Alphabet sells custom AI chips to boost cloud growth while facing cash drain

A new hardware business model

Alphabet is diversifying its business by selling its homegrown AI chips to external customers. These custom processors, known as Tensor Processing Units or TPUs, were originally designed for internal use to power Google's own AI projects. Now, the company is positioning itself as a direct competitor to Nvidia in the hardware space.

As reported by The Drive, this move is helping to drive significant growth in Google Cloud. The company did not provide specific revenue figures for the chips, but noted that sales made a "meaningful" contribution to the cloud division. This suggests that Google is successfully turning its internal tools into a viable product for other tech companies.

The shift represents a change in how Alphabet generates money. Instead of just selling software services or advertising, they are now involved in the physical manufacturing and distribution of high-end hardware. This transition is happening as investors watch closely to see if AI can offset potential threats to Google's traditional search advertising revenue.

The cost of building inventory

While selling chips brings in revenue, it also creates a significant drain on Alphabet's cash flow. Unlike traditional software sales, hardware requires the company to spend billions on manufacturing and inventory before a customer ever pays a cent. This creates a "different cycle of cash flow generation" compared to their usual business model.

Finance chief Anat Ashkenazi explained that inventory balances will now include TPU systems that have not yet been sold. This means Alphabet must front the costs of building these systems to meet surging demand. This upfront spending puts pressure on the company's liquid cash, even as the long-term sales potential grows.

Investors have reacted nervously to this trend, with shares falling nearly 11% on Thursday. The stock drop followed a report of another quarter of cash burn and a reiteration that spending on AI infrastructure will rise sharply next year. The company is essentially spending heavily today to build the infrastructure that will power the AI economy of tomorrow.

Who is buying Google's chips

Several major AI players are already moving away from relying solely on Nvidia and toward Google's hardware. Alphabet reported last year that it would supply Anthropic with TPUs worth tens of billions of dollars. This move helps Anthropic scale its own models while giving Google a foothold in the broader AI ecosystem.

Beyond Anthropic, media reports suggest that Alphabet is currently in talks with OpenAI for a similar deal. By providing these chips, Google is attempting to become the foundational infrastructure provider for the world's leading AI developers. This strategy aims to make Google Cloud the primary destination for companies building large-scale models.

Thomas Kurian, the head of Google Cloud, expects sales of these TPU systems to rise further in the current quarter. He predicts that these sales will ramp up significantly by 2027. This long-term outlook suggests that Google is playing a marathon game to secure its place in the hardware supply chain.

The competitive landscape

The primary challenge for Alphabet is competing with Nvidia, the current leader in the GPU market. Nvidia's graphic processing units are the industry standard for training and running AI models. Alphabet must prove that its TPUs offer enough performance or cost-efficiency to lure customers away from the Nvidia ecosystem.

Analysts are watching the profit margins of these sales very closely. Jeremy Saltzberg of Check Capital noted that there is no reason to believe Alphabet's margins would be significantly different than what an Nvidia GPU would sell for. Because Nvidia's margins are currently very high, Alphabet faces a difficult task in maintaining profitability while scaling production.

The trade-off for Alphabet is clear: they are sacrificing immediate cash flow to build a massive hardware footprint. If they can successfully capture a large portion of the market, they could become a dominant force in AI infrastructure. If they cannot, the high costs of building and maintaining these data centers could become a long-term burden on the company's finances.

What comes next for Alphabet

The next year will be a critical period for Alphabet as it continues to scale its AI infrastructure. The company has already signaled that spending will increase sharply, which may continue to put pressure on its quarterly cash reports. Investors will be looking for signs that this spending is translating into sustainable, high-margin revenue.

As more companies like OpenAI and Anthropic integrate TPUs into their workflows, Google will need to maintain a steady supply of hardware. Any bottlenecks in manufacturing or software integration could hurt their ability to compete with Nvidia. The company must balance the need to sell hardware with the need to keep its own cloud services running smoothly.

Ultimately, the success of this strategy depends on whether Google can turn its hardware sales into a primary growth engine. If the TPU systems become the backbone of the AI industry, Alphabet could redefine its role in the tech economy. For now, the company remains in a high-stakes transition from a software giant to a hardware powerhouse.