NVDA Q2 2027 Earnings Analysis
NVIDIA delivered record Q2 revenue of $96B (+2x YoY) with data center up 18% QoQ; FY2028 guidance of ~70% growth constrained by supply, with gross margins expected to bottom in Q4 at 71-72%.
Key Metrics
要点总结
- Q2 revenue hit $96B with data center growing 18% QoQ; gross margin held at 75% despite component cost pressures.
- FY2028 revenue expected to grow ~70% YoY constrained by supply; margins to bottom in Q4 at 71-72% before recovery.
- AWS adding 2M GPUs through Q2 FY2029; Vera Rubin ramp expected to be fastest in NVIDIA's history.
收听平台
可用语言
Transcript
// Full episode scriptBeta Finch: NVIDIA Q2 FY2027 Earnings Breakdown
Welcome to Beta Finch, your AI-powered earnings breakdown. I'm Alex.
And I'm Jordan. Today we're diving into NVIDIA's fiscal Q2 2027 print — and there's a lot to unpack.
Before we get into it — quick note. This podcast is AI-generated content for educational and entertainment purposes only. Nothing we discuss should be considered investment advice. Always do your own research and consult a qualified financial advisor before making any investment decisions.
Alright, with that out of the way — Alex, NVIDIA just put up $96 billion in quarterly revenue. That's more than double year-over-year.
And it's the fourth straight quarter of accelerating growth, which is honestly wild for a company already this large. Data center revenue alone hit $89 billion, up 18% sequentially. Split that into two buckets: hyperscale revenue was $49 billion, up 13%, and then this other segment they call ACIE — that's NeoClouds, enterprise, sovereign AI — came in at $40 billion, up 25% sequentially and 138% year-over-year.
That ACIE number jumps out to me. It's growing faster than the hyperscale business, and CFO Colette Kress made a point of saying it now represents roughly half of NVIDIA's data center business. That's not a side hustle anymore — sovereign AI alone tripled year-over-year.
Right, and they're guiding Q3 revenue to $108 billion, plus or minus 2%. For the full fiscal year 2028, they're now projecting about 70% revenue growth — but here's the catch, Jordan.
The catch is that's a supply-constrained number, not a demand number. Jensen Huang was pretty blunt on the call — actual demand is closer to 100% growth. They simply can't build enough. Every cloud they serve is fully utilized.
Which is a good problem to have, but it does mean gross margins are getting squeezed. Memory pricing — DRAM specifically — has spiked way beyond what they expected. Gross margin is guided down to 74% in Q3, and they said it'll bottom out around 71-72% in Q4 before recovering to 72-73% next fiscal year once price increases kick in.
That's the number I think investors are going to fixate on. Component costs eating into margins is a real headwind, even with revenue this strong.
Let's talk strategy, because there's a lot of new stuff here. First, the AWS expansion — Amazon's deploying an additional 2 million GPUs through fiscal 2029, plus their new Vera CPUs, some paired with the upcoming Rubin GPU.
And Vera Rubin itself is the big story architecturally. They started production shipments this month, and Jensen's calling it the fastest product ramp in company history — already has purchase orders from every major hyperscaler, cloud, and OEM. The pitch is 30x higher throughput per megawatt and 35x lower token cost versus the previous Blackwell Ultra generation.
There's also this revenue-per-gigawatt trend they keep highlighting. Hopper was about $18 billion of NVIDIA revenue opportunity per gigawatt of data center capacity. Blackwell pushed that to $25 billion. Vera Rubin is $40 billion.
That's the flywheel — each generation captures a bigger slice of the data center build-out, not just GPUs but CPUs, networking, even their new Groq LPU line for high-speed inference. It's less "sell a chip" and more "sell the whole factory."
Speaking of financing — this is where it gets a little more complicated. NVIDIA disclosed nearly $50 billion invested directly into frontier AI labs, plus new financing partnerships with six major capital providers — Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, KKR — aiming to raise over $500 billion in third-party capital for AI infrastructure.
And they addressed the elephant in the room directly. Analyst Vivek Arya basically asked "isn't this circular financing?" — NVIDIA investing in labs that then buy NVIDIA chips. Jensen's answer was essentially: yes, we know what it looks like, but we think the returns justify it because these will be the biggest tech companies in history.
That's a confident stance, but it's also a real risk worth flagging for listeners — concentration risk in a handful of AI lab customers, some of whom are also designing their own custom chips. OpenAI's Jalapeño chip came up specifically in the Q&A.
Jensen's response there was that NVIDIA's differentiator isn't a single chip — it's a full-stack platform that runs every model, everywhere, across the whole AI lifecycle. He argued custom inference chips from single labs can't match that flexibility. Worth remembering that's the CEO's framing, not a neutral assessment.
One more thing worth flagging for listeners — China. Data center revenue to China was under 1% this quarter, and management said there's zero China data center revenue baked into the forward outlook. So that's already de-risked in the guide, for better or worse.
And on capital returns, they handed back a record $26 billion to shareholders in the quarter — $20 billion in buybacks, $6 billion in dividends.
So stepping back — the headline is: massive, accelerating demand, a supply-constrained but still enormous growth guide, margin pressure from memory costs, and NVIDIA doubling down on being the financier as well as the chip supplier for the AI buildout.
Before we sign off — everything discussed is AI-generated analysis for educational purposes. Past performance doesn't guarantee future results. Please do your own due diligence.
NVIDIA's next earnings call is scheduled for November 17, and Jensen's got a keynote in San Francisco next month we'll be watching for updates.
Thanks for listening to Beta Finch — we'll catch you next time.