Piper Sandler Sees Google TPU Sales Reaching $104 Billion by 2028

Google’s artificial-intelligence chip business could become a much larger source of revenue than investors currently expect, according to a new forecast from Piper Sandler.

Analyst Thomas Champion estimates that Google’s external Tensor Processing Unit (TPU) business could generate about $104 billion in hardware revenue by 2028. The forecast highlights the growing commercial opportunity around Google’s in-house AI chips as the company expands beyond using TPUs primarily for its own cloud and AI workloads.

The estimate also adds another dimension to the competition in AI infrastructure. Nvidia remains the dominant supplier of high-end AI accelerators, but Google, Amazon and Microsoft have increasingly developed their own chips to reduce dependence on third-party processors and tailor computing hardware to their workloads.

For Alphabet investors, the potential significance is much larger than the chip business itself. Successful TPU commercialization could strengthen Google Cloud, create another AI revenue stream and give Alphabet greater control over the economics of AI computing.

What Are Google TPUs?

TPU stands for Tensor Processing Unit. These are custom-designed chips developed by Google specifically for machine-learning and artificial-intelligence workloads.

Unlike general-purpose processors, TPUs are optimized for the mathematical operations used extensively in AI models.

Google has used TPUs internally for years and has also made them available to customers through Google Cloud. More recently, the company has been expanding the ways outside customers can access or purchase TPU capacity.

That creates a potentially significant business opportunity.

Instead of relying entirely on Nvidia GPUs to supply AI computing capacity, Google can use its own chip architecture across its cloud infrastructure and potentially sell TPU systems to external customers.

Piper Sandler's $104 Billion TPU Forecast

Piper Sandler's Thomas Champion estimates that Google's external TPU business could reach approximately $104 billion in hardware revenue by 2028. The firm argues that the opportunity has not yet been fully reflected in broader market expectations.

The forecast is an analyst estimate, not Google's official revenue guidance. Investors should therefore treat the $104 billion figure as a scenario based on assumptions about TPU demand, production capacity, pricing and commercialization.

Other Wall Street research has also produced substantial estimates for Google's TPU opportunity, although the figures vary considerably.

For example, Morgan Stanley recently estimated TPU-related Google Cloud revenue of $84 billion in 2027 and $108 billion in 2028, while Oppenheimer has estimated about $170 billion of cumulative incremental revenue through 2028 from external TPU sales.

The differences between these forecasts underline an important point: there is considerable uncertainty around the eventual size of Google's external TPU business.

Why Google's TPU Business Is Expanding

The AI boom has created an enormous need for computing power.

Training and running advanced AI models require large numbers of accelerators operating together inside data centers. Nvidia has captured a substantial portion of this market through its GPUs and software ecosystem.

Google has a different advantage: it designs both the AI chips and much of the software and cloud infrastructure surrounding them.

That vertical integration can allow Google to optimize its computing systems for specific workloads.

For customers, the attraction can include performance, availability, energy efficiency and potentially different economics from using conventional GPU infrastructure.

Google's latest TPU generation, Ironwood, is designed for AI inference as well as training. The company has been increasingly positioning its custom silicon as an important component of its broader AI infrastructure strategy.

Google Is Turning Internal Technology Into a Commercial Business

The most important change is that TPUs are increasingly becoming more than an internal Google advantage.

Google Cloud customers can access TPU computing capacity, while the company has also been pursuing arrangements that allow external customers to obtain TPU systems.

Recent reports have highlighted agreements involving large AI-computing deployments and infrastructure providers.

This changes the economics of Google's custom-chip program.

If Google can monetize TPUs through external customers at significant scale, the company does not have to rely exclusively on advertising or conventional cloud services to capture value from its AI hardware investment.

Instead, the semiconductor layer itself can become part of the revenue opportunity.

What This Means for Google Cloud

The TPU opportunity could have an important effect on Google Cloud, which has already been growing rapidly.

Alphabet reported that Google Cloud revenue increased 32% year over year to $13.6 billion in the first quarter of 2026, while operating income reached $2.8 billion. The company has also been investing heavily in AI infrastructure to meet rising demand.

More recent analyst estimates have become considerably more aggressive about the potential contribution from TPU commercialization.

Morgan Stanley, for example, expects TPU-related Google Cloud revenue to reach $84 billion in 2027 and $108 billion in 2028 under its revised assumptions. The firm also estimated that Google Cloud could account for approximately 48% of Alphabet's company-wide EBIT by 2028.

These are analyst projections rather than company guidance, but they demonstrate why investors are paying increasing attention to Google's custom AI silicon.

Could TPUs Challenge Nvidia?

Google's TPU expansion does not necessarily mean Nvidia's position is immediately threatened.

Nvidia has built a broad ecosystem around its GPUs, including networking, software libraries, developer tools and systems designed for large-scale AI computing.

That ecosystem is one of Nvidia's major competitive advantages.

Google, however, has a different model. It controls its own cloud infrastructure and can optimize TPUs for Google's AI models and workloads. It also has the ability to offer TPU access directly through Google Cloud.

Other hyperscalers are pursuing similar strategies.

Amazon has its Trainium and Inferentia chips, while Microsoft has developed its own AI accelerators. The result is a gradual diversification of the AI-chip market.

For investors, the important question is therefore not simply whether Google will replace Nvidia. It is whether custom accelerators will capture an increasing share of the enormous AI-computing market.

The Financial Opportunity Comes With Costs

A $104 billion revenue forecast sounds substantial, but investors should also consider the capital required to generate that revenue.

AI chips require advanced semiconductor manufacturing, packaging, networking equipment and data-center infrastructure.

Google is already spending heavily on AI infrastructure, and higher TPU demand could require additional investment.

There is also a margin question.

Hardware revenue does not automatically translate into equivalent levels of profit. The economics depend on manufacturing costs, system configuration, depreciation, data-center expenses and the pricing Google can command.

Morgan Stanley, for example, recently raised its assumed TPU gross margin to 30% from 20% in its model, illustrating how sensitive the business case can be to margin assumptions.

Why This Matters for Alphabet Stock

For Alphabet investors, TPU commercialization could provide another way for the company to monetize its AI capabilities.

Google already has several major AI assets:

• Gemini and other AI models

• Google Cloud

• Search and advertising

• YouTube

• Custom TPU hardware

• AI software and infrastructure

The TPU business connects several of these pieces.

Google can use its chips internally, rent computing capacity through Google Cloud and potentially sell systems or capacity to external customers.

If demand grows as analysts expect, Alphabet could capture more value from the AI infrastructure chain rather than paying external suppliers for every accelerator it needs.

However, the size of the opportunity remains uncertain.

Risks Investors Should Watch

The biggest risk is that analyst estimates could prove too aggressive.

TPU adoption depends on customers being willing to use Google's hardware instead of Nvidia GPUs or competing accelerators.

Manufacturing capacity is another constraint. Scaling custom AI chips requires access to advanced semiconductor manufacturing and packaging.

Customer concentration could also matter if a relatively small number of large AI companies account for much of external TPU demand.

There is also competition from Nvidia and other custom-chip programs. If competing accelerators improve rapidly, Google's pricing power could be weaker than expected.

Finally, investors should remember that the $104 billion figure is an analyst forecast rather than confirmed future revenue.

What Investors Should Watch Next

The most important signals will be Google's disclosures about TPU customers, production capacity and revenue.

Investors should also monitor:

• Google Cloud revenue growth

• AI-related capital expenditure

• TPU deployment and customer adoption

• TPU pricing and margins

• Demand from major AI companies

• Nvidia's market share and competitive response

• Alphabet's overall AI infrastructure spending

Another important indicator will be whether Google's TPU business begins appearing as a meaningful contributor to reported revenue rather than remaining primarily an internal infrastructure advantage.

Conclusion

Piper Sandler estimates that Google's external TPU hardware business could generate about $104 billion in sales by 2028, highlighting the rapidly expanding commercial opportunity around Alphabet's custom AI chips.

The forecast is significant because Google is attempting to capture more of the AI infrastructure value chain. Its TPUs can support Google's own AI workloads, power Google Cloud services and potentially become a major external hardware business.

The $104 billion figure remains a forecast, not guaranteed revenue. Investors should therefore focus on actual customer adoption, production capacity, margins and Google Cloud growth as the TPU business scales.

If Google's custom silicon gains meaningful traction, the impact could extend beyond semiconductors by strengthening Google Cloud and changing the competitive dynamics of the AI infrastructure market.

Follow the blog for more updates on Alphabet, Google Cloud, AI chips, Nvidia, technology stocks and global markets.

This article is for informational and educational purposes only and should not be considered investment advice.