Google Cloud revenue grew 82% to $24.77 billion in the summer quarter, powered in part by Alphabet's own TPU silicon, while Nvidia's revenue rose 105.8% on hyperscaler demand — a pairing that frames the central question in AI infrastructure: whether Alphabet's in-house chip program can dent Nvidia's trajectory before it runs out of room to grow.

Alphabet and Nvidia reported the same AI boom from opposite ends of the supply chain. Alphabet designs its own Tensor Processing Units, operates its own cloud, and controls its own models, yet it still buys Nvidia systems every quarter while shifting steadily toward cheaper, application-specific silicon. Nvidia's Data Center segment reached $89.02 billion for the quarter, with networking revenue up 138%, as customers buy complete systems rather than discrete GPUs. "Now, compute is revenue," Jensen Huang said.

The capital intensity gap between the two companies is stark. Alphabet spent $44.92 billion in capex during the quarter and posted negative free cash flow of $5.86 billion, prompting the company to suspend buybacks. Nvidia spent $2.68 billion on capex and generated $21.34 billion in free cash flow. Alphabet is effectively financing the challenger position; Nvidia collects the cash.

Sundar Pichai defended the spending on the earnings call, saying Alphabet's "differentiated, full stack approach to AI is delivering real, measurable value." Gemini now processes 22 billion API tokens per minute, and Google can route that traffic onto its own TPUs rather than rented GPUs — a flexibility no other hyperscaler matching Nvidia's purchase volumes possesses.

How much Google actually matters to Nvidia

Nvidia does not disclose Google's share of its sales. Google sits inside a hyperscale customer category worth $49 billion per quarter, which grew 13% sequentially, according to the company's reporting. Revenue from neoclouds, enterprises and government buyers reached $40 billion, and management expects non-hyperscaler customers to make up roughly half of the data center business going forward.

Huang has argued that competing XPUs "are inference-specific chips for one cloud or one service." That framing weakens when applied to Google, which owns the models, the cloud and the workloads that those chips serve. A prior 24/7 Wall St. analysis projected TPU shipment volumes could triple to 8.8 million units by 2027, though that figure remains a forecast rather than a confirmed order book.

Margins will expose any TPU damage first

The clearest early indicator will be Nvidia's gross margin. Management guides it to bottom at 71% to 72% in the fourth quarter. If TPU expansion starts pulling GPU pricing down, the margin line will register it before revenue does. Alphabet, for its part, held a Google Cloud backlog above $460 billion after Q1, and the recent launch of Gemini 4 Argon gives TPU capacity another major internal workload to absorb.

The competitive dynamics cut both ways in the near term. Nvidia's outlook for roughly 70% growth in fiscal 2028 is, in management's words, "supply-constrained." Under tight supply, every TPU Google deploys for itself frees Nvidia GPU inventory for other buyers — meaning Alphabet's silicon strategy caps how far Nvidia can grow later more than it threatens current sales.

Valuations reflect the asymmetry. Nvidia trades at 46 times earnings on far stronger cash generation; Alphabet trades at 15 times earnings while carrying the heavy spending burden of the challenger role.

If TPU volumes do triple toward 2027 and Nvidia's gross margin breaks below the guided 71% floor, the pricing power question answers itself; until then, the divergence shows up only in capex lines and backlog figures.