Google Cloud Doubles to $30.71B as TPU Sales Move the Revenue Line

Google Cloud revenue hit $30.71 billion in the third quarter, up 103%, and Alphabet now books TPU system sales directly as cloud product revenue — a chip business now sits inside the cloud line. That's the headline number in a quarter where both Alphabet (NASDAQ:GOOGL) and Amazon (NASDAQ:AMZN) posted their fastest cloud growth in years, built on two very different custom-silicon strategies.

Alphabet runs TPUs behind its own Gemini models and Search. Amazon uses Trainium and Graviton to cut the cost of running other companies' models inside AWS. Same hardware thesis, opposite monetization paths. For practitioners, that split decides where the next dollar of AI compute spend lands.

What do the numbers show?

Alphabet's cloud operating income rose to $11.12 billion from $3.59 billion a year ago — the unit makes real money at scale, not just growth. Search query growth hit its highest rate in five years, and the Gemini app crossed one billion users. Each layer of the stack feeds demand to the others.

Sundar Pichai said Google is "seeing strong demand for our TPU systems across a wide range of customers." That quote matters because it confirms TPU sales aren't captive to Alphabet's own workloads anymore.

AWS grew revenue 39% to $45.98 billion, its fastest pace in nearly five years. Its operating income of $17.77 billion still tops Alphabet's, even at half the growth rate. Andy Jassy said "most of Bedrock today is being run on top of Trainium."

The customer economics are concrete: buyers running on Amazon's chips saved more than $2 billion in Q3 alone. That's roughly a lock-in mechanism priced into the hardware.

Business Driver	Alphabet	Amazon
Cloud growth	103%	39%
Cloud operating income	$11.12B	$17.77B
Chip showcase	TPU systems sold to outside customers	Trainium running most of Bedrock
Model strategy	First-party Gemini	Multi-model marketplace

Who owns frontier training, and who undercuts inference?

Google spent a decade building TPUs into a platform for frontier training and native JAX scaling. Amazon designed its chips for a different job: breaking Nvidia's (NASDAQ:NVDA) pricing power on inference.

Jassy's view explains that choice: "There is not one tool to rule the world." He expects six to eight models to reach roughly the same quality. If he's right, AWS makes money whichever lab wins. If Alphabet is right, owning the best model on the cheapest chip takes the market.

Both strategies are expensive:

• Alphabet's capex hit $51.36 billion, and free cash flow fell to negative $9.07 billion.
• Alphabet also sold new shares through its at-the-market (ATM) program, which dilutes existing holders.
• Amazon's capex reached $77.26 billion, with roughly $220 billion expected for 2026.

That Amazon figure for 2026 is nearly triple its 2025 run rate — worth sitting with for a second before assuming margin expansion.

Could Trainium outside AWS reshape the race?

Amazon plans to sell Trainium outside AWS late in 2027. That puts it in direct competition with Google's TPU sales for the first time, turning two parallel strategies into a head-to-head merchant silicon market.

Watch AWS margins in the meantime. CFO Brian Olsavsky noted that AI margins run lower than core AWS margins, so today's 38.7% operating margin could narrow as Bedrock grows. On the Alphabet side, the marker to track is whether TPU customers keep spreading beyond its own products.

Which stock setup looks better?

Alphabet trades at about 13 times earnings versus Amazon's 37. The gap looks narrower than it reads, though: big equity gains earlier in the year inflated Alphabet's trailing earnings, so treat the multiple with caution. Alphabet shares fell 18.06% over the past month, while Amazon is up 15.64% year to date.

The original analysis finds Alphabet more compelling because it owns every layer, from hardware to a billion Gemini users. Amazon's pitch is AI exposure without betting on a single winning model. That view could shift if Alphabet's share dilution speeds up or TPU demand stalls.

For anyone allocating compute or capital, the practical read is simpler: Alphabet is selling a vertically integrated stack, and Amazon is selling the cheapest place to run everyone else's models. Q3 proved both models work. 2027 decides which one scales further.