NVIDIA Reports Quarterly Earnings: Revenue Beats Forecasts by 7%, Stock Surges 6.5%
Chipmaker earned $32.4 billion against expected $30.3 billion, and raised annual forecast for Blackwell shipments. NVDA shares closed at $158.20.
NVIDIA reported perfectly, but insiders are fleeing: anatomy of a 'silent' reversal you won't notice in the headlines
[The Gist]: What's Really Happening
You see the numbers: revenue $81.62 billion vs. forecast $78.42 billion — a 4% beat. EPS $1.87 vs. $1.76 — beats consensus. Data Center — $75.25 billion, up 92% year-over-year. Shares closed at $208.64. The market applauds. Bloggers scream 'new cycle.' But institutional traders' terminals show a different picture.
Over the past 90 days, 52 insider trades in NVIDIA — and ALL were sales. A director sold shares at $214 as early as May 27. CFO Colette Kress executed a series of planned sales worth hundreds of thousands of dollars in March. This is not 'dividends' or 'tax planning' — it's a systematic capital withdrawal by people who know where the bodies are buried.
Here's what's really happening: NVIDIA is undergoing a tectonic shift from a 'growth at any cost' story to an 'execution with no margin for error' story. The market has stopped rewarding the company for simply beating forecasts. Now it demands flawless execution on three fronts: Blackwell cooling, timely Rubin launch, and margin protection amid exponential cost growth. And on each of these three fronts, there are serious questions.
Analysts still give an average price target of $305.67. Bank of America sets $325. But the gap between 'Buy rating' and the actual behavior of smart money (insiders selling, Allstate Corp buying — but that's an insurer, not a tech hedge fund) has reached a critical level. My insider view: we are witnessing a classic divergence before a correction that mass media will only notice in hindsight.
Timeline and Context
To understand why a 'perfect quarter' could be a peak rather than a platform for growth, we need to look at the sequence of events. The table below is the hidden nerve of the story that won't be told in press releases.
| Date | Event | Market Reaction | What's Hidden |
|---|---|---|---|
| May 11, 2026 | The Information: Blackwell NVL72 overheats in 72-GPU configuration, requires redesign | NVDA -1.2% next day | Nvidia engineers confirmed the 'density' issue but publicly declined to comment |
| May 19, 2026 | Q1 FY2027 report: $81.62 billion revenue, +85% YoY | After-hours -1% (disappointment) | Market expected $83+ billion — expectations exceeded even the record result |
| May 20, 2026 | Conference call: dividend raised to $0.25, $80 billion buyback | +0.5% | Buyback signals lack of better investment ideas within the company |
| May 21, 2026 | Truist Financial raises target to $307, Wolfe Research to $275 | NVDA +1.8% | But both analysts lower multiplier ratios |
| May 28, 2026 | TrendForce: Rubin may be delayed due to HBM4 and cooling | Futures -1.5% | First systemic warning from an independent source on supply chain |
| June 4, 2026 | Ex-dividend date — shares trade without right to $0.25 | -0.8% | Mechanical decline, but coincided with start of sell-off |
| June 5–10, 2026 | Shares fall from $236 to $208 despite no bad news | -12% in a week | Market rates AI Capex as 'lumpy' rather than smooth |
| June 10, 2026 | Fed holds rate, geopolitics with Iran pressures oil | S&P falls 1.6%, NVDA loses 2.1% | Rotation from growth to value begins |
Key takeaway: the peak of optimism came on May 20–21, right after the report. Since then, each day brought either bad news (Blackwell overheating, Rubin risks) or no good news. And in the absence of news, the market by default sells the leader that has been a leader for too long.
Who Wins and Who Loses
Winners (explicit and implicit):
- AMD and Intel: while NVIDIA spends engineering resources fixing Blackwell and preparing Rubin, its competitors gain a window. AMD MI350 (announced for 2026) could capture 10-15% of the market, especially in inference, where cooling requirements are lower. Intel Gaudi 3 targets cost-sensitive enterprise clients — those unwilling to pay a premium for the NVIDIA brand.
- Broadcom and Marvell: custom ASICs for cloud giants are a growing trend. Microsoft already tested AMD and its own chips instead of NVIDIA in early 2026. Broadcom trades at a lower multiple than NVIDIA and grows faster in the custom silicon segment.
- Insurers and pension funds: Allstate Corp increased its NVDA position by 109% in Q4. Norway's sovereign wealth fund bought $51 billion worth of shares. These players don't trade on news — they enter for the long haul. For them, the current correction is an opportunity, not a disaster.
Losers:
- Retail investors who bought at highs: those who entered at $230–236 in late May are already seeing a 10-12% loss. If the market re-rates the multiplier for AI infrastructure, the drop could be 20-25%.
- NVIDIA ecosystem suppliers: Foxconn, Quanta, Wistron — Taiwanese server assemblers. If Rubin slips, their revenue in H2 2026 will be under pressure. Foxconn shares have already lost 6% since news of delays broke.
- Investors in leveraged semiconductor ETFs: funds like SOXL (3x long semi) amplify gains, but on a 10% correction, they fall 30%. Over the past two weeks, SOXL lost 18%.
Who's in the 'gray zone': Microsoft, Google, Amazon, Meta. They are simultaneously NVIDIA's largest customers (buying Blackwell and Rubin) and its competitors (developing their own chips — TPU, Trainium, MTIA). If NVIDIA slows down, these companies will simply switch faster to in-house silicon. Their loss won't be explicit — it will be hidden in reports under 'reduced Capex on external equipment.'
What the Media Isn't Saying
First. Insiders are not selling because the company is bad. They are selling because the next 6-9 months will be turbulent. CFO Colette Kress sold shares worth hundreds of thousands of dollars in March at prices $171–177. Now shares have fallen to $208 after rising to $236. She wasn't trying to 'catch the high' — she was locking in profits according to a plan set in 2025. But the fact that no top manager has bought shares on the open market in the last six months speaks volumes. If insiders believed in +50% over a year, they would be buying more. They are not.
Second. The NVL72 redesign is not a 'normal engineering iteration.' It is a scaling crisis. The Information reported that 72 GPUs in one rack overheat, and Nvidia demanded 'multiple changes' from suppliers in the design. The company calls it 'routine engineering.' But the fact that information leaked, and from the company's own employees, points to internal tension. Engineers are worried. Clients (Microsoft, CoreWeave) are worried. The public only hears 'demand exceeds supply.'
Third. $1 trillion in revenue from Blackwell and Rubin by 2027 is marketing, not a plan. Joe Jensen announced this figure at GTC. But TrendForce has already lowered its forecast for Rubin's share of total shipments from 29% to 22% due to delays with HBM4 and cooling. If Rubin comes later, the $1 trillion is achieved either through higher prices (which reduces demand) or through selling older Blackwell longer (which reduces margins as competitors catch up). Media write 'NVIDIA sees $1 trillion' — they don't write 'but not in 2027, rather in 2028, and with a 60% margin, not 75%.'
Fourth, most importantly. The market is changing the valuation rules for AI companies. Previously, NVIDIA was valued as a growth company with a P/E of 40-50. Now the forward P/E is 25, PEG is 0.7. That's a 'value company' valuation, not a growth leader. Reason: investors realized that AI Capex is cyclical, not linear. Hyperscalers buy chips in waves — first build one data center, then pause, then a second. NVIDIA cannot control this rhythm. And the market is now pricing in a discount for this uncertainty. This shift is structural. It won't disappear after one good quarter.
Forecast: Next 30 Days and 90 Days
Next 30 days (until July 11):
- NVIDIA shares will trade in the range of $195–$225. The upper bound is achievable only with no negative news about Rubin and a general recovery in the Nasdaq.
- Key trigger: any leaks from Taiwanese ODMs (Foxconn, Quanta) about the start of Rubin production. If they confirm July-August — that's positive. If a delay to September or later — expect a test of support at $190–$195.
- Watch for Microsoft, Google, and Amazon reports (due in July). Their Capex forecasts will encode demand for NVIDIA. If they announce 10-15% QoQ growth in AI Capex — NVDA goes up. If they cut or redirect to their own chips — NVDA falls 5-7% in a day.
Next 90 days (until September 2026):
- Rubin is the main catalyst. If shipments start in July-August as planned and initial customer feedback is positive, shares could return to $240–$250 by the end of Q3.
- If Rubin is delayed to 2027 (and TrendForce points to this risk), NVIDIA is stuck with Blackwell for all of 2026. In this scenario, competitors (AMD MI350, Google TPU v7) will bite off 15-20% market share. Year-end price target: $170–$180.
- External factor: geopolitics and chip trade with China. NVIDIA has already excluded Data Center compute revenue from China in its forecasts. If sanctions tighten, losing 10-15% of the market becomes permanent. Compensating with growth in the US and Europe will be difficult.
Editorial Forecast
Asset: NVIDIA shares (NVDA). Direction: moderate decline in the next 24–72 hours to the $200–$205 zone. Key levels: resistance — $212 (50-day moving average), support — $200 (psychological level), if broken — $195. Confidence level: medium (55%). Main risk: a sudden announcement of early Rubin shipments (e.g., June 15) could trigger a 5-8% rally. However, given insider sales and the structural re-rating of AI multipliers, the probability of a short-term decline is higher. We recommend watching levels $200 and $195 for a possible entry with a 3-6 month horizon, but not opening large positions until the July hyperscaler reports.
This analysis is the opinion of the editorial board and does not constitute individual investment advice. All decisions to buy or sell assets are your own.
— Editorial Team
