NVIDIA Corporation (NVDA)

News Summary

• Vera Rubin entered production shipments in August with purchase orders from every major hyperscaler, AI cloud, and OEM; Nvidia expects it to be the fastest product ramp in company history, contributing ~20% of data center revenue in Q3, and delivers 30x higher throughput per megawatt and 35x lower token costs versus Grace Blackwell Ultra — a direct benchmark for competitive positioning against alternative AI hardware.
• Jensen Huang declared that agentic AI requires 15–100x more compute per user interaction versus traditional prompting and stated that as of last month most AI workloads have crossed from prompted to agentic, framing this as a structural demand multiplier just beginning to hit infrastructure buildout; separately, Nvidia agreed to acquire Hugging Face for $12.9 billion, extending its reach into the open-source model and developer ecosystem.
• Demand is running at roughly 2x what supply can fulfill — the binding constraint is supply, not demand — with Nvidia guiding FY2028 revenue growth of ~70% YoY while Huang characterized unconstrained demand growth as closer to 100%; the company also flagged a gross margin headwind from extreme HBM memory pricing, guiding Q3 non-GAAP gross margin to 74% and a Q4 trough of 71–72% before recovering to 72–73% in FY2028 as price increases take effect.
• Q2 FY2027 revenue came in at $96.2B (+106% YoY, +18% QoQ), beating consensus of ~$92.2B; data center revenue hit a record $89.0B with ACIE (AI clouds, industrial, enterprise) up 138% YoY to $40.3B; Q3 guidance of $108B midpoint exceeded the Street's ~$104B estimate; GAAP operating margin expanded to 66.2% and the company returned a record $26B to shareholders.
• NVDA initially dipped ~1.2% in after-hours trading before reversing sharply to close the next session up ~7%, with the broader chip ecosystem rallying in sympathy (Micron +4.5%, Marvell +5.7%, CoreWeave +6%, Nebius +7.5%); press framing was uniformly bullish — WSJ called it a 'blowout quarter,' CNBC headlined a '7% jump' — while the key analyst concern flagged was the emerging threat from hyperscaler and OpenAI custom silicon to Nvidia's near-monopoly on frontier AI chips.

Financial Highlights

Earnings Call Highlights

This was a triumphant, almost euphoric call from NVIDIA management, with Jensen Huang projecting absolute confidence in the durability and acceleration of AI infrastructure demand while Colette Kress delivered a rare full-year forward revenue guide of approximately 70% growth for fiscal 2028 — a signal of extraordinary visibility. The dominant theme was that NVIDIA has transcended being a chip company and is now a full-stack AI factory platform capturing an expanding share of every gigawatt of data center capacity deployed globally, with revenue opportunity per gigawatt scaling from $18B at Hopper to $40B at Vera Rubin. Analysts probed hard on inference market share, open-source model risk, circular financing of frontier AI labs, and gross margin compression from memory pricing — all of which management addressed with unusual directness and confidence. The tone was one of controlled scarcity: demand is running at roughly 2x what supply can fulfill, and the constraint is explicitly supply, not demand. Sentiment on the call was overwhelmingly bullish, with management's posture being that every structural trend — agentic AI, sovereign AI, NeoCloud buildout, enterprise adoption — is simultaneously accelerating.

• NVIDIA guided fiscal 2028 revenue growth of approximately 70% year-over-year, explicitly characterizing this as supply-constrained and stating that unconstrained demand growth is running closer to 100%; Jensen said 'the unconstrained would be a lot higher' and that the company must work with its supply chain to close the gap.
• Vera Rubin commenced production shipments earlier in August with purchase orders already in hand from every major hyperscaler, AI cloud, and system OEM; NVIDIA expects it to be the fastest product ramp in company history, with Vera Rubin accounting for approximately 20% of data center revenue in Q3 and delivering 30x higher throughput per megawatt and 35x lower token costs versus Grace Blackwell Ultra.
• The Groq 3 LPX rack-scale LPU system entered full production and is already claiming nearly 4x tokens per second versus the next best alternative on the Artificial Analysis benchmark, with volume shipments expected later this quarter; Nebius will be the first customer, and Jensen framed it as a high-ASP, high-interactivity complement to the core Vera Rubin NVLink 72 platform rather than a replacement.
• NVIDIA disclosed a significant gross margin reset driven by extreme HBM memory pricing, guiding Q3 non-GAAP gross margins to 74%, bottoming in Q4 at 71-72%, before recovering to 72-73% in fiscal 2028 as executed price increases take effect in Q1; management was unusually direct, stating 'we want to be direct about this rather than let it linger as an open question.'
• Jensen described agentic AI as requiring 15 to 100 times more compute per user interaction versus traditional prompting, and stated that as of last month he believes most AI workloads have crossed from prompted to agentic — framing this inflection as a structural demand multiplier that is only beginning to be reflected in infrastructure buildout.

Sources

• Nvidia's to report Q2 earnings amid chip stock struggles - Yahoo Finance
• NVIDIA Announces Financial Results for Second Quarter ...
• Nvidia forecasts 70% sales growth next year, signals AI ...