Altair
Skip to content

Trading Analysis Report: MKSI

Analysis date: 2026-10-14 · Model: claude-opus-5-5 · AI-generated, not financial advice.

  • Analysis date: 2026-10-14
  • Rating: Underweight
  • Generated: 2026-10-14 17:36:12
  • TradingAgents 0.6.0: anthropic, deep claude-opus-5-5, quick claude-sonnet-5-5
  • Analysts: market, sentiment, news, fundamentals; research debate rounds 5, risk debate rounds 5
  • Data vendors: core_stock_apis yfinance, technical_indicators yfinance, fundamental_data sec_edgar,yfinance, news_data yfinance, macro_data fred, prediction_markets polymarket

V. Portfolio Manager Decision

Rating: Underweight

Executive Summary: Underweight MKSI. Cut to a residual of one-third of a standard allocation, or one-quarter if the checks below are not all run and passed by Friday morning or if the book already holds other high-beta semis. If you are flat, do not initiate. Sell about three-quarters of the reduction on Thu 10/15 and Fri 10/16 at VWAP, using 234.50 as the reference rather than 256.70, and rest the last quarter (no more than a sixth of standard) at 247-250 through Monday 10/19's close, selling anything unfilled at market then. A close below 228 finishes the reduction the next session and steps the residual down one rung; the residual goes to zero on two failed checks, on an adverse or unread 10-Q debt footnote by the last close before the Q3 release, or on a close in the 207-208 zone, and nothing is added before the print because size, not stops, is the defense against a 20-33% gap over the roughly 1-2 months through the Q3 print and Q4 guide.

PRICE TARGET: $229. Method 1, the valuation objective from the analysts' shared run-rate EV/EBITDA grid. Equal-weight mean of -11%, -2% and +6% is -2.3%, and 234.50 x 0.977 is about 229. Equal weights are used because the bull's weights assumed a guide stronger than his own gate. Cross-checks: - The 14x leg (234.50 x 0.89, about 209) sits on the 208.12 March low. - The 16x leg (about 249) sits on the 247-250 gap fill. - The 15x leg (about 230) sits on the 228.46-231.50 support zone. On the bull's weights the mean is about +1.7%, or roughly 238.5.

Current Price: 234.5

Price Target: 229.0

Confidence: Medium

Time Horizon: 1-2 months

IV. Risk Management Team Decision

Aggressive Analyst

Aggressive Analyst: Let me open, since nobody else has spoken yet. I'll also take on the objections I expect, because I think this Sell is more aggressive than it looks, and the cautious camps will miss that.

The conservative analyst will probably say that if you're worried about a beta-2 stock with a possible maturity wall, you should be out entirely. Waiting for a bounce to sell into is just hoping. I'd push back on that. Dumping the whole position at 234.50 means selling at the lower Bollinger band, with a daily z-score of -2.13 and the stock one point above the September closing low. That is the classic place where a forced seller hands the next buyer a gift. In a downtrend with ADX climbing from 7 to 23 in a week, bounces are violent. The September low produced a 22% rally, from 231.50 to 281.41, in about three weeks. This plan sells half the reduction now and puts the rest on resting limits at 256.70 and 261, about 9 to 11% higher. That is active risk management that gets paid for the volatility instead of just eating it. Selling the whole position to avoid risk gives up the one thing a beta of 2.03 reliably offers, which is big counter-trend moves.

The neutral analyst will probably say to hold the core position, wait for the 10-Q footnote and the Q3 print, and not trade on incomplete information. That sounds reasonable, but look at what waiting costs. Four single-day drops of roughly 10% since July: 07-02, 07-28, 07-29 and 09-14. The tape is bad on every tier: SuperTrend is down on the weekly, monthly and daily, the 50 SMA is under the 200, MACD crossed down on 10/13, and the retest of the September low is happening on weaker money flow, 0.343 against 0.505 to 0.530 the first time. The earnings gap alone could be 20 to 33% in either direction, and the position's size shouldn't be set by hope. Also, this plan doesn't claim certainty. It keeps about half the standard allocation, so there is still real exposure to a good print and a rebound, and the plan assumes a big gap on that remainder. That is a calculated bet on the upside, not a retreat.

Now the bull case, which I take seriously. Q2 was a breakout: revenue up 28%, operating margin 20.1%, EPS of $2.41 against $0.92 a year earlier. The forward P/E is 13.4x, and the Street's targets imply about 60% upside. Retail on StockTwits is saying accumulate. My answer is that most of this is already in the price, and the "cheap" numbers are fragile. The 13.4x depends entirely on forward EPS of $17.55 being delivered. On GAAP numbers, Q2 annualized is about 24x, and the trailing multiple is 37x. The 60% upside is mechanical target math, and those targets get cut after drops like this one. The retail sample is four posts with no bearish tags, which tells us almost nothing. And the stock didn't fall 47% from its June peak for no reason. A semis relative-strength slump since the MU miss, a Fed where the market prices 97% odds of no cuts and 86% odds of another hike, and an oil price near $103 are all hitting a high-multiple, high-beta name together.

The piece I care most about is the current liabilities jump, from $927M in December to $2.29B in March and $2.42B in June, with total liabilities flat. That looks like about $1.4B of debt rolling inside twelve months, on a company with roughly $4.2B of implied debt, $454M of cash and net debt around 3.5x EBITDA. I can't verify the cause from these tools, and it may well be benign, such as a planned refinancing. But that is exactly the situation where you size for the tail. The bold move isn't pretending the unknown is fine. It's cutting exposure first, then keeping a smaller stake that can profit if the lookups come back clean.

I'm not leaning on the insider selling, about $43M with no buys. It looks routine and post-vest, and it would be weak evidence either way.

I'll also admit what could go wrong. If the stock reverses sharply from here, the first half sold near 234.50 will look poor. The plan answers that with discipline: it finishes the trim only on a close below 228, and it exits the remainder below about 207 unless the debt footnote, Q3 guide and peer lookups have been checked and passed. Support below is thin, with the March lows at 208.12 and 200.16. Selling some here and some into 256 to 261 has better odds than either holding everything or dumping everything.

Overall, I think the case for the Sell is that a zero expected return on a 5% ATR, beta-2 stock with unresolved balance-sheet questions is poor use of capital, and trimming into strength while keeping a meaningful stake for the upside is the better trade. Whatever your standard position is, cut it to roughly half, and don't initiate if you're flat.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Let me start with what I got wrong, and then I'll go through where I think you're both still giving away too much.

The neutral is right that the stock is three points above the September closing low, not one. It's 1.3%, which is a quarter of an ATR, so that's noise around support and not a gift to the next buyer. The ADX point was also weak. ADX tells you trend strength, not bounce size, and I shouldn't have leaned on it. He's right that 256.70 is a moving target too. It's today's 10-day EMA, and it falls about four points a day if the stock sits here. And the conservative is right that a trade sheet headlined at 256.7 overstates what we're selling at. The anchor is 234.50 and everything above it is upside.

The conservative's best point is the sequencing. A close-based trigger at 228 is 6.50 away on a 12.25 ATR, while the 256-261 limits are nearly two ATRs up. On a coin-flip tape, the trigger beats the limit maybe three times out of four. That's my rough estimate, not a measured figure. So the lottery ticket should be small, and the neutral's fix of putting the bulk of the staged tranche near 247-250 on the gap fill and keeping only a small slice at 256-261 is better than what I opened with. I'll take it.

Where I won't follow the conservative is the full dump. Selling the whole reduction in one or two sessions means finishing before the first possible TD exhaustion date on the 20th, with a daily z-score near -2.1 on a gap-down bar. That is the forced-seller print I warned about. The neutral is also right that an intraday stop doesn't protect against gaps, because it fills at the open and not at your level. A stop 6.50 inside a 12-point daily range mostly shakes you out on noise. Against a 20-33% earnings gap, the real protection is size, and the plan already cuts the size.

The conservative's "give up a possible 10% bounce" understates the cost of going to zero. If you hold nothing into the print and the quarter is good, the stock gaps up and you rebuy "with better information" at prices that already reflect it. Information arrives together with the repricing. The last quarter in the data was the best one in it, with revenue up 28%, operating margin of 20.1% and EPS of $2.41 against $1.18 in Q1. The stock is now below where the CEO was selling in February on much lower earnings. I'm not saying Q3 will be good. I'm saying that after a 47% drawdown the bar is lower than it was at 444, and a zero residual means being fully short that gap relative to your standard weight.

On zero expected return, the conservative says the half has no edge because the whole has none. That's true, but zero on a grid centered on today's price is not negative. Going to zero is a claim that the expectation is below zero, and nobody has shown that. The footnote and the print will move the number, and we can read one of them this week. A third to a half of a standard position is a modest variance premium for staying exposed to that resolution. I'll also take back "calculated bet on the upside." It's not an edge. It's a hedge against being wrong about the sell.

On the debt, the conservative is right that it's a lookup, and the plan already says not to carry the remainder through the print unless the footnote has been read and passed. But I overweighted the item in my opening, and the neutral caught it. The reclassification was on the March balance sheet and the stock ran from about 315 in late May to 444 at the end of June. It isn't new information, so it doesn't explain the 47% decline. The coupon math is small too. Roughly $1.4B refinanced two points higher is about $28M pre-tax. On about 68M shares that's around 30 cents after tax, under 2% of the $17.55 forward EPS, though the two-point step is an assumption. The real risk is market access or covenants, which is a tail worth reading for. The cash flow worry is also narrower than it looks. Q2 free cash flow of $189M was the best quarter in the table, and TTM FCF of $480M is above TTM net income of $439M. The first-half weakness looks like a working-capital build during a 28% revenue ramp.

I'd push back on the neutral in two places. First, the "both bounces followed 2.5 million share washout days" point doesn't line up cleanly. The 2.71M day was 9/22, a week after the 9/15 closing low, unless that day was a retest I can't see in the summary. That makes the precedent shakier than he describes, though it also means low volume today doesn't prove the low isn't in. Second, I'd rather the residual downgrade depend on what the footnote says than on the clock. If it's unread a few days before earnings, that's a process failure, so fix the process by reading it Thursday. Cut to a sixth or zero if it shows covenant pressure or a financing we can't see being done.

Here's where I land. Sell the immediate tranche over the next one or two sessions near 234.50 with VWAP or limit orders, using the neutral's four-session slicing if you want smoother fills. Put the staged tranche mostly at 247-250 with a small slice at 256-261, and finish the reduction the next session after any close below 228. Start the residual at a third to a half of standard, size it up or down on the footnote, and treat 207 as a disaster line and not as protection. If you're flat, don't initiate, and if you're at or above standard, reduce. The conservative's plan makes us certain to pay the gap if we're wrong on the upside, and the neutral's is close to mine. Both of us are trying not to have to be right about the debt and the print at the same time, and sizing does that without giving up the gap.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: We're down to a few real gaps. I'll start with what I'm dropping, then argue for the three things I'm keeping.

I'm done with "forced seller." Nobody is facing a margin call, and we're choosing to cut a position our own grid says pays nothing. I'm also dropping the daily z-score of minus 2.13 as a bounce argument. The conservative is right that it's measured against a 20-day mean that includes prices from when the stock was 20% higher, and the weekly reading is only minus 1.42. I accept the rest of what we've agreed:

  • A resting stop at 228 protects nothing against a gap.
  • The 256.7 headline comes off the trade sheet, and 234.50 becomes the reference.
  • The footnote can shrink the residual but never grow it.
  • Nothing gets added before the print.
  • The residual starts around a third, or a quarter if the book already holds other high-beta semis.

First, staging. The conservative says the stock closes below 228 before it reaches 247.56 about two times in three, so the staged tranche blends to about 232 and we should take the price in hand. Run his own numbers. On a driftless walk, a limit above plus a trigger below has an expected fill equal to the starting price minus the overshoot. The 2.5 points he calls a shortfall is that overshoot, about 1% on a tranche that is now a third of the reduction. That's roughly a third of a percent of the dollars sold.

That's what we pay to avoid pushing the whole reduction through two sessions that open right after a 5% gap-down, in a stock that ranges 12 points a day. As the neutral said, 234.50 isn't in hand, because it's a close that's already over. And the driftless model is one the conservative doesn't believe elsewhere, since he reads rising ADX, volume and weaker MFI as momentum. If momentum is right, sell fast, and two-thirds of the plan already does.

But look at the last two big down bars. After the 7/29 close of 257.37, an insider sold at 288 on August 3 and the CEO sold at 302 on August 14, a gain of about 17% in 2.5 weeks. After the 9/15 close of 231.50, the stock reached 281.41 on October 6, about 22% in three weeks. Two precedents aren't a law, and I gave up the washout-volume version of the argument. But when the tape supports both a momentum read and a bounce read, a third of the reduction on limits is cheap insurance against selling the low of the week.

Second, the 228 rule. The conservative wants a close below 228 to halve the residual on top of finishing the reduction. By his own two-in-three odds, that fires in most paths within days. His "I'm fine with a third" quietly becomes a sixth. The rule is also cheap to skip. From 228 to the March low at 208 is about 9%, and on a third of a standard position that's under 3% of a standard position, or under a tenth of a percent of the book if standard is 2 to 3 percent. One trigger should do one job. A close below 228 finishes the reduction, and the residual gets cut by the lookups or by a close under the 207 to 208 zone.

Third, why the residual isn't zero. I'm not claiming a third has an edge today. The zero-EV grid leans on two things nobody has checked, the possible maturity wall and four unexplained 10% days. Both resolve in days. If the footnote shows a routine refinancing and the drops line up with sector and rate moves, EV at 234.50 is probably above zero, and the person holding nothing finds that out by paying the gap. The conservative says rebuying after good news costs little because of the 269 to 286 resistance stack. But he told us a stop at 228 doesn't work because gaps skip levels, and that cuts both ways, as the neutral pointed out. I'd go further and write the rebuy rule now, so the sale doesn't become an anchor. If the print and guide clear, rebuild in stages toward standard.

On the debt, I'll concede what I can't answer and bound the rest. The neutral's bound holds. Even if the whole $4.2B were floating, 50 basis points is about 25 cents a share and a full 200 basis points is about a dollar, around 5% of forward EPS. A 47% price decline is a lot bigger than that.

On cash, TTM free cash flow of $480M is above TTM net income of $439M, and Q2's $189M was the best quarter in the table. Total liabilities fell from $6.18B to $5.95B over three quarters, which may mean debt is being paid down, though the $285M financing outflow isn't split.

The conservative's inventory point is the real one. Current assets up $229M in a quarter is what a ramp looks like, and also how a write-down starts in a downcycle, so the working-capital lines belong in the lookups. His peak-earnings argument is the best bear case at this table. A 13.4x forward multiple on a 20% operating margin is also what a cyclical looks like at the top, and I can't refute it with this data. That's why I won't go above a third. It's also why the first lookup I'd run is the neutral's: did peers fall on 7/2, 7/28, 7/29 and 9/14? If they did, a third is comfortable. If MKSI fell alone, I'd go to a quarter.

On whether the lookups are a precondition or run in parallel, the dispute is moot if they're done before the open Thursday. If they can't be, sell the first slice anyway, because no result argues for holding more than the plan.

So here's where I land. Do the footnote, earnings-date and drop-explanation lookups before the open Thursday. Sell about two-thirds of the intended reduction Thursday and Friday at VWAP. Put the last third on limits at 247 to 250, with a small piece at 256 to 261 only if you want the lottery ticket. A close below 228 finishes the reduction the next session. Size the residual at a third of standard, or a quarter if the book already holds high-beta semis. The lookups can cut it to a sixth or zero, and so can the footnote simply being unread days before the print. Add nothing before earnings. Treat 207 as a disaster line and not protection. If you're flat, don't initiate, and if you're at or above standard, reduce.

The cost is real. If the stock falls another 10% in two sessions, the limit tranche sells lower than the conservative's quick exit. If it bounces right away, we're lighter than I'd like. But this plan doesn't require us to be right on timing, the debt and the print all at once. It also keeps the one thing a beta-2 stock reliably offers, which is a tail that pays if the lookups come back clean.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: We're close enough that I'll be quick on what I'm giving up and spend my words on where the plan still has a trap in it.

On the limit tranche, the conservative's point is fair. I can't call the cost of waiting trivial using a driftless walk and then call the benefit real using two bounce precedents. Two precedents is thin evidence, so the slice should be small. A quarter of the reduction at 247 to 250, good through Monday's close and then sold at market, works for me. The neutral also caught that my opening plan carried three-quarters of standard until its limits filled, so comparing five-ninths to "the half the aggressive opened with" mixed up the residual with the interim holding. Everything on the table now is more conservative than what I started with. I'd also drop the 256 to 261 slice. If the stock sits at 234.50, standard 10-day EMA smoothing decays that average to roughly 247 by Monday's close, which is my arithmetic and not a reported number. So 247 to 250 is where the EMA lands anyway, and 256 to 261 targets a level that is moving away from us.

On the 228 rule, I'm convinced by the neutral. The only reason the trader gives for keeping anything is that 231.50 held once. If it fails on a close, that reason is gone, so the rule should follow the reason and not my cost argument. One notch, once.

Now the trap. As written, three things push the residual down: mixed lookups give a quarter, a close below 228 steps it down a notch, and an unread footnote near the print cuts it again. Read literally, mixed lookups plus a 228 close give a sixth. By the conservative's own touch-versus-close logic, the 228 close is a live possibility within days. So a third is really the number for the narrow path where the stock holds and everything comes back clean. A sixth of standard taking a 25% gap is about 4% of standard, around a tenth of a percent of the book if standard is 2 to 3 percent. That doesn't hedge being wrong about the sell, and it can't pay for the attention it takes. If the tape and the lookups are going to decide between a quarter and nothing, we should say so openly instead of hiding the choice in a ladder. My ask is simple. A quarter is the floor whenever the footnote has been read and isn't ugly, and a mixed result and a 228 close don't stack. A sixth or zero is for an ugly footnote, an unread one with earnings close, or a close in the 207 to 208 zone. That sits inside the conservative's own opening range of a quarter to a third. It's a twelfth of standard, so I won't die on it, but the stacking is invisible until it fires.

The rebuy rule needs one fix. The neutral pairs the second tranche with a close above the daily SuperTrend, quoted at 275 today. In a downtrend that line only ratchets down, so 275 is a ceiling that falls. Write the rule as a close above whatever the daily SuperTrend reads at the time, not a price. I'd also keep the first tranche tied to the first post-print session and add no further waiting period. The gap is the price of the insurance, and every extra session of waiting adds to it.

On the lookups, the trader's own plan lists the Q3 guide among the pending items. Companies like this usually give next-quarter guidance with the prior quarter's results, so I'd expect it's already public. That's an expectation, not something in our data. Paired with the neutral's estimate-revision check, it tests the conservative's peak-earnings case directly. If consensus held near $17.55 and the guide stepped up sequentially, the bear case needs the cycle to turn inside a quarter, which is a specific claim. If estimates were cut, I side with the conservative and go to a quarter. I'd also date the Q2 release against the four big drops. The CEO's sale at 302 on August 14, after the 257 close on July 29, hints the print landed in that window. If so, three of the four 10% days came before the best quarter in the data was reported. That's inference from sale timing, so confirm it.

I agree on the rest. The trade sheet shows 234.50 as the reference plus the blended fill, the final 10/14 bar gets re-pulled before anchoring on 234.50 and 228, nothing gets added before earnings, and 207 is a disaster line, not protection.

The cost is real. If the stock slides another 10% before Monday, the limit slice sells lower than a quick exit would have. If it bounces hard, we're lighter than I'd like. Neither case requires us to be right on timing, the debt and the print at once.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: We agree on the trade, and the remaining gaps are about a twelfth of a standard position, so I'll keep the concessions short and spend my time on the soft spots in the final plan.

The neutral is right that I leaned too hard on "the bar is lower after a 47% drop," and I'm dropping it. If the print landed in early August, the market had the best quarter in the file and still gave back about 22% from the CEO's sale at 302. That could be good news fading or it could be multiple compression, and our data can't separate them. The conservative is right that rate-driven drops aren't comfort, because the driver is still moving. That's why the residual stays capped at a third even on a clean read.

I'm also taking several of his mechanics: - The Friday-close deadline on the footnote. - The limit slice fixed at a sixth of standard, good through Monday's close. - Any extra reduction from the lookups going into the VWAP leg, not stretching the limit window. - His rebuy rule: nothing before the print, then up to half of standard in total after the first post-print session if the footnote passed and the guide wasn't cut. The rest goes in only on a close above wherever the daily SuperTrend reads at the time, never above standard.

On the ladder, I'll take the neutral's failure count over the conservative's clean/mixed/ugly. His definition of clean requires the four drops to line up with sector or rate news, and a headline search often can't explain a 10% day. That would make a quarter the real default while we call it a third. Not finding an explanation isn't a failure, and I agree a sixth isn't worth carrying, so the rungs are a third, a quarter and zero.

I'd tighten the peer test before anyone runs it. This stock has a beta of 2.03, so a 10% day can be a 5% day for the group and still be perfectly ordinary. "MKSI fell alone" should mean peers were roughly flat or up while MKSI dropped about 10%. Also, the four drops are really three episodes, since 7/28 and 7/29 were back to back for about 18% over two sessions. I'd weight 9/14 most. It's the one most likely to have come after the Q2 numbers were public, so a pending print can't explain it, and it's where an 8-K or a financing headline would show up if one exists.

On stacking, the 228 close is the failure most likely to fire, and it could fire within days. So the real question is whether either other test fails. If estimates were cut, or MKSI fell alone with the level already broken, we're at zero. I'm fine with that, because the stake would then have lost both reasons the trader gave for keeping it.

I don't accept the conservative's argument that on the one-failure path it's cheap to go to zero. It's equally cheap to hold. A quarter taking a 25% gap costs about 0.12 to 0.19% of the book, and in the good case it earns a similar amount. One failure means the other tests passed, and nobody has shown that the expectation on that path is negative. The exposure also isn't a stub. At a beta of 2.03, a quarter of standard carries about half the market sensitivity of a full position in an average stock, and a third carries about two-thirds. The rate regime is also thin and fragile. The 86% hike odds come from a $0.76M market, so one soft inflation print could move it. That second point is my inference, not a data point, but it's why I want some beta left.

The point I think we've all been glossing over is that none of these lookups was something our tools could do. The fundamentals report says the tools don't show the debt cause. The news feed was headlines only, and the earnings date wasn't retrieved. "A lookup, not an event" is true for someone with EDGAR and a consensus screen, which is the caller. So the plan has to say what happens if they don't run them. The default is to sell on schedule and hold the third in the interim, since unresolved isn't a failure. If the footnote is unread by Friday's close, hold a quarter at most. If it's still unread at the print, the residual goes to zero. Last year's big gap was November 6, which hints the print is about three weeks out, but that needs confirming. Either way there's time to read one footnote.

Here's the plan as I'd hand it over. 1. Before Thursday's open, run the footnote, the earnings date, the estimate and guide revisions, and the peer comparison on the three drop episodes. 2. Sell about three-quarters of the reduction Thursday and Friday at VWAP, and put the last quarter on limits at 247 to 250 through Monday's close. Sell whatever is unfilled then, or the next session after any close below 228. 3. Size the residual at a third on zero failures, a quarter on one, and zero on two. An ugly footnote or an unread one at the print is zero on its own. 4. Add nothing before earnings, and treat a close in the 207 to 208 zone as zero. 5. Replace the 256.7 on the trade sheet with 234.50 as the reference plus the blended fill, and re-pull the final 10/14 bar before anchoring on it. 6. If the caller is flat, don't initiate. If they're at or above standard, reduce, and cap it at a third of standard until the checks are done.

The cost is real. If the stock slides another 10% before Monday, the limit slice sells lower than a quick exit would have. If it bounces hard, we're lighter than I'd like. Neither case requires us to be right on timing, the debt and the print at once.

FINAL TRANSACTION PROPOSAL: SELL