E-Paper
Alphabet stock jumps as Google reportedly builds AI chip to boost Gemini efficiency
Alphabet shares rose after a report said Google is building the Frozen v2 AI chip to make Gemini faster, more efficient, and reduce computing power needs.
Alphabet shares rose about 3% on Monday after a report said Google is developing a new AI server chip called "Frozen v2" to make its Gemini AI models run more efficiently. The new chip is being built specially for Gemini AI. It is designed to improve speed and reduce the amount of computing power needed to answer user questions.
Frozen v2 would permanently build parts of Gemini's AI architecture directly into the chip. This means the chip can perform many tasks without repeatedly running the same calculations. Because of this design, the chip would reduce both the number of calculations and the movement of data inside the system. That would make Gemini faster and more energy efficient.
Google explains AI chip plans
Alphabet did not directly confirm the project, but the company told CNBC that its engineers are constantly researching and testing new technologies to improve performance and efficiency. Alphabet said not every research project becomes a real product, but experimenting with new ideas is an important part of its strategy. The company also said it designs both its hardware and software together so that its AI systems are highly optimized for real-world workloads.
Google engineers believe Frozen v2 could generate six to ten times more AI tokens using the same amount of electricity compared with Google's latest Tensor Processing Units (TPUs). AI tokens are the small pieces of text an AI model reads and generates. Producing more tokens with the same power means better efficiency and lower operating costs. Frozen v2 is not expected to replace Google's TPUs. Instead, it would become a more specialized chip within Google's custom AI hardware lineup.
Also read: Microsoft backs AMD Helios AI system to challenge Nvidia's dominance
Google targets 2028 launch
Google is reportedly aiming to deploy the new chip in 2028. One major goal of the project is to solve Google's internal shortage of AI computing power. The company has been struggling to meet growing demand for AI infrastructure. The computer shortage has reportedly created internal pressure and even forced Google Cloud to reject some outside customer business. Last month, Google reportedly agreed to pay SpaceX nearly $1 billion every month to help increase computing capacity and meet enterprise AI commitments, according to CNBC.
There is one important downside to Frozen v2. The chip would work well only if future Gemini models continue using the same core architecture. If Google makes major changes to Gemini's design in the future, Frozen v2 may become less useful. That is because the chip is built for a specific AI architecture, according to The Information. Google reportedly sees Frozen v2 as a test project for now. The company does not plan to manufacture it in the same large volumes as its general-purpose TPUs.
Google faces AI competition
Google's AI business is also facing other challenges. The next version of Gemini Pro has reportedly been delayed. Google has also lost several senior AI researchers to competing companies, increasing pressure in the fast-moving AI race. Chinese AI models are becoming stronger competitors. According to the report, they now account for about 45% of AI token usage by US businesses.
Competition increased further over the weekend after new AI model releases from Moonshot AI and Alibaba, which reportedly narrowed the performance gap with leading US AI models. At the same time, Google DeepMind CEO Demis Hassabis is in Washington, D.C. meeting US lawmakers to discuss AI regulation. Hassabis is reportedly proposing a FINRA-style AI watchdog that would be overseen by the federal government, funded largely by the AI industry, and responsible for testing advanced AI models for national security risks before release, according to CNBC.
Alphabet invests in AI chips
Alphabet has been investing heavily in custom AI chips as part of its long-term strategy to reduce dependence on third-party hardware suppliers and build more of its own AI infrastructure. A more efficient AI chip could help Google lower computing costs, improve Gemini's performance, and strengthen its position against rivals like OpenAI, Anthropic, and leading Chinese AI companies.
Investors welcomed the report, sending Alphabet shares higher, as they believe stronger in-house AI hardware could support the company's long-term AI growth and competitiveness.
