Yesterday we wrote about Alphabet triggering an extreme percentile reading — extension and momentum both in the low teens, a setup that has historically preceded significant recoveries for that stock. The natural question: Microsoft just triggered the same alert. Is it the same opportunity?
It isn't. And the reason it isn't reveals something important about how different stocks behave under identical-looking technical conditions.
Microsoft is down 31.4% from its October peak of $541. It closed at $371 on March 25. Extension at the 5th percentile, momentum at the 14th. The composite scores say "weak" — trend at 6, flow at 19. The same flat label Google got. The same label Tesla got. The same label GLD got.
But when we run the historical comparison, Microsoft's data tells a story the composites can't: this stock isn't done falling.
Two months of cascading decline
The trajectory data tells you everything about the character of this selloff. On January 28, Microsoft was at $482. Extension at the 59th percentile. Momentum at the 66th. Flow at the 75th. Everything healthy.
Then the earnings report landed. Revenue up 18% to $81.3 billion. Net income surging 60% to $38.5 billion. Objectively strong numbers. But capital expenditure had hit $37.5 billion in a single quarter, and Microsoft Cloud gross margins slipped from 70% to 67%. The market didn't like the trade-off.
The stock dropped $50 in a single session — from $482 to $434. Extension went from the 59th percentile to the 1st. Momentum from the 66th to the 21st. Volatility exploded from the 67th to the 89th. It was the sharpest single-day percentile shift in Microsoft's recent history.
What happened next is more concerning than the initial drop. The stock never recovered. It bounced briefly to $414, then $401, then $397. Each rally was weaker than the last. Extension and momentum briefly recovered into the 30s and 40s in early March, flow climbed back above 60, and it looked — for about a week — like a base was forming.
Then a second wave hit. From March 11 to March 25, the stock fell from $405 to $371 in a grinding, persistent decline. Extension slid from the 27th back to the 5th. Momentum from the 55th to the 14th. Flow, which had recovered to the 65th, collapsed back to the 26th.
This is the pattern the percentile framework captures and the composites miss entirely. It's not one selloff. It's two. And the second one — happening now — is occurring with volatility still elevated at the 70th percentile. That specific combination has been dismal for Microsoft historically.
The volatility split
We tested every day since November 2021 where Microsoft's extension and momentum were both below the 20th percentile — 181 observations. Then we split by volatility regime.
Calm oversold (volatility below the 40th percentile): 35 observations. Hit rate at 21 days: 74.3%. Hit rate at 63 days: 88.2%. Median 63-day return: +7.75%.
Volatile oversold (volatility above the 50th percentile): 122 observations. Hit rate at 21 days: 38.3%. Hit rate at 63 days: 52.0%. Median 63-day return: +0.32%.
The difference is dramatic. When Microsoft reaches oversold conditions in a calm environment, it recovers nearly nine times out of ten within 63 days, with a median gain of almost 8%. When it reaches oversold conditions in a volatile environment, the odds are barely better than a coin flip and the median return is essentially zero.
Today's volatility is at the 70th percentile. That's firmly in the volatile regime — the one where oversold conditions don't reliably lead to recovery.
But it gets worse. When we narrow to the specific profile matching today — extension below 10, momentum below 20, flow between 15 and 35, volatility above 50 — there are 24 observations. The 21-day hit rate drops to 26%. The 63-day hit rate falls to 14%. The median 63-day return is -5.5%.
Out of 24 prior episodes that looked like today's Microsoft, only 3 were positive at 63 days.
Google versus Microsoft: same alert, opposite signal
This is the sharpest illustration of why a single "extreme reading" alert means nothing without the dimensional breakdown.
| Dimension | Google (Mar 24) | Microsoft (Mar 25) |
|---|---|---|
| Extension | 13th | 5th |
| Momentum | 14th | 14th |
| Flow | 33rd | 26th |
| Volatility | 30th | 70th |
| Risk profile | 91 | 13 |
Extension and momentum are nearly identical. Both are compressed, both would trigger the same "oversold" alert on any traditional platform. But the other three dimensions paint entirely different pictures.
Google's volatility at the 30th percentile tells you the selloff is orderly — a methodical repricing, not a panic. Microsoft's at the 70th tells you the environment is still chaotic, still dominated by forced selling and position unwinds.
Google's flow at the 33rd is depressed but holding. Microsoft's at the 26th is weaker and, critically, it was at the 65th just three weeks ago — the deterioration is accelerating.
And then there's the risk profile. Google's at 91 out of 100 tells you the underlying risk-adjusted return quality is still excellent — the drawdown hasn't damaged the Sortino ratio or capture metrics because the prior trend was so strong. Microsoft's at 13 out of 100 — the lowest we've seen in any of the four assets we've analysed this week — tells you the damage is real. The drawdown has been deep enough and prolonged enough that the risk-adjusted return metrics have genuinely deteriorated. This isn't a narrative repricing. The numbers are broken.
Why Microsoft isn't Google, despite looking similar
The fundamental stories explain why the same technical setup produces different outcomes.
Google's selloff is driven by macro (energy costs), capex concerns ($175 to $185 billion guidance), and the broad software rotation. But the operational results are strong: search revenue grew 17%, Cloud grew 48%, Gemini has 750 million users. The business is delivering. The market is re-rating the multiple, not questioning the business.
Microsoft's selloff has a deeper problem. The operational results were strong on paper — revenue up 18%, net income up 60% — but the details worry investors in ways that Google's don't.
The OpenAI partnership, once Microsoft's defining AI advantage, is fracturing. In February, OpenAI launched its "Frontier" enterprise platform and signed a $50 billion cloud deal with Amazon, arguing it fell outside its Microsoft contract. Microsoft's stake has been restructured to roughly 27%. A single OpenAI contract now represents about 40% of Microsoft's $625 billion backlog. If OpenAI's trajectory shifts, Microsoft's revenue visibility narrows considerably. This isn't an abstract risk — it's actively unfolding.
Copilot adoption has plateaued at 15 million paid users — just 3.3% of Microsoft's 450 million commercial customer base. For a product priced at a $30 monthly premium that was supposed to be the primary AI monetisation vehicle, the conversion rate is troublingly low. Enterprises appear to be waiting for more tangible productivity evidence before committing to wide-scale deployment. Microsoft's upcoming E7 enterprise package in May could be a catalyst, but it could also be a test the market watches closely for signs of demand.
Azure, the other pillar of the AI bull case, grew 38% on a constant currency basis — respectable, but unchanged from prior quarters despite massive capex acceleration. Management attributed this to diverting data centre capacity toward internal AI development rather than selling it externally. That's a strategic choice, but it means the capex isn't translating into visible revenue growth yet, which is precisely what spooked investors after the January earnings.
And then there's the broader "SaaSpocalypse" narrative — the fear that agentic AI will disrupt per-seat software licensing, the business model that Microsoft dominates more than any company on earth. Google faces this risk too, but Google's revenue is primarily advertising-based, not subscription-based. Microsoft's entire $80 billion Productivity and Business Processes segment is built on the per-seat model. If AI agents genuinely reduce the number of human seats needed for enterprise workflows, Microsoft has the most to lose.
The one setup that works — and why we're not there
There's an important finding buried in the data. The 5 observations where Microsoft was oversold with flow below the 20th percentile — a total washout — produced a 100% hit rate at both 21 and 63 days, with a median 63-day return of +8.9%. All five occurred in April 2024 when volatility was between the 9th and 24th percentile. Calm, washed out, everything exhausted. That was the buy.
Today, flow is at the 26th — depressed, but not washed out. Volatility is at the 70th — elevated, not calm. The risk profile is at 13 — broken, not merely compressed.
For Microsoft to set up the kind of oversold condition that has historically produced genuine recoveries, two things need to happen. First, volatility needs to come down substantially — from the current 70th to below 40. That would signal the chaotic phase of the selloff is over and the environment has stabilised. Second, flow needs to either wash out completely (dropping below 20, signalling full capitulation) or stabilise and begin recovering (climbing back above 35 to 40).
Neither is happening yet. The trajectory over the past two weeks shows flow declining from 44 to 26 and volatility climbing from 54 to 70. The selloff is intensifying, not resolving.
The earnings catalyst — and the risk
Microsoft's Q3 fiscal 2026 earnings are expected in late April. The report will answer the two questions the market needs resolved: is Azure growth accelerating to justify the capex (analysts want to see above 40% constant currency growth), and is Copilot adoption gaining momentum (anything materially above 15 million seats).
A strong report could break the pattern. If Azure shows demand acceleration and Copilot penetration reaches a meaningful inflection, the fundamental narrative shifts and the technical setup becomes less relevant. Markets can override historical patterns when the catalyst is powerful enough.
But the risk runs the other direction too. If Azure growth stays flat at 38% and Copilot adoption remains at 3%, the "show me the money" narrative hardens further. Morningstar's fair value of $600 and the analysts targeting $460+ assume these investments pay off. The market at $371 is starting to price the scenario where they don't — at least not on the originally expected timeline.
What this means
Microsoft and Google both triggered extreme readings this week. A traditional analysis platform would show the same signal for both: "weak," "oversold," "extreme." The composite scores are essentially identical — trend in single digits, flow in the teens to twenties.
The percentile framework sees something the composites can't. Google is in calm compression with a pristine risk profile — an 80% historical hit rate at 63 days. Microsoft is in volatile compression with a deteriorated risk profile — a 14% hit rate.
This isn't a call that Microsoft is a bad company or a permanently impaired business. Revenue is still growing 18%. Azure is a generational franchise. The AI infrastructure being built has genuine long-term value. Morningstar's $600 fair value and the analyst consensus above $460 may prove correct over a multi-year horizon.
But the data says this isn't the right kind of oversold for Microsoft. The volatility is too high. The risk metrics are too damaged. The flow is still weakening. The prior episodes that matched this profile produced a median decline of 5.5% over 63 days, with only 14% of observations positive.
The entry point that works for Microsoft historically — calm, washed out, low volatility — hasn't arrived. Getting there may require the Q1 earnings report to either resolve the uncertainty (bullish catalyst) or intensify it (driving the volatile selloff to its conclusion). Either way, the current setup favours patience over action.
Sometimes the most useful thing the data can tell you is: not yet.
Data as of March 25, 2026. Analysis based on MSFT 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.