Alphabet (GOOGL) shares rose approximately 3% on Monday following a report that Google is developing a new custom AI server chip designed specifically to optimize its Gemini large language models. The chip, internally codenamed 'Frozen v2,' would embed parts of Gemini's architecture directly into silicon, potentially reducing computation and data movement for AI queries while lowering power consumption.
According to The Information, the chip is intended to improve efficiency by a factor of six to ten times more tokens per unit of power compared to Google's latest Tensor Processing Units (TPUs). Unlike TPUs, which support a broad range of AI models, Frozen v2 is being built exclusively for Gemini, making it a specialized addition to Google's custom-chip portfolio rather than a replacement for general-purpose processors.
Alphabet responded to the report by emphasizing its ongoing investment in AI hardware innovation. The company stated that its teams are 'constantly researching and experimenting with new innovations to deliver maximum performance and efficiency for our users and customers' and that 'by co-designing our hardware and software from the ground up, we ensure our systems are integrated and highly optimized for real-world workloads.' The stock was trading 1.3% higher at the time of writing.
The project, with a target deployment around 2028, aims to address internal compute shortages that have reportedly limited Google's cloud capacity. Last month, Google agreed to pay SpaceX nearly $1 billion per month to help meet enterprise computing commitments, underscoring the urgency of expanding its AI infrastructure. However, the report notes that Frozen v2 would remain effective only if future Gemini models continue using the same underlying architecture, and Google views the project partly as a trial run, not expecting to manufacture the chips at the same scale as its TPUs.
Google's hardware ambitions come as its AI business faces intensifying competition. A recent Bloomberg report indicated that the next Gemini Pro release has been delayed, and Google has lost several senior researchers to rivals. Chinese AI developers have also gained traction, with their models now accounting for 45% of token usage among US companies. Releases from Moonshot AI and Alibaba have further narrowed the performance gap with leading US AI models.
Meanwhile, Google DeepMind Chief Executive Demis Hassabis is on Capitol Hill this week advocating for a federally overseen, industry-funded AI watchdog modeled after FINRA to evaluate advanced AI systems for national security risks before deployment.
Wall Street remains constructive on Alphabet. BMO Capital raised its price target on Alphabet to $455 from $435 while maintaining an Outperform rating, citing stronger cloud demand, expanding capacity, and a substantial backlog. However, BMO also noted questions remain regarding Gemini model performance following reports that Gemini Pro 3.5 has been delayed as it falls short on certain benchmarks.
Google continues to expand its in-house AI hardware efforts to reduce dependence on Nvidia while lowering the cost of running Gemini. Earlier this year, the company introduced its eighth-generation TPU and has increasingly marketed its chips to external cloud customers, including a multibillion-dollar agreement to supply TPUs to Meta Platforms. Google has also approached other cloud providers that have traditionally relied on Nvidia GPUs.
For more context on Alphabet's recent market moves, see Alphabet Stock Dips 2% on Gemini 3.5 Pro Delay, Yet Analysts Remain Bullish Ahead of Earnings and Morgan Stanley Picks Alphabet, Amazon, Meta as AI Trade Shifts from Chips. Additionally, the broader tech rally has been fueled by AI catalysts, as highlighted in Dow Breaks 52,000 as Alphabet Surge and Chip Rebound Fuel Rally and Alphabet Surges 4% on Dow Debut, AI Capacity Constraints Fuel Investor Interest.
This article is for informational purposes only and does not constitute financial advice.
