Google is accelerating its artificial intelligence (AI) hardware strategy as competition intensifies in the global AI chip market. The company is introducing specialised versions of its custom-built Tensor Processing Units (TPUs) to meet the growing demand for computing power and strengthen its position against market leader Nvidia.
The move marks a change in Google's traditional hardware development cycle. Instead of following its usual two-year launch schedule, the company is speeding up the introduction of new processors. Its strategy focuses on developing chips for specific workloads rather than relying on a single design for every AI application.
Google is expanding its TPU portfolio with two specialised processors: the TPU 8tx and the TPU 8i Max. The chips are designed to address different requirements, including secure computing and demanding AI workloads.
Google's head of chip development, Amin Vahdat, has highlighted specialisation as an important direction for the future of AI infrastructure. The company's latest processors reflect this approach by targeting distinct computing needs.
The TPU 8tx is designed for organisations that handle sensitive information and require strong data protection. Its capabilities are aimed particularly at government agencies and businesses operating in highly regulated industries.
The chip supports confidential computing, an approach that helps protect data and AI models while they are being processed. This can be important for organisations that need to maintain strict control over information and prevent unauthorised access.
The TPU 8i Max is built to handle computationally intensive AI tasks. Google has combined two TPU chips into a single unit, providing greater memory and computing capacity.
This configuration is intended to support applications that need to process large volumes of information. Potential use cases include advanced mathematical calculations, software development, complex reasoning and drug discovery.
Google has partnered with Advanced Micro Devices (AMD) to support the secure computing architecture associated with the TPU 8tx.
The system combines Google's custom AI hardware with an AMD processor. The partnership is intended to help protect sensitive data and AI models through hardware-level encryption during processing.
The collaboration also adds an interesting dimension to competition in the semiconductor industry. AMD is led by CEO Lisa Su, who is a cousin of Nvidia CEO Jensen Huang. While the family connection has attracted attention, the business significance of the partnership lies in Google's effort to strengthen its hardware ecosystem and reduce dependence on Nvidia-based infrastructure.
Pricing is emerging as an important factor in Google's attempt to gain a larger share of the AI chip market. According to market analysts cited in the report, Google's TPUs can cost around 30% less than Nvidia's popular graphics processing units (GPUs).
The processors are also positioned as competitive in data-processing performance and power efficiency. If these cost advantages hold across different workloads, they could help businesses manage the growing expenses associated with AI infrastructure.
Several major technology companies are reportedly interested in Google's hardware, including:
Apple
Meta
OpenAI
Oracle
Anthropic
Anthropic, the company behind the Claude AI chatbot, reportedly has an option to purchase up to $150 billion worth of Google's chips between 2028 and 2031.
Financial firms are also exploring alternative hardware. Jump Trading, for example, is adopting TPUs for AI model training. Using different types of processors can help organisations select hardware based on the specific requirements of individual workloads.
Google's growing focus on selling AI chips could develop into a major source of revenue. Bloomberg Intelligence estimates that the company's TPU sales could increase from approximately $12 billion in 2026 to more than $135 billion by 2028.
These figures are projections, not confirmed future revenue.
However, expanding the business will depend on Google's ability to increase production, secure components and meet customer demand. The company is already facing supply constraints affecting memory components and semiconductor manufacturing capacity.
Despite strong demand, Google faces several challenges as it expands its AI hardware operations.
New companies are entering the specialised chip market. Tenspresso, founded by former Google engineers, is among the startups working on competing chip technologies. Such developments could increase competition and give AI companies more hardware options.
The semiconductor industry is dealing with limited supplies of important memory components and manufacturing capacity. Reports indicate that Google's available TPU production is already committed to buyers, highlighting the gap between demand and supply.
To reduce the risk of future bottlenecks, Google is exploring partnerships with hardware companies such as Marvell Technology and MediaTek. Diversifying its supply chain could help the company improve production flexibility and support future expansion.
Analysts have also warned that global data centre infrastructure spending could slow by 5% to 7% by 2027. Any slowdown in investment could influence demand for AI chips and related equipment.
Google's latest TPU developments reflect a broader shift towards specialised AI infrastructure. As businesses adopt generative AI, AI agents and advanced computing applications, demand is increasing for processors that can deliver strong performance without excessive energy and infrastructure costs.
By offering different chips for secure computing and intensive workloads, Google is attempting to serve a wider range of customers. Its partnership with AMD, interest from major technology companies and plans to diversify manufacturing support this strategy.
However, competing with Nvidia will require more than competitive pricing. Consistent chip supply, software compatibility, reliability and performance across real-world applications will also influence customer decisions. Google's ability to address these challenges will help determine how much influence its TPU business can gain in the global AI market.