Amazon and Google both have custom chips. Here's what they said about their future.

Amid their respective surges in artificial intelligence-related spending, Amazon (AMZN) and Alphabet (GOOG) (GOOGL) have seen strong internal demand for their respective custom chip businesses.

Both companies still rely heavily on Nvidia (NVDA) for access to GPUs and will likely continue to do so for the foreseeable future. However, it's clear the ambitions of both companies go beyond just internal use to generate revenue. Alphabet started selling its tensor processing units externally earlier this year, and Amazon is set to do so with its Trainium offerings in late 2027.

And given the insatiable demand for GPUs and CPUs to run AI models, agents, and more, both tech titans see promising futures that could generate billions of dollars in added revenue.

'Strong demand' for TPUs, with more to come?

For Alphabet, its TPU systems have seen “strong demand” already, CEO Sundar Pichai said. The company started offering them externally starting in the middle of this year.

“We are working hard to bring more capacity online to meet this long-term demand,” Pichai said on the company's third-quarter earnings call. “At the same time, we are seeing strong demand for our TPU systems across a wide range of customers, and we expect that to continue as we scale.”

The company has worked with tech giants such as Meta (META) and Apple (AAPL), AI model labs like Mirendil and Physical Intelligence, and financial firms like Citadel Securities to utilize its TPUs. In total, nearly 100 customers already use TPUs, and the company expects that trend to continue.

“Given continued strong and accelerating demand, we've recently entered into partnerships to deliver TPU systems with both Crux AI and Oracle, with more in the works,” Google Cloud CEO Thomas Kurian said on the earnings call.

Yorkville Ives analyst Dan Ives said he was pleased regarding the demand for TPU systems that Alphabet is already seeing, citing it as a “second growth engine on the field the other hyperscalers do not have.”

In recent memory, Alphabet has expanded the distribution for TPUs, including deals with Oracle (ORCL) and Crux AI, Kurian explained. And while the company declined to provide specific financial information for TPU sales, CFO Anat Ashkenazi said there will be a “meaningful revenue ramp in 2027.”

Amazon's chips are growing at triple digits

Amazon said in July that its custom chips business (which includes Graviton and Trainium) had reached a $25B revenue run rate. This week, the company delved further into detail and said chip-related revenue grew at a triple-digit percentage pace year-over-year.

Revenue commitments from its Graviton CPUs and Trainium AI accelerators have each risen more than 20 times. By bringing Trainium outside of Amazon Web Services later next year, that could expand its market considerably, Amazon added.

Amazon CEO Andy Jassy explained “most customers are going to use multiple models,” and said its approach is different from Alphabet when it comes to selling chips externally versus what is used in the cloud. (There is still strong demand for older generations of Amazon's Trainium chips that support the economic longevity of the existing infrastructure.)

“I would say that the number of companies who have a significant cloud business and a chip strong enough to sell externally is effectively two,” Jassy said on the company's earnings call, referring to Alphabet. “So there really aren't many customers that have to grapple with the same type of problem at this type of scale. And at this point, everybody is producing their advanced chips with the same provider. There's only so many chips that provider can provide each year. Everybody gets an allocation. And you have to decide with that allocation how much of it you want to sell externally and how much of it you want to use in your cloud. And there's a trade-off. If you sell externally, you're able to realize more revenue in the short term because you get to recognize the revenue as soon as it leaves your inventory. And on the flip side, if you decide to use it inside your cloud in the form of EC2 instances or Bedrock, you actually will end up driving more absolute revenue and free cash flow from those chips because you get to use them over 5 to 6 years and drive revenue for a longer period of time. You have to really think through what you want to do. To date, I would say that we've taken a different approach in how we've used our chips than the other provider who has the same type of balancing act going on.”

“Our ability to create price performance, scalable chips is going to be a very significant strategic advantage for us at scale, both in our own cost structure as well as building a separate business. And selling external chips really is a very different business than your cloud infrastructure business,” Jassy added.