NVIDIA's 71% to 72% Q4 margin is where Alphabet's TPU threat will show

Alphabet and NVIDIA reported summer quarters showing the same AI boom from opposite sides. Google Cloud brought in $24.77B, driven by enterprise AI infrastructure, and Sundar Pichai said Alphabet's "differentiated, full stack approach to AI is delivering real, measurable value." NVIDIA's data center segment reached $89.02B, and networking grew 138%, with Jensen Huang saying: "Now, compute is revenue." The two are linked — Alphabet still buys NVIDIA systems while moving toward cheaper, more application-specific chips.

The financial pictures diverge sharply. Alphabet's free cash flow came in at negative $5.86B, and it suspended buybacks, while NVIDIA generated $21.34B in Q2 free cash flow on $2.68B in capex, against Alphabet's $44.92B. NVIDIA's revenue from neoclouds, enterprises and government buyers reached $40 billion, and management expects non-hyperscalers to make up roughly half of the data center business. Huang argued that many XPUs "are inference-specific chips for one cloud or one service" — an argument that works less well against Google, which has the models, cloud and workloads. Google Cloud's backlog was over $460 billion after Q1, and Google just launched Gemini 4 Argon, giving TPU capacity another major workload.

The stakes appear to hinge on timing and margins. NVIDIA's outlook for about 70% fiscal 2028 growth is "supply-constrained," and as long as supply stays tight, TPUs Google uses can free up NVIDIA GPU supply for other buyers — so Alphabet may limit how far NVIDIA can grow later more than it threatens today's sales, while the gross margin line is likely to show any pricing damage first.