even with internal chip advancements, the company continues to lean on Nvidia's technology for large-scale AI deployment.
Bullish signals for Nvidia also emerged from memory maker SK Hynix, a key supplier of high-bandwidth memory essential for AI chips. The company reported blowout quarterly profits driven by AI demand and stated that its order book for the next three years “far exceeds our production capacity,” with customer requests continuing to flow in.
However, the competitive landscape is widening. Broadcom recently secured a long-term agreement to develop and supply Google's next-generation custom AI chips through 2031, while Anthropic runs its Claude models across a mix of platforms including AWS Trainium, Google TPUs, and Nvidia GPUs. Nvidia remains a central player but no longer operates in a monopoly.
China-related uncertainties persist. U.S. Commerce Secretary Howard Lutnick told a Senate panel that Nvidia's H200 chips have not yet reached Chinese buyers, despite receiving U.S. approval in January, because Beijing has not given its sign-off. “We have not sold them chips as of yet,” Lutnick said, leaving a key market in limbo.
Analysts note potential headwinds: custom chips could capture more inference market share, China may remain closed, and oil prices above $100 could reignite inflation fears, reviving concerns about whether AI investment is outpacing returns. Capital.com analyst Kyle Rodda flagged AI overinvestment and “diminishing future returns” as a “forgotten theme.”
Despite these risks, Nvidia shares have held above $200, and cash is beginning to flow back into U.S. equities. “The bigger risk may be staying on the sidelines too long,” said State Street's Michael Arone, as traders bet that AI investment and strong earnings will continue to support big tech names. The stock remains caught between growing competition and sustained customer demand, a tension that will likely define its next move.
