Forecasting a company's quarterly revenue is easiest in the middle of a trend.
When business is steady, last quarter is a good guide to next quarter, and Wall Street's consensus is hard to beat.
But the real game is at the turns: the moment a boom rolls over, or a slump starts to recover. That is where most analyst models get caught, because they extend the old direction for too long.
We are publishing this before it is proven because that is the only honest way to test it. We don’t want to show you a polished backtest. Instead we’re showing the method, making forward calls, and letting our record accumulate in public.
Why one company is not enough
A sell-side analyst usually knows a narrow set of companies deeply. That is the strength of the model. It is also the limitation.
A memory maker like Micron does not turn in isolation. Its revenue is driven by demand that shows up first across the rest of the chain:
what its customers are buying
what its competitors are signalling
orders flowing to the equipment needed to build the chips.
The turn often appears in the supply chain before it appears in any single company's reported revenue. That’s where our model lives. Tessara reads the chain as a system, instead of siloed company by company.
A worked example: Micron, 2022 to 2024
The clearest way to show what our model does is to replay a cycle that has already happened…
Micron’s 2022 to 2024 memory cycle is a textbook case.
In June 2022, Micron reported peak quarterly revenue of $8.6B. On the surface, the business still looked strong. But Tessara’s demand read (historical) had already peaked one quarter earlier and was moving down before the revenue decline appeared.
The revenue collapse came next. Micron’s sales fell 23%, then 38%, and continued sliding for a year, eventually bottoming near $3.7B. From peak to trough, quarterly revenue fell 57%. Gross margin moved even harder, from +47% to -31%, an almost 80-point swing.
Then the signal reversed.
At the bottom in 2023, Micron’s revenue looked dead flat - $3.7B, $3.8B, $4.0B across three quarters. But Tessara’s demand read bottomed one quarter into that stretch, then jumped roughly 15 points while reported sales were still scraping along the floor.
Revenue followed. One quarter later, Micron grew 18%, then 23%, and by late 2024 quarterly revenue was back above $7.8B.
At both turns, the demand read moved first. It weakened before the revenue collapse showed up, and it recovered before the revenue rebound appeared. In that cycle, the lead time was roughly one quarter.
What this is, and what it is not
We want to be precise here, because it is easy to oversell.
This is one company, one cycle, and two turns inside that cycle. It shows the mechanism Tessara is built around: a cross-chain demand read can move before a single company’s reported revenue, because the model is watching customers, competitors, suppliers, equipment makers, and end-market commentary together. Not just Micron in isolation.
That matters. But it is not yet a proven edge.
The read is built partly from the same earnings calls, trade press, and supply-chain commentary that any serious analyst could read. The difference is breadth, structure, and discipline: Tessara reads across the whole chain at once, calibrated across history and across companies.
Why the forecast starts at the bottleneck
Let me spend a moment to explain why our model can do that at all.
We start from a different place than The Street. A sell-side analyst models one company from the inside out: its products, its pricing, its costs, line by line.
Earnings are downstream of physical constraints.
A company’s revenue is where the cycle shows up late. The bottleneck is where it shows up first: HBM availability, wafer capacity, packaging lead times, substrate tightness, power delivery, networking gear, freight flows, spot prices, customer pull-ins, supplier commentary.
So Tessara starts there.
We track more than 80 chokepoints across memory, packaging, foundry, power, and networking. Each one is scored from real-world evidence: prices, lead times, capacity plans, customs flows, and what suppliers, customers, and competitors are saying.
Then we map those constraints onto companies.
HBM constraint read on Tessara
Our forecast is not built from the income statement outward. It is derived from the constraints the company is exposed to.
That creates three advantages.
One read travels across the whole chain. If HBM is sold out, that does not matter only for Micron. It hits memory makers, packaging companies, equipment suppliers, hyperscalers, accelerator vendors, and server OEMs, all in different directions. We map who benefits, who gets squeezed, and how the signal propagates.
The cycle turns before the numbers do. The AI buildout does not turn one company at a time. It rolls through the stack. The first signs appear in the constraints: a lead time stretches, a price slips, an order is pulled forward, a supplier changes tone. By the time it reaches reported revenue, the move has already started.
We can be wrong out loud. We have no direct relationships with the companies we analyse and no incentive to smooth the message. When the evidence moves against consensus, we say so and put a number to it.
But Tessara does not ignore Wall St consensus. Analysts do serious work, and their models are the right starting point.
Our constraint read is the tilt.
When bottlenecks say conditions are tighter or looser than consensus implies, we move the forecast. How far we move it depends on two things: the quality of the evidence, and the strength of our live, graded record in that domain
Tessara's Forecast for Micron Q3 FY2026
Micron reports Q3 FY2026 on 24 June 2026. This is our first public test.
The memory cycle today is the opposite of 2022. The cycle is supply-bound. Micron's last quarter printed $23.9B in revenue, up 76% from the quarter before, at a 75% gross margin.
Our constraint model expects Micron’s revenue to beat the street’s consensus by ~11%, coming in at $37B-$41B
Q3 FY2026 print | Revenue |
|---|---|
Micron's own guide | $33.5B |
Street consensus | $35.6B |
Tessara forecast (p50) | ~$39.4B |
Tessara range (p10 to p90) | $37.3B to $41.7B |
Several factors drive our above-consensus call:
Our Memory Regime Score is near the top of its range (81 out of 100, indicating a "sold out" regime)
The underlying constraint reads for HBM and leading-edge DRAM are pinned at their highest levels. Practically, this means that conditions are tighter than most of their history. This pushes our forecast range higher. This is corroborated across the chain: spot and contract pricing, the pace of orders to suppliers and equipment makers, export and customs flows, and demand from the largest buyers.
From the filings & news we track: Every unit of HBM Micron will build in 2026 is already spoken for. And the company guided 2026 DRAM prices up on the order of 200%.
When a company is supply-constrained and pricing is running, its own guidance is usually a floor. We looked at the long record of how this company's actual results have related to its guidance and factor that in
Our turn detector (designed to look for cycle inflections) is deliberately quiet. i.e. no signs of turning
What we are watching when Micron reports
Revenue versus the $35.6B Street estimate. This is the most important test. Above $38B confirms that cross-tier DRAM and HBM tightening is outrunning the new supply coming from Samsung and CXMT. At or below $35.6B, the tightening read was too hot and capacity is catching demand faster than we modeled.
Gross margin, against the 81% guide. At or above 81% says HBM and DDR5 pricing power is still flowing straight through to the bottom line. Below 78% is the first crack in that story.
The Q4 guide (the real forward tell). A guide that implies sequential growth from Q3 says the cycle extends into the back half of the year. A guide that implies sequential decline is the bear case landing, and it would matter even if Q3 itself prints strong.
HBM4 and DRAM pricing commentary. The HBM4 qualification and ramp timeline (does the customer base widen beyond the current high-volume accounts?) and where management points DRAM pricing into the next quarter. Those decide whether the pricing thesis carries into 2027 or stalls.
The risk to our own thesis, stated plainly. This could be the quarter the supply-deficit narrative peaks: HBM4 and the 1-gamma node are ramping into 2027, and at some point the new capacity catches the demand. A strong print with a soft forward guide would be exactly what that top looks like. We would rather name that risk now than explain it after.
If Tessara is right, it will not be because we built a prettier income-statement model. It will be because the physical system moved before the reported numbers did, and Tessara was watching the system.
Tessara is the research terminal for the AI buildout. We track what is binding across compute, memory, foundry, networking, packaging, and power, then map those constraints to the companies exposed. Get access today →
This article is for informational and research purposes only. It is not financial advice, investment advice, or a recommendation to buy or sell any security. Tessara Research does not publish price targets. The views expressed here reflect our analysis at the time of publication and may change as new evidence arrives. Readers should do their own research and consult a qualified financial adviser before making investment decisions.