Nvidia Corp. (NASDAQ:NVDA) just posted a quarter that would make any other chipmaker blush. On its August 26 earnings call, the company reported $96.2 billion in revenue, more than double what it made a year earlier, and said AI demand has crossed into something it calls an inflection point. But buried inside the good news sat two admissions that matter just as much: memory costs are rising faster than expected, and Nvidia is now underwriting some of its own customers' growth. Both cut against the simple growth story.
Data center revenue reached $89 billion, up 117% year over year, and the ACIE segment, which covers AI labs, cloud providers, industrial and enterprise customers outside the big hyperscalers, grew 138% year over year to $40.0 billion. Management said that segment now represents roughly half of Nvidia's data center business, a sign that governments and specialized cloud operators are becoming nearly as important as Amazon or Microsoft. Sovereign AI revenue, sold mostly through regional NeoCloud partners, grew 35% sequentially and more than tripled from a year ago, and those partners are expected to exit the year with 8 gigawatts of installed capacity, up from roughly 3 gigawatts at the end of 2025.
The bigger shift is how much of each data center dollar Nvidia now keeps for itself. Management said the revenue potential per gigawatt of capacity has climbed from $18 billion in the Hopper generation to $40 billion with the upcoming Vera Rubin platform, as Nvidia sells the CPUs, networking gear and software around its chips rather than just the chips themselves. Networking revenue hit a record, up 18% sequentially, with Spectrum-X Ethernet sales growing 2.6 times year over year. Amazon deepened its own commitment too, agreeing to deploy an additional 2 million Nvidia GPUs through the second quarter of fiscal 2029 alongside new Vera CPUs, while adopting Nvidia's Omniverse and robotics software for its warehouse fleet.
None of this looks like a company running out of runway. Nvidia returned $26 billion to shareholders in the quarter, split between $20 billion in buybacks and $6 billion in dividends, with about $99 billion still left on its repurchase authorization. Global venture funding into AI topped $400 billion in the first half of 2026 alone, with roughly 70% of that money earmarked for compute, which happens to be exactly what Nvidia sells.
Growth this fast is not free. Gross margin held at 75% this quarter, but CFO Colette Kress told investors it will bottom out at 71% to 72% in the fourth quarter as memory component costs spike, and that the size of those price increases has already exceeded the company's own expectations and is set to climb further into next year. Operating expenses are rising too, up 11% sequentially to $8.2 billion, with guidance near $9 billion for the next quarter, and inventory swelled to $31.6 billion as Nvidia stocks up ahead of the Vera Rubin launch.
#vera
Data center revenue reached $89 billion, up 117% year over year, and the ACIE segment, which covers AI labs, cloud providers, industrial and enterprise customers outside the big hyperscalers, grew 138% year over year to $40.0 billion. Management said that segment now represents roughly half of Nvidia's data center business, a sign that governments and specialized cloud operators are becoming nearly as important as Amazon or Microsoft. Sovereign AI revenue, sold mostly through regional NeoCloud partners, grew 35% sequentially and more than tripled from a year ago, and those partners are expected to exit the year with 8 gigawatts of installed capacity, up from roughly 3 gigawatts at the end of 2025.
The bigger shift is how much of each data center dollar Nvidia now keeps for itself. Management said the revenue potential per gigawatt of capacity has climbed from $18 billion in the Hopper generation to $40 billion with the upcoming Vera Rubin platform, as Nvidia sells the CPUs, networking gear and software around its chips rather than just the chips themselves. Networking revenue hit a record, up 18% sequentially, with Spectrum-X Ethernet sales growing 2.6 times year over year. Amazon deepened its own commitment too, agreeing to deploy an additional 2 million Nvidia GPUs through the second quarter of fiscal 2029 alongside new Vera CPUs, while adopting Nvidia's Omniverse and robotics software for its warehouse fleet.
None of this looks like a company running out of runway. Nvidia returned $26 billion to shareholders in the quarter, split between $20 billion in buybacks and $6 billion in dividends, with about $99 billion still left on its repurchase authorization. Global venture funding into AI topped $400 billion in the first half of 2026 alone, with roughly 70% of that money earmarked for compute, which happens to be exactly what Nvidia sells.
Growth this fast is not free. Gross margin held at 75% this quarter, but CFO Colette Kress told investors it will bottom out at 71% to 72% in the fourth quarter as memory component costs spike, and that the size of those price increases has already exceeded the company's own expectations and is set to climb further into next year. Operating expenses are rising too, up 11% sequentially to $8.2 billion, with guidance near $9 billion for the next quarter, and inventory swelled to $31.6 billion as Nvidia stocks up ahead of the Vera Rubin launch.
#vera
3 hours ago
