NVIDIA’s Earnings Reversal: Decoding the 70% FY2028 Guidance and the Future of AI Compute

Tuesday night, after the bell. NVIDIA earnings night is the closest thing the market has to a Super Bowl, and I’ve watched more than my share of them. This one broke the pattern: revenue $96.2 billion, up 106%; data center $89 billion; net income nearly doubled; next-quarter guidance a record $108 billion. Every number beat. The stock dropped anyway — straight to $203 in after-hours trading.

Then the CFO picked up the phone, said something that wasn’t in the press release, and everything changed — the stock ripped to $220 and the whole AI trade repriced.

If you sold at $203 that night, or chased the pop at $220 the next morning, this breakdown is for you: the gap between press release and tape has never felt wider than on this nvidia earnings night.

What Actually Happened on NVIDIA’s Earnings Night

Revenue of $96.2 billion beat the $92.1 billion consensus by roughly $4 billion — a mid-cap company’s entire quarter, as pocket change. Data center hit $89 billion, up 117%, now 92% of revenue: this is a pure AI compute company now, not a graphics-card maker. Net income: $59.62 billion, EPS $2.22. The Q3 guide of $108 billion crushed consensus — the first $100-billion quarter any chipmaker has ever guided to. All of it is in the official Q2 FY2027 release, and the SEC filing carries the same numbers.

So why did the stock fall first? Layer one: the beat was priced in. Options implied a ±5.4% move, the smallest earnings-day expectation since 2021, and beat margins have decayed from north of 20% two years ago to under 5% last quarter. A beat that is the default answer earns no reward. Layer two: the new information pointed down — the nvidia gross margin guide landed at 74%, a step below the 75% just reported, as NVIDIA absorbs climbing memory costs. One percentage point was all it took.

Then the call. The CFO put fiscal 2028 revenue growth at roughly 70% — the Street was modeling 45%. Against ~$397 billion of FY2027 consensus, that’s a ~$670 billion year. The tape went from $203 to $220 in minutes — memory, optical, server, and power stocks followed. The reversal wasn’t about the numbers; it was the growth endpoint moving a year further out.

Why the 70% Guidance Matters More Than the Beat

• One stock moves the whole index. NVIDIA is roughly 7–8% of the S&P 500 — it swings your funds before you touch anything.
• It neutralized the two biggest bear arguments at once — custom silicon and margin pressure are real, yet neither stopped a 70% guide.

My honest opinion: the 70% matters less than what it implies. Growth stocks trade on the distance to the finish line — NVIDIA just moved it a year out. That’s why the multiple holds.

Insider Strategies — How to Read the Next NVIDIA Report Before the Market Does

Here’s the playbook I ran that night — and will run next quarter: not predicting numbers, but reading what the market is already telling you.

1. Check the options-implied move before reading a single headline. My secret weapon: consensus tells you what analysts think, options tell you what money thinks. At ±5.4%, a routine beat had nowhere to go but down.
2. Never trade the press release — wait for the call transcript. The nvidia guidance number wasn’t in the release; the drop to $203 was an information vacuum, not a verdict. The CFO’s 70% projection exists only in the call — that asymmetry is the trade.
3. Separate margin from growth. One point of gross margin is a quarterly cost story; 70% growth is structural demand. If gross margin holds near 70% for two quarters, NVIDIA is winning the memory pass-through war.
4. Track supply commitments — and ask who funds the demand. Commitments jumped from $119 billion to $279 billion, but NVIDIA also helps finance roughly $500 billion of customer data center projects. Is this revenue customer-earned or NVIDIA-brokered? Even the famous shorts hedge with NVIDIA calls.

Common Mistakes I’ve Made (So You Don’t Have To)

1. Selling the after-hours dip before the call ended. I’ve made this exact mistake — watching $203 become $220 an hour after I sold is a feeling you don’t forget. Fix: no earnings-night trade until the transcript is out.

2. Breaking a long-term thesis over one percentage point. Margin whiplash is how positions get donated. Fix: sort your signals — endpoint guidance is structural, quarterly cost lines are noise until they trend.
3. Shorting the “NVIDIA killer” narrative. It feels smart until guidance jumps 25 points. Fix: think both/and — mature workloads escape NVIDIA, frontier workloads deepen its grip, from orbital data centers to the next Rubin buildout. Both are true while the pie grows.

4. Ignoring who funds the demand. Fix: ask the funded-versus-organic question before celebrating. It’s the difference between a cycle and a franchise.

FAQ

Why did the stock drop after NVIDIA beat on everything?
Because the beat was priced in: with a ±5.4% implied move and shrinking beat margins, only genuinely new information could move the stock — and it pointed down, until the call added a new number.

Is the OpenAI–Broadcom custom chip a threat?
It’s an ai inference chip — 700 watts, 1.5–1.9x better performance per watt, lower latency than the flagship — deploying by end of 2026. Yet NVIDIA raised its own guidance anyway: custom ai chips are eating the increment of a fast-growing market, not its share — for now. The first benchmark results are public if you want the raw numbers.

Why is gross margin declining?
Memory costs. NVIDIA raised AI server prices roughly 15% yet still guided margins down — it’s absorbing part of the bill. Stabilization near 70% means the pass-through is working.

Is this an AI bubble?
The guidance extended the growth endpoint rather than confirming a peak. The test: margin stabilization, independent chip benchmarks, and the Rubin ramp.

Final Thoughts

My verdict: the bull case got stronger on this call, but the honest question — how much of this demand is genuinely earned — didn’t get answered. I’m holding, and respecting the funding question. Next quarter, three signals: gross margin stabilizing near 70%, independent benchmarks of the custom inference chips, and the Rubin ramp landing on that $108 billion quarter. That scorecard is the whole game.

We runs a small Telegram channel where our mob talks through exactly these kinds of market moves and setups in real time — worth a look if you want to follow along between reports. And bookmark this page; We’ll update the scorecard every quarter.