Which Semiconductor Stocks Saw the Biggest Estimate Revisions This Quarter?

Wall Street's semiconductor forecasts did not simply move higher in the third quarter of 2026. They split into three distinct groups.

Memory producers experienced the largest reset. AI compute and custom-silicon companies recorded meaningful but smaller upgrades. Semiconductor-equipment estimates improved more gradually, even as demand for additional fabrication capacity remained strong.

AlgovestIQ's 90-day estimate-revision data places Micron Technology (MU) at the top of the screen by a wide margin. SanDisk (SNDK) ranks second, followed by Broadcom (AVGO), Nvidia (NVDA) and Advanced Micro Devices (AMD). Lam Research and Applied Materials also show positive revisions, but their changes are much smaller than those of the memory leaders.

Every figure in this analysis is drawn from a page any reader can open without an account. Each number is linked to the page that displays it.

The Q3 2026 revision leaders

The table below uses the 90-day revision field and current upcoming-quarter consensus EPS displayed on AlgovestIQ's stock catalyst pages. The window ends with the latest available AlgovestIQ snapshot through September 17, 2026, so it closely approximates — but does not exactly match — the calendar quarter beginning July 1.

The "90 days earlier" column is derived from the current estimate and the revision rate, not read from a stored historical snapshot.

Sources: AlgovestIQ catalyst pages for MU, SNDK, AVGO, NVDA, AMD, LRCX and AMAT. The AIQ Score is a descriptive snapshot and can update independently of estimate data.

The ranking answers one question: whose near-term earnings expectations changed the most? It does not answer whether a stock is attractively valued or likely to outperform. A large revision can follow an earnings surprise, a change in guidance, a low initial estimate, or all three.

1. Micron: the quarter's dominant estimate reset

Micron's +117.6% 90-day revision is in a different category from the rest of the group — more than three times the largest move outside memory. Its current upcoming-quarter consensus stands at $31.14 per share, against roughly $14.31 ninety days earlier, according to AlgovestIQ's MU catalyst tracker.

The size of the change reflects how quickly the earnings baseline became obsolete. In fiscal Q3 2026, Micron reported revenue of $41.46 billion, up from $23.86 billion in the prior quarter. Non-GAAP EPS reached $25.11, compared with the $20.98 consensus captured in AlgovestIQ's earnings history — a 19.7% surprise. Micron then guided fiscal Q4 revenue to $50.0 billion, plus or minus $1.0 billion, and non-GAAP EPS to $31.00, plus or minus $1.00. Those figures are directly consistent with the subsequent consensus reset. Micron's earnings release also showed GAAP gross margin rising to 84.6% from 74.4% in the prior quarter. What management said on the call is covered in our MU fiscal Q3 earnings insight.

This was not simply a revenue story. It was an earnings-conversion story. Higher DRAM and NAND pricing, a richer data-center mix and tight supply pushed incremental revenue through at unusually high margins. Micron also cited multi-year strategic customer agreements as a source of improved visibility.

AlgovestIQ's broader read is constructive but not unqualified. MU's stock analysis carried an AIQ Score of 65/100, and its risk page showed beta of 2.21 with one-year volatility of 81.7% — among the highest in this group.

That combination matters. Micron had the strongest revision signal in the screen, but it also carried one of the clearest reminders that a favorable earnings cycle can coexist with substantial price risk.

2. SanDisk: NAND and data-center storage drive the second-largest upgrade

SanDisk's upcoming-quarter EPS estimate reached $46.23, accompanied by a +92.8% 90-day revision on AlgovestIQ's SNDK catalyst page. Ninety days earlier the same consensus sat near $23.98.

Its latest reported quarter explains much of the move. SanDisk generated fiscal Q4 2026 revenue of $8.97 billion, up 51% sequentially and 372% year over year. Non-GAAP diluted EPS was $39.25, exceeding the $34.96 estimate in AlgovestIQ's earnings history by 12.3%. Data-center revenue more than doubled sequentially to $2.98 billion.

Management then guided fiscal Q1 2027 revenue to $10.3 billion to $10.8 billion and non-GAAP EPS to $44 to $46. The midpoint of that guidance is close to the consensus now displayed by AlgovestIQ. SanDisk also said it had signed additional customer agreements and identified data center as a core growth pillar. The complete figures are available in SanDisk's fiscal Q4 release.

AlgovestIQ's SNDK fundamentals page adds an important distinction: the latest quarter showed revenue growth, margin expansion and cash generation moving together. Revenue rose 50.7% sequentially, gross profit increased 62.6%, and free cash flow increased 136.7%.

As with Micron, the earnings revision was powerful because both price and mix changed at once. The risk is equally familiar: NAND remains cyclical, and expectations can reverse when supply catches up or contract pricing weakens. The two memory names can be read against each other on the MU vs SNDK comparison.

3. Broadcom: custom AI silicon lifts the next-quarter bar

Broadcom's upcoming-quarter EPS estimate rose 37.2% over 90 days to $3.82, placing it third in the screen and first among the non-memory names.

Broadcom's fiscal Q3 results supplied a clear fundamental catalyst. Revenue reached $29.6 billion, up 86% year over year, while non-GAAP EPS increased 96% to $3.32 against a $3.22 estimate. AI semiconductor revenue rose 221% year over year and 54% sequentially to $16.7 billion. Management expected that figure to reach $21.7 billion in Q4 and guided total quarterly revenue to approximately $34.8 billion. Those figures come from Broadcom's fiscal Q3 release, and the call itself is covered in our AVGO fiscal Q3 earnings insight.

Yet the price evidence pointed the other way. On the September 17 close, AVGO's technicals page showed the stock at $347.30, below both its 50-day moving average of $380.62 and its 200-day average of $369.24, with RSI (14) at 32.31 and MACD negative at -10.54. Analysts raised earnings expectations while the price trend weakened.

That divergence is worth watching. Positive estimate revisions are supportive, but they do not guarantee a positive near-term price response when valuation, positioning or expectations are already demanding.

4. Nvidia: large dollar growth, smaller percentage revision

Nvidia's current upcoming-quarter EPS estimate of $2.47 represented a +34.5% 90-day revision in AlgovestIQ's NVDA catalyst data.

That change is substantial, but smaller than Micron's or SanDisk's because Nvidia entered the quarter with an already elevated earnings base. AlgovestIQ's earnings history shows Nvidia reporting $2.22 in the August quarter versus a $2.09 estimate, a 6.2% surprise, on revenue of $96.22 billion versus $92.27 billion expected. Nvidia's fiscal Q2 2027 release confirms the reported revenue and earnings figures, and our NVDA fiscal Q2 earnings insight covers what analysts pressed on.

This distinction is central to reading revision data. A company can contribute more dollars of sector earnings growth while producing a smaller percentage change in consensus because analysts already expected strong results. Percentage revisions identify where expectations changed fastest; they do not measure absolute earnings contribution. The scale difference is easiest to see side by side on the NVDA vs AVGO comparison.

5. AMD: revisions broaden beyond memory

AMD ranks fifth in the screen, with a +33.4% 90-day revision and a current upcoming-quarter EPS estimate of $1.90, up from roughly $1.42 ninety days earlier.

The driver is different from Micron and SanDisk. AMD's revision is tied more directly to compute demand and product mix than to memory pricing. The company reported Q2 2026 revenue of $11.54 billion, up 50% year over year, while non-GAAP EPS reached $1.66 against a $1.62 estimate. Data Center segment revenue rose 107% to $6.7 billion and represented 58% of company revenue. AMD attributed the growth to EPYC processors and Instinct GPUs and said it expected Data Center sales to accelerate in the second half. See AMD's Q2 release and our AMD Q2 earnings insight.

AMD is the one name in the screen where the estimate signal and the price signal point the same way. On the September 17 close, its technicals page showed the stock at $545.09, above both its 50-day average of $495.58 and its 200-day average of $352.92, with RSI (14) at 67.09 and MACD positive at +8.92. Estimates rose modestly by this screen's standards; the price had already moved. The NVDA vs AMD comparison sets the two AI-compute names against each other across the same factor set.

Equipment suppliers: positive revisions, but a smaller step-up

The semiconductor-equipment group also benefited from the AI infrastructure cycle, though its estimate changes were more measured.

Lam Research showed a +22.0% 90-day revision to $2.15 in upcoming-quarter EPS. Its AlgovestIQ catalyst page also recorded six consecutive covered EPS beats, averaging a 6.9% surprise. Applied Materials showed a +14.9% revision to $4.05, with six consecutive covered beats averaging 4.8% on its AMAT catalyst page; its most recent call is covered in the AMAT fiscal Q3 earnings insight.

The mechanism here is one step removed from the chip designers and memory producers. Rising demand for HBM, advanced DRAM, leading-edge logic and packaging eventually requires more deposition, etch, inspection and other fabrication equipment. Those capital-spending plans can be large and durable, but revenue recognition often follows customer capacity decisions with a lag.

The market and the earnings models can therefore move on different schedules. On the September 17 close, Lam traded at $269.31 against a 50-day average of $308.68, with RSI (14) at 28.40; Applied Materials traded at $417.40 against a 50-day average of $504.00, with RSI (14) at 28.01. Both sat in territory normally described as oversold while their earnings expectations were rising — the opposite of the names surfacing on the stocks gaining momentum screen.

What the revisions say about the semiconductor cycle

Three conclusions stand out.

1. Memory produced the largest forecasting error

Micron and SanDisk did not merely grow faster than expected. Their revenue, pricing and margin structure changed faster than analysts' models. That is why their estimate revisions were roughly three times those of Nvidia and Broadcom, even though the latter companies remain central to the AI infrastructure buildout.

2. AI demand is reaching several layers of the supply chain

The revision leaders span memory, flash storage, CPUs, GPUs, custom accelerators, networking and fabrication equipment. The common demand source is AI infrastructure, but the path to earnings differs across each layer.

For memory companies, pricing and utilization can drive sharp operating leverage. For AMD and Nvidia, the key variables include product ramps, accelerator volumes and data-center mix. For Broadcom, custom silicon and networking visibility matter. For equipment vendors, customer capital spending and shipment timing are critical.

3. Estimate strength and price strength are not interchangeable

Outside memory, the revision ranking and the price trend almost invert. Broadcom, Lam Research and Applied Materials all posted positive revisions while trading below their 50-day moving averages, two of them with RSI readings near 28. AMD posted the smallest revision in the AI-compute group while trading above both its moving averages. The market can discount stronger earnings before they appear in consensus, or it can focus on valuation, cycle duration and execution risk instead.

Risks that could reverse the trend

The strongest revisions often appear when operating leverage is high. That also makes the forecasts more sensitive to small changes in demand, price or supply.

The main risks are:

• ·Memory normalization: Additional DRAM or NAND supply could slow price increases and compress the margins behind the largest revisions.
• ·AI capital-spending concentration: A small group of hyperscalers and AI developers accounts for a large share of infrastructure demand.
• ·Execution risk: Product delays, qualification issues or supply constraints can shift revenue between quarters. The earnings calendar tracks when each of these companies next reports.
• ·Estimate-base effects: Percentage revisions can look extreme when the starting estimate was unusually low.
• ·Period and corporate-action comparability: Fiscal calendars, spin-offs, stock splits and share-count changes can distort cross-company comparisons. This screen excluded anomalous negative observations rather than forcing unlike periods into the same table.
• ·Valuation and volatility: Better estimates can already be priced in. Micron, for example, paired the strongest revision in the screen with one-year volatility of 81.7%.

The bottom line

The largest semiconductor-linked estimate revisions of Q3 2026 were concentrated in memory.

Micron was the clear leader, with SanDisk second. Broadcom, Nvidia and AMD formed the next tier, showing that positive revisions had broadened from memory pricing into AI compute, custom accelerators and networking — but at roughly a third of the memory names' magnitude. Lam Research and Applied Materials showed a smaller but still constructive improvement in the equipment layer.

The most important takeaway is not that every stock on the list should rise. It is that analyst models moved fastest where AI demand collided with constrained supply, improving product mix or unexpectedly strong operating leverage.

Investors should treat revisions as one evidence layer. The more complete question is whether stronger estimates are supported by cash flow, balance-sheet quality, valuation, price momentum and risk resilience. Registered users can review the complete AIQ component breakdown on each company's research page. The underlying pages are open to everyone: the stock research hub, the stock comparison tool, and the stock screener for building a wider list than the seven names here.

Sample and method

This analysis uses the "90-day revisions" and upcoming-quarter EPS consensus fields displayed on AlgovestIQ catalyst pages, with data available through September 17, 2026. The screen covers selected U.S.-listed semiconductor companies and adjacent memory hardware companies with comparable, available catalyst data. It is not a complete ranking of every listed semiconductor security, and it is not a random sample.

The 90-day measure compares the current upcoming-quarter consensus EPS estimate with the corresponding estimate approximately 90 days earlier. The earlier estimates in the table are derived as current estimate divided by one plus the revision rate; they are not read from a stored historical snapshot, and they inherit the precision of the displayed revision rate. Because companies use different fiscal calendars, "upcoming quarter" does not refer to the same fiscal period for every company, so the revision percentages are comparable as a measure of expectation change but not as a measure of the same accounting period.

Reported results, guidance and operating metrics were checked against company investor-relations releases, which are linked in full below. Price and technical readings are as of the September 17, 2026 close. AIQ Scores are descriptive research snapshots, not forecasts, and carry a September 17, 2026 as-of date; how each component is calculated is set out in the calculation and interpretation reference.

Limitations

A percentage revision is sensitive to its starting point: a low prior estimate produces a large percentage change without implying a large change in absolute earnings. The seven companies here do not share a fiscal calendar, so "upcoming quarter" spans reporting dates from September 30 to December 9, 2026. Consensus estimates change without notice, and both the estimates and the scores on this page may have moved since the as-of date. Cross-sectional readings of this kind do not demonstrate persistence, causality, investment suitability or future performance.

This article is for informational purposes only and is not investment advice or a recommendation to buy or sell any security. Historical earnings surprises and estimate revisions do not guarantee future results.