Alphabet-owned Google is aggressively stepping up its game to challenge Nvidia's dominance in the booming artificial intelligence chip market. Breaking away from its traditional two-year launch cycle, Google is accelerating its hardware rollout to meet the massive demand for AI computing.The company is introducing new variations of its custom-made processors, known as Tensor Processing Units (TPUs), marking a major shift towards creating specialised chips for specific tasks rather than a one-size-fits-all solution.The new powerhouses: TPU 8tx and 8i MaxGoogle’s head of chip development, Amin Vahdat, noted that the future of the AI industry lies in specialisation.

To cater to this, Google has unveiled two new chip variations. The first is the TPU 8tx, which is heavily focused on top-tier security and confidential computing.

It is specifically designed for government agencies and businesses in highly regulated sectors that need to keep their data completely isolated.The second chip, the TPU 8i Max, is built for heavy-duty AI workloads. Google has essentially packed two TPU chips into a single unit to deliver double the memory and computing power.

This makes it perfect for processing massive amounts of information at once, a feature essential for complex tasks like advanced mathematics, coding, and drug discovery. The AMD connection and a family rivalryTo power the highly secure TPU 8tx, Google has partnered with Advanced Micro Devices (AMD).

The system pairs Google's technology with an AMD processor to ensure that sensitive data and AI models remain encrypted at the hardware level while being processed.Interestingly, this collaboration adds a layer of family drama to the corporate AI race. AMD is led by CEO Lisa Su, who is actually the cousin of Nvidia CEO Jensen Huang.

By teaming up with AMD, Google is effectively joining forces with Huang's own family rival to chip away at Nvidia’s massive market share.Big clients and a major price advantage A major reason tech companies are turning to Google’s hardware is the cost. Market analysts point out that Google's TPUs cost roughly 30 per cent less than Nvidia’s highly sought-after graphics processing units (GPUs).

Even with the lower price tag, TPUs offer highly competitive data processing speeds and are much more power-efficient.This cost-to-performance ratio has already attracted a roster of tech heavyweights. Companies like Apple, Meta, OpenAI, and Oracle are already in line to use Google’s hardware.

Anthropic, the creator of the Claude AI chatbot, reportedly has an option to purchase up to $150 billion worth of these chips between 2028 and 2031.Even financial firms like Jump Trading are adopting TPUs to train their AI models, noting the benefits of using diverse hardware for different problems.The multi-billion-dollar road aheadGoogle’s pivot to selling chips is turning into a massive revenue stream. According to Bloomberg Intelligence, the company's TPU sales are projected to skyrocket from roughly $12 billion this year to over $135 billion by 2028.Despite the incredible growth, Google still faces significant hurdles.

Startups like Tenspresso, founded by former Google engineers, are stepping in to build competing chips. Furthermore, analysts warn of a potential 5 to 7 per cent slowdown in global data centre infrastructure spending by 2027.Google is also dealing with severe industry-wide shortages for key memory components and manufacturing space.

Every TPU the company makes is currently contracted to a buyer, pushing Google to diversify its supply chain by looking toward hardware partners like Marvell Technology and MediaTek to prevent future bottlenecks.- Ends