Google Has Two New Weapons in Its Fight Against Nvidia

It seems like Google has no problem kicking a horse when it’s down. The tech giant announced Monday two new versions of its existing AI chip systems, designed to offer stronger security and handle more demanding AI workloads. The news comes just days after Nvidia CEO Jensen Huang, whose company dominates the AI chip market, admitted to having an affair with a woman suspected of being a Chinese spy. Nvidia’s board of directors has since launched an independent investigation into the matter. The company said it has no reason to believe national security or company secrets were compromised. Nvidia stock was down roughly 5.6% at the time of writing on Monday. But that doesn’t mean Google has to go easy on Nvidia. Google first unveiled its eighth-generation Tensor Processing Units (TPUs), custom-made AI processors, back in April. The company has been developing its own TPUs for more than a decade, initially for tasks like voice recognition and translation. Today, they’re used to train and run Google’s Gemini AI models, power its Google Cloud business, and as of recently being sold to customers for use in their own data centers. Google Chief Technologist and Senior Vice President of AI and Infrastructure Amin Vahdat wrote in a blog post Monday that the two newest systems were designed in part based on feedback from outside customers about their specific needs. “With computing moving from human to agent scale, businesses are asking for accelerators that can reason across AI models’ massive context windows,” wrote Vahdat. “In addition, some highly regulated customers are asking for specialized encryption capabilities for unique workloads.” The first system, called TPU 8tx, is designed for governments and highly regulated industries that need to keep sensitive data secure while running AI models. Google says the system uses hardware-level security to protect data even while it’s being processed. Meanwhile, the TPU 8i Max is designed for AI tasks that require large amounts of memory, such as dozens of AI agents working together on an entire software codebase. The system combines two existing TPU 8i chips into a single package, doubling computing power and memory bandwidth. Both systems are expected to become available in 2027, according to Bloomberg. The announcement also comes at a difficult time for the broader AI industry. On top of an ongoing memory shortage, the industry is still reeling from last week’s midterm elections, which saw Democrats win control of Congress and voters overwhelmingly back measures restricting data center development across the country. AI-related stocks took a hit after the elections as investors worried that tougher regulations could slow down data center construction. Still, Google’s TPU business appears to be doing pretty well. In its most recent quarterly earnings report, Alphabet said TPU systems sales made a meaningful contribution to Google Cloud’s revenue growth. Revenue from the division more than doubled year over year to $30.7 billion in the quarter ended Sept. 30. Google also said Monday that it now has nearly 100 TPU customers. That includes major companies like Apple, Meta, and Oracle, according to Bloomberg. In other news, AI chips aren’t the only area where Google’s parent company, Alphabet, is looking to expand. Alphabet’s drone delivery subsidiary, Wing, announced Friday that it plans to invest up to $125 million in manufacturing operations at its new facility in Fort Worth, Texas, over the next two years. The factory will be capable of producing tens of thousands of delivery drones each year and could create up to 100 jobs. Wing says it has already completed more than 1 million deliveries in the U.S. this year. The company is now looking to expand its delivery service through its partnership with Walmart, with plans to reach nearly 20 U.S. markets by 2027.