Nvidia's Fundamentals Are Perfect. So Why Is Money Quietly Leaving?

Record quarterly revenue of $68.1 billion. Data centre sales up 75% year-over-year. Forward guidance of $78 billion for Q1 — five billion above consensus. Supply commitments nearly doubling in a single quarter to $95.2 billion. Gross margins at 75%.

By every traditional measure, Nvidia just reported the most impressive earnings in the semiconductor industry's history. Jensen Huang told investors to expect sequential revenue growth throughout 2026 that would exceed the already staggering $500 billion Blackwell architecture commitment.

The stock is at $171. That's 17% below its October peak of $207. It hasn't made a new high in five months. And the money flow data tells you why.

The $170 level: nine months of battle

To understand what's happening with Nvidia, you have to zoom out from the current numbers and look at the price structure.

The stock first crossed $170 in mid-July 2025. Since then, it has tested the $170 to $175 zone at least eleven separate times — touching or breaching that level in July, August, September (twice), November, December (twice), February 2026, and March 2026 (three times in the last week alone).

The pattern is unmistakable. Nvidia runs up — to $183 in August, $188 in September, $202 in October, $191 in late January — and each time it rolls back toward $170. The highs are getting lower. $202, $191, $190, $183, $180, $175. The floor at $170 has held, but the ceiling is compressing toward it.

This is what technical analysts call a descending triangle — a pattern where buyers hold a level but sellers are becoming more aggressive at progressively lower prices. The pattern resolves when either the buyers give up (breakdown) or the sellers exhaust themselves (breakout). After nine months and eleven tests, one side is running out of patience.

What the flow data reveals

Here's the part that most analysis misses. The price chart shows a stock going sideways. The flow data shows something very different happening beneath the surface.

We track three independent flow indicators, each measuring a different dimension of buying versus selling pressure:

Chaikin Money Flow (CMF): Measures whether volume is associated with price closing near the high (buying pressure) or the low (selling pressure) of each day's range. A positive reading means money is flowing in. Negative means money is flowing out.

OBV Slope (21-day): The rate of change of On-Balance Volume — whether cumulative volume is trending up (accumulation) or down (distribution) over the past month.

Force Index (13-day): Combines price change with volume to measure the actual force behind moves. Large up days on heavy volume produce positive readings. Large down days on heavy volume produce negative readings.

We convert each into a rolling percentile — where today's reading sits relative to the past year of data. The average of the three is the flow percentile.

Here's Nvidia's flow trajectory since October 2025:

The pattern is distribution, and it's been going on for five months. Flow was healthy in early October at the 81st percentile. It has spent the majority of the time since then below the 40th. CMF has been negative for 20 of the last 25 weekly observations. Each brief recovery in flow (late December, late February) has been followed by a deeper decline.

This is what distribution looks like in practice. The stock price holds a range — $170 to $190 — while the flow indicators systematically deteriorate beneath it. Buyers keep defending $170. Sellers keep distributing on every rally toward $185 to $190. The CMF at -0.18 and the 8th percentile tells you that over the last 20 trading days, volume has been overwhelmingly associated with closes near the daily low. Money is leaving this stock on a net basis, and has been for months.

The obvious question: if the business is this good, why is money leaving?

There are several theories, and they're not mutually exclusive.

The first is simple maths. Nvidia's revenue has grown from $26 billion to $216 billion in three years. The stock is up 446,000% since its IPO. At some point, the magnitude of the run creates a gravitational pull toward mean reversion — not because anything is wrong, but because the stock has outperformed so dramatically that portfolio rebalancing forces are enormous. Every index fund, every pension, every balanced portfolio that holds Nvidia is now massively overweight relative to their targets. The selling may be mechanical, not judgemental.

The second is the capex anxiety that's hitting all AI infrastructure stocks. Nvidia's customers — Microsoft, Alphabet, Meta, Amazon — are committing $500 to $700 billion in combined capex for 2026. The market is no longer asking "Is AI demand real?" It's asking "Will the return on this spending justify it?" If Nvidia's customers eventually slow their purchasing because the ROI timeline extends, Nvidia is the most exposed company in the supply chain. The market may be pricing a deceleration that hasn't appeared in the earnings yet.

The third is competitive concern. While Nvidia dominates GPU-based AI compute, every hyperscaler is developing custom silicon. Google has TPUs. Amazon has Trainium and Inferentia. Microsoft has Maia. These aren't replacing Nvidia today, but each generation gets closer. Nvidia's gross margins at 75% are extraordinary — and extraordinary margins attract extraordinary competition.

The fourth — and this is what the flow data uniquely captures — is that large institutional holders may be taking profits over an extended period, selling into strength to avoid moving the market. This is exactly what distribution looks like: the price holds while the smart money exits gradually. You don't see it in the price chart. You see it in the CMF sitting at -0.18 and the flow percentile at the 10th.

What the percentile framework shows

Here's Nvidia's current positioning:

This is what our system flagged as a "triple low" — extension, momentum, and flow all below the 15th percentile simultaneously. It's the most extreme multi-dimension reading Nvidia has shown since December 2023.

The triple-low condition has occurred only 14 times in Nvidia's history since November 2021. The forward returns are striking: a 100% hit rate at 63 days with a median return of +37.7%. Every single time Nvidia hit this level of multi-dimension compression, the stock was substantially higher three months later.

But — and this is an important but — those prior episodes were different in character.

September/October 2021: Nvidia at $20, during the initial AI infrastructure buildout. The stock was early in its mega-rally. Triple-low conditions resolved with gains of 18% to 34% at 21 days and 43% to 53% at 63 days. This was a stock with enormous upward momentum temporarily interrupted.

September 2022: Nvidia at $13 to $14, during the post-COVID tech crash. Flow was weak, volatility was high (around 80th percentile). The stock was deeply oversold in a genuine bear market. It declined further in the short term (-3% to -13% at 21 days) before ripping higher (+17% to +38% at 63 days). Classic volatile washout recovery.

December 2023: Nvidia at $45. The single most powerful episode — the stock gained 15% at 21 days and 92% at 63 days. This was the launch point for the Blackwell-driven surge that took the stock to $207. Flow was deeply washed out, volatility was moderate (34th percentile), and the catalyst (AI demand acceleration) was immense.

February 2026 (last month): Nvidia at $172. Triple-low conditions resolved with a modest 2% to 6% gain at 21 days. We don't have the 63-day outcome yet. But the fact that the stock bounced to $190 and then fell back to $171 within a month is telling — the bounce was real, but it was sold into.

Today versus history

Today's Nvidia at $171 shares some characteristics with those prior episodes and diverges on others.

The similarities: extension at the 10th and momentum at the 8th are as compressed as most of the prior triple-low episodes. The risk profile score at 83 remains high — the underlying risk-adjusted return quality is strong, similar to the 2021 episodes. RSI at 39 is the lowest it's been since the September 2022 bear market.

The differences are significant. The prior triple-low episodes all occurred during either early-stage rallies (2021), genuine bear markets (2022), or before a massive catalyst (December 2023). Today's episode occurs after nine months of sideways distribution, with progressively lower highs, and with a CMF that has been persistently negative. The character of the selloff is different — it's not a sharp correction from a rally (like 2021 or 2023) and it's not a bear market capitulation (like 2022). It's a slow grind lower from a prolonged distribution range.

The volatility percentile adds nuance. At the 47th, it's right in the middle — neither the calm conditions (below 30th) that historically preceded the best recoveries for quality stocks, nor the volatile conditions (above 70th) that signal capitulation. It's ambiguous.

And the flow decomposition tells you which component is most damaged. CMF is at the 8th percentile — the buying-versus-selling pressure within each day's range is extremely negative. OBV slope is at the 11th — the trend of cumulative volume is declining. Force Index is at the 30th — less extreme, suggesting the selling isn't happening in dramatic fashion but rather as a persistent grind. This is consistent with institutional distribution rather than panic selling.

The question that defines the trade

Nvidia's situation comes down to a single question: is the $170 level a floor that will hold for a twelfth time, or is the distribution pattern about to resolve downward?

The fundamental case for the floor is overwhelming. Revenue guidance of $78 billion for Q1. Supply commitments of $95 billion (nearly doubled in one quarter). Hyperscalers committing $500 billion or more in combined AI capex. Data centre revenue scaled 13 times since ChatGPT emerged. Gross margins at 75%. No other semiconductor company in history has shown this combination of growth rate and profitability at this scale.

The technical case for a breakdown is building. Five months of declining flow. CMF persistently negative. Progressively lower highs compressing toward the floor. Eleven tests of the $170 level — each test weakens it, because the pool of buyers willing to step in at that price gradually depletes. The sellers have been patient, distributing over months rather than dumping all at once. That patience is what makes the pattern so concerning — it suggests the selling is deliberate, not panicked.

The historical triple-low data (100% hit rate at 63 days) argues for a recovery. But every prior episode had a clear catalyst or occurred during a different market regime. The current episode is the first to occur after a prolonged distribution pattern at a major structural level.

What to watch

Three things will determine whether the $170 floor holds or breaks.

CMF direction. The flow recovery is the single most important signal. If CMF turns positive over the next two weeks — even slightly — it would suggest the distribution phase is ending and buyers are regaining control of the daily ranges. If CMF stays below -0.10, the selling pressure is ongoing and the floor is under increasing stress.

Volume on the next test of $170. Nvidia touched $171.14 today — the lowest intraday of 2026. If the stock tests $170 again on increasing volume, it's a warning. Distribution patterns break down when the final test comes on heavy volume as the last buyers capitulate. If it tests $170 on light volume and bounces, the floor is still intact.

The May earnings report. Nvidia guided Q1 revenue to $78 billion. If the report delivers at or above that level with sustained margins, the fundamental case reasserts itself and the distribution pattern may resolve upward — similar to how the December 2023 triple-low preceded a massive rally into the Blackwell announcement. If there's any hint of deceleration — even a modest miss or softer guidance — the technical picture and the flow data suggest the stock is vulnerable to a meaningful move lower.

The honest read: Nvidia's fundamentals are the strongest of any company we track across 118 assets. But the flow data has been telling a different story for five months, and the pattern of progressively lower highs compressing toward a repeatedly-tested support level is one of the most reliable precursors to a breakdown in technical analysis. The market is deciding right now whether Nvidia's earnings power is sufficient to overcome the weight of five months of distribution.

The triple-low percentile reading says this is a historically rare condition that has always resolved higher. The flow data says something has been changing beneath the surface for months. Both are true. The resolution of that tension will likely be one of the most important moves in the market over the next quarter.

Data as of March 26, 2026. Analysis based on NVDA daily indicator data from November 2021 to present. Percentile calculations use a rolling 365-day window. This is not investment advice — it is a description of how the current conditions compare to historical episodes, based on our indicator framework. Past patterns do not guarantee future outcomes.