Altair
Skip to content

Trading Analysis Report: SMCI

Analysis date: 2026-10-21 · Model: claude-opus-5-5 · AI-generated, not financial advice.

  • Analysis date: 2026-10-21
  • Rating: Underweight
  • Generated: 2026-10-21 11:23:14
  • 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 SMCI. Sell the larger of one-third of the position or everything above 2% of portfolio weight now: today if the session is still open, otherwise through Thursday 10/22's first hour, rather than waiting on resting orders at $37-39.5. If flat, stay flat, and do not short. After reading the 10-K and raise 8-Ks tonight, hold the smaller of two-thirds weight or 2% if the filing is clean, cut to the smaller of half weight or 1.5% by the Friday 10/23 close if it is bad or unresolved, and exit the residual that day if it is ugly; a daily close below the $32.52 200 SMA exits the residual, and re-adds come only after the September-quarter report and an on-time 10-Q.

Current Price: 35.75

Price Target: 31.7

Confidence: Medium

Time Horizon: 1-2 months

IV. Risk Management Team Decision

Aggressive Analyst

Aggressive Analyst: Since nobody has spoken yet, I'll open, and I'll answer in advance the objections I expect from both of you.

First, selling here is the aggressive call, not the timid one. Everything on the surface says buy: a pre-announced revenue beat, an 11 to 12 percent pop, Dell and HPE jumping, the Nasdaq up over 1 percent, and about 88 percent of labeled StockTwits posts bullish (15 of 17). Selling into that is the contrarian move, and the tape isn't backing the crowd. The stock gapped from 32.47 to a 36.42 open, sold off all day, and closed at 35.75 in the bottom fifth of the range. Volume was 58 million shares, the heaviest since mid-September. A weak close on heavy volume after good news looks like distribution.

All three SuperTrend tiers are still down. Price is under the 10 EMA and the 50 SMA, and the MACD histogram is still -1.11. The whole bounce has recovered less than a third of the drop from 44.94 to 31.63. That drop happened before the beat, and nothing in the headlines explains it.

To the neutral analyst, who will probably say hold because it's 6.4 times forward earnings: that multiple rests on two things nobody has verified. One is the June quarter. The 17.5 percent gross margin was backed out of TTM figures, not reported, and every reported FY26 quarter was between 6 and 10 percent. The other is the share count. Market cap implies about 647 million shares, while equity divided by book value implies about 928 million. That's a 43 percent spread in the denominator. Equity also jumped roughly $6.9B in one quarter, with about $5.6B unexplained by earnings, which smells like an issuance.

If you spread the same forward earnings dollars over 928 million shares, forward EPS drops from 5.59 to about 3.90 and the multiple goes from 6.4 to about 9. That is still not expensive, but it is no longer a screaming bargain for a beta-2 hardware company. The plan's own math says that if the count is above roughly 683 million, the "holds" outcome earns close to nothing.

And the earnings aren't cash. Operating cash flow was -$7.6B over nine months, TTM free cash flow is -$8.25B, and the burn was funded externally. The CEO and director also sold 200,000 shares on September 4 at $36.60 to $40, and there's no insider buying anywhere in the data. BofA's concern is margin, and the pre-announcement was a revenue beat. Revenue was never the problem.

The "wait for confirmation" instinct also fails on this name. SMCI has had five single-day moves of 19 to 33 percent in seven months, and ATR is 7.4 percent of price. On March 20 it fell 33 percent in a day. A stop is a wish with a gap like that, so size is the only protection that works.

To the conservative analyst, who will want a full exit or a short: I agree with the diagnosis and not the prescription. A full exit walks away from a company that grew revenue 123 percent year over year and net income about 4.4x, and the Q4 growth looks real. If the 10-K shows 683 million shares or fewer and June margins hold, this goes back to hold, and a stock this violent could gap up 20 percent without warning. Retail is already shouting squeeze. Shorting a name with this gap history is where the loss is unbounded, so I won't do it.

Instead, trim about half and cap the residual near 2 percent of the portfolio. A 50 percent gap against us then costs about 1 percent, and we still participate if the filing comes back clean.

Execution matters too. Rest sell orders at $37 to $39.50, just under the cluster of the 50 SMA at 38.79, the daily SuperTrend at 39.43 and the Bollinger midline at 39.62. That means selling into strength that has already shown it fades. If there's no rally in about five sessions, finish at market. If the stock closes below 32.52 and fills the $33.17 to $35.27 gap, finish immediately. Exit whatever is left on a close below about $30.50, just under the $30.92 low.

On reward and risk, the cost of being wrong on the trim is giving up part of a run toward $39. The plan's fail case is $16 to $25, which is 30 to 55 percent below here. That fail case is a judgment, not something my data confirms, and I can't verify the 68 percent confidence figure either. But the asymmetry is clear even if you haircut it.

I'm also upfront about what I don't know: the June P&L, the share count, the raise terms and your actual holdings. The sizing applies to whatever you really hold, and if you're flat, stay flat. The catalysts are close: the full fiscal Q1 report, which I believe is early November but you should verify, the Iran ceasefire deadline on October 31, a December Fed hike that prediction markets price at 78 percent, and Nvidia around November 18. I'd rather be the one deciding what to hold going into that than the one forced to react.

So the bold play is to sell into the pop, size for the fat tails, and keep a small residual as cheap optionality. Pull the 10-K now, and don't re-add until the share count and June gross profit are confirmed.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Neutral, you're right about the tape, and I oversold it. 58 million shares is the heaviest print since mid-September only because the sell-off days ran 25 to 44 million, and the real gap days this year ran 127 to 243 million. You're also right that 35.75 is basically the October 14 close of 35.78, so this isn't a euphoric high. I read that fact differently, though. A pre-announced beat, Dell and HPE up, the Nasdaq up over 1%, and the stock only gets back to where it was a week ago, then fades from 37.45. That doesn't prove distribution. It does say the good news bought a retracement and nothing more. I'm treating the tape as a tiebreaker, and the 15-of-17 StockTwits figure as color.

You're both right about my execution. Resting orders at $37 to $39.50 mean I carry full risk while waiting for a 4 to 10 percent rally in a name with a 7.4 percent ATR, and that resistance cluster has never been tested. So flip the order: sell a first tranche tomorrow at the open or in the first hour, then work the rest. The stop was never meant to define the worst case. It covers ordinary drift, and the cap covers the gap. The difference between conservative's 1 percent residual and my 2 percent is a few tenths of a point either way, so I'll take neutral's 1.5 and move on.

Neither of you took the share count far enough. Neutral, you bracketed it at 650 to 850 million. I agree 928 is a ceiling, but your floor is too low. Divide net income by diluted EPS for the reported quarters and you get roughly 646 million in September, 668 in December and 671 in March. The company was already near 670 million before June, so the 647 million implied by market cap is the stale number. The back-solved June quarter, about $1.15B over $1.68, gives roughly 683 million, which is the plan's threshold. That's a weighted average, and a raise late in the quarter barely moves it. Start at 670 million and add 125 to 190 million shares for your own $5.6B at $30 to $45, and you land around 800 to 860 million. Forward P/E there is eight to nine. With your numerator haircut to about $2B of earnings, it's mid-teens. So I agree 35.75 is priced for neither outcome, but 6.4x is the best case, not the base case. That's the argument for half weight in a beta-2 name, and you made it better than I did.

On cash, conservative, your instinct is right but your re-entry bar is too strict. From the balance sheet alone, non-cash current assets were $20.28B in March and $20.21B in June, essentially flat. Cash rose $6.2B while equity rose $6.9B and liabilities fell $0.4B. That's a rough inference, but it says June cash came mostly from the raise and operations were near break-even. The burn probably stopped, yet $20B of receivables and inventory hasn't turned into cash. I'd require evidence those are turning, plus the raise terms, rather than a positive operating cash flow quarter. The Q3 outflow of $6.6B lines up with the $7.3B payables unwind, so neutral is right that part of it is timing. And neutral is right that the missing June P&L isn't a red flag, since neither fiscal Q4 shows up. It costs one lookup, so I wouldn't size on it.

Where I still disagree with conservative is on selling down to 1 percent before October 31 regardless of what the filing says. The 10-K and the raise-related 8-Ks are a day away and answer the numerator, the denominator and the funding question together. Committing to a blind exit when the answer is a lookup away is paying twice. Sell the insurance tranche now and let the filing decide the rest.

Where I disagree with neutral is the clean branch. If the filing shows shares near 700 million and June margins genuinely mid-teens, parking at two-thirds weight out of habit is the timid move. That's when I'd rebuy toward full weight, even at a higher price. You said recapturing means buying the gap. Yes, and I'd pay it. On an illustrative 4 percent standard weight, paying 15 percent more to rebuy the third I sold costs about 0.2 points of portfolio. A March 20-style 33 percent gap on an unverified full position costs over a full point. Paying a bounded premium for information is what the trim is for.

On the fail case, neither of you should take $16 to $25 from me on faith. It's a judgment. But the stock closed at 20.53 in March and 23.83 in July, and the 52-week low is 19.48. Even if the realistic fail zone is $22 to $28, that's 20 to 40 percent below here, against a bounded cost of being wrong on the trim. I won't recommend puts or collars, because there's no options data in front of us.

So here's where I land. Sell roughly a third tomorrow. Read the 10-K and the 8-Ks. If the count is near 700 million and June margin is real, buy back toward full weight on confirmation. If it's near 800 million or the margin looks like 10 to 12 percent, finish down to the smaller of half weight or about 1.5 percent without waiting for $37. In the murky middle, leave resting orders at $37 to $39.50 with a hard time stop at the October 28 close, ahead of the October 31 ceasefire deadline (82 percent to hold), the full fiscal Q1 report (early November, verify), a December hike priced at 78 percent, and Nvidia around November 18. Keep the close-below-32.52 and gap-fill rules as backstops on the residual. No short, since the gap history makes that loss unbounded.

All of this is in fractions of whatever is actually held. If you're flat, nothing here makes me want to initiate at 35.75.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Let me start with what I'm giving up, because you both earned it.

Neutral, October 31 is a Saturday, so the last tradable moment before the ceasefire deadline is the Friday close. I'll keep your Wednesday October 28 close as the time stop on anything still above target.

Conservative, you're right that my 0.2-point arithmetic cuts both ways. If unwinding the trim is that cheap, waiting for the print to unwind it is just as cheap, so I can't use cheap reversibility to justify rebuying toward full weight on a four-month-old quarter. I withdraw that.

Neutral, you're also right about the five giant days. They compound to about 15 percent below flat, but the stock is up 16 percent from the March 19 close. The gap history is an argument about how much to hold, not which way to lean, so I'll stop using it for direction.

Here's where I think we're still building a decision tree around a lookup. A June fiscal year-end 10-K is due about two months after the quarter closes. If the vendor has a June 30 balance sheet, the filing has almost certainly been public since late summer. If it's late again, that's the bad branch on its own. The share count is on the cover page, June gross profit is on the income statement, and the equity jump is in the statement of stockholders' equity. That's a short job, and it can be done tonight, before the open. The murky bucket should be tiny. If something is still unresolved at Thursday's close, treat it as bad.

For the bad branch, I'd finish Thursday once it's clear, with Friday's close as the outer limit. Each day of delay is roughly one ATR of variance with no expected gain. A stock trading 25 to 60 million shares a day doesn't have a spread problem worth waiting out. Resting orders at 37 to 39.5 are a bonus, not the plan.

The uncomfortable corollary is that if that filing has been public for weeks, the market already had the share count and June margin while the stock ran from 35.64 on September 15 to 44.94 on October 7 and then fell 30 percent. We don't hold an informational edge. We're missing pages the market has read. That's the honest reason to trim. Earnings could be about $2 billion on neutral's haircut or about $4.5 billion if June annualizes, and a beta-2 name shouldn't be held at standard weight across a range that wide.

On the clean branch, I'm with neutral over conservative, and not out of habit. Say June is real: $1.14 billion of earnings on roughly 683 million shares. Annualized, that's about $6.70 a share, above the 5.59 forward figure, so today's price is around 5.3 times run-rate earnings. Sitting at half weight on that leaves too much on the table. I wouldn't go to full weight, because one quarter of margin nearly doubling is exactly what the November report has to prove. So hold two-thirds and bring the last third back on either of two triggers, whichever comes first. One is a daily close above about 39.6, which clears the SuperTrend stop, the 50 SMA and the Bollinger midline together. The other is a print showing margin holding and receivables or inventory turning.

I prefer that to conservative's bar of margin and cash together. I'm not sure the earnings release carries a full cash flow statement. If it only arrives with the 10-Q, we'd be parked at half through Nvidia on the 18th.

On cash, I'll correct myself and push back on conservative's neat number. The TTM route says June free cash flow was around minus half a billion. The balance sheet says roughly zero to modestly positive: cash up $6.2 billion, equity up $6.9 billion, liabilities down $0.4 billion, and non-cash current assets flat. That assumes the raise was cash and net of fees. The two routes bracket zero. The burn probably stopped, but about $1.1 billion of reported profit produced roughly no cash, and $20 billion of receivables and inventory hasn't turned. That's a quality-of-earnings question. It supports the haircut and the receivables test, not a requirement for a positive operating cash flow quarter.

Neutral, your working-capital point is better than I expected. Working capital excluding cash went from about 0.96 times quarterly revenue to about 1.2 times. So a bigger September revenue number can mean a bigger funding need, and "revenue was never the problem" overstated it. That strengthens the trim.

So here's the plan. Read the filing tonight and sell about a third on Thursday, worked through the first hour. If it's clean, meaning shares near 700 million or below, mid-teens margin and an explained raise, stop selling, hold about two-thirds, and re-add on the triggers above. If it's bad or unresolved by Thursday's close, finish down to the smaller of half weight or about 1.5 percent of the portfolio, by Friday at the latest. The close below 32.52 and the gap-fill rules stay as backstops on the residual, and the 30.5 close is an ordinary-drift exit, not the worst-case definition. No short, because a stock with 19 to 25 percent up-gaps makes that loss unbounded. Everything applies as fractions of what's actually held, and if you're flat, nothing here makes me want to initiate at 35.75.

Being bold here doesn't mean holding more. It means acting tonight and tomorrow while others debate branches, paying a higher price to rebuy if the filing earns it, and refusing to take the one trade, the short, where the tail is unlimited.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: We're down to arguing over roughly half a point of portfolio weight, so I'll skip the long version and take only what still matters.

First, what I'm giving up. Neutral, you're right that I can't call the June quarter unproven and then annualize it as my base case. I wanted it both ways, and I'll stop. Conservative and Neutral, you're both right about the price trigger. A close above 39.6 is a counter-trend flip with the weekly and monthly SuperTrends still down at 44.40 and 69.53. The likeliest way it happens is a run into the November print, which would have us buying back a third at the top just before the test. I'm dropping it. Re-adds come after the report, on evidence, in two pieces. I'll also take the 2 percent cap in the clean branch, and Neutral's point that a pre-announced quarter probably puts the full report well ahead of Nvidia. My worry about sitting through November 18 was thin.

I also need to correct my own opening, because it changes how much weight the multiple deserves. I took the 5.59 forward EPS, multiplied by 647 million shares to get dollars, and re-divided by 928 million to get 3.90 and a 9x multiple. That only works if the 5.59 was built on the stale count. Forward EPS is most likely an analyst consensus per-share figure, and analysts have read the filing. If they've updated for the raise, the dilution is already in the 5.59 and the 6.4x doesn't move with the share count at all. June annualized is $4.56 billion, and dividing that by 5.59 gives about 816 million shares, right in the 800 to 860 range we've been calling post-raise. So one reading is that consensus assumes June-level earnings on a diluted count. The other is that it assumes $3.6 to $3.8 billion on the old count. I can't tell which from here. Either way, the share count may matter less for the multiple than we've said, and the margin and cash conversion matter more. My sell never rested on the multiple. It rests on a beta of 2, an ATR of 7.4 percent, a dense calendar and a quarter nobody has seen. But it's another reason the filing is worth reading tonight.

Conservative, you say you won't rebuild while the drop is unexplained. Neutral said that cuts both ways, and I'd add that it proves too much. The November report may not explain a seven-session, 30 percent fall either, and if "explained" is the bar, we never re-add. Your balance-sheet bar has the same flaw. If receivables and inventory must fall in absolute terms from $20 billion, a company that just beat on revenue will almost certainly fail, because a bigger quarter ties up more working capital. Neutral's ratio of non-cash current assets to quarterly revenue has held between 1.7 and 2.0 times. That test can tell a scaling business from one that's building stock it can't collect on. A bar a growing company can't clear is a decision not to re-add, not a condition.

You are right about what the 5x multiple is saying. The market is asking whether earnings turn into cash, and the ratio test answers that same question, so we're closer than the wording suggests.

On rates, I'm with Neutral, and the debt figure is shakier than either of you said. The $9.3 billion is the profile's 64 percent debt-to-equity times June equity, and this profile already mixes dates. On March equity of $7.6 billion, the same ratio gives about $4.8 billion. A hundred basis points is $50 to $90 million a year, a few percent of operating income even at the March run rate. A hawkish Fed compresses multiples, but this one is already near six times. The funding risk is dilution and terms, which is why the raise terms are on the read list.

One addition to tonight's list. The pre-announcement 8-K came out today and is the document that caused the gap. I haven't seen it, but I'd read it for what it leaves out. BofA's margin headline came from somewhere. If it states revenue and says nothing on margin or EPS, that silence tells us where the uncertainty sits.

Neither of you should mistake the first third for a bet on direction. On an illustrative 4 percent weight, it costs about a third of a point if the stock retakes the October 7 high of 44.94, and it saves about 0.45 if March 20 repeats. It buys variance reduction, and it's cheap.

Here's the plan. Tonight, read the 10-K cover page, the June income statement, the equity statement, the raise-related 8-Ks and today's pre-announcement. Thursday, sell about a third, worked through the first hour. If analyst upgrades lift the stock, treat that as a better price, not as news about margin. If the filing is clean, meaning about 700 million shares or fewer, mid-teens June margin and an explained raise, stop selling. Hold the smaller of two-thirds weight or 2 percent and add nothing. If it's bad, thinly explained and unresolved at Thursday's close, or not filed, finish down to the smaller of half weight or 1.5 percent by Friday the 23rd at the latest, without waiting for $37. October 31 is a Saturday, so that leaves a week of cushion before the last tradable session. A close below 32.52 with the gap filled means finishing the residual immediately. A 30.5 close is an ordinary-drift exit, not a worst-case definition. Re-add only after the full report, in two pieces, with margin holding, the ratio flat or falling, and the share count and raise terms known. No short, and no puts we can't price. Everything applies to whatever is actually held, and if you're flat, nothing here makes me want to initiate at 35.75.

Being aggressive here doesn't mean holding more. It means moving first, reading the actual documents tonight, paying up to re-enter if the evidence earns it, and refusing the one trade, the short, where the loss is unbounded.

FINAL TRANSACTION PROPOSAL: SELL Aggressive Analyst: Neutral, your "sell the larger of a third or everything above 2 percent" is the best refinement anyone has offered, and it follows from the caps we all accepted. If every branch ends at or below 2 percent, any weight above that gets sold either way, so waiting for the filing to decide it just buys overnight gap risk with nothing in return. I'm taking it. If the session is still open, work it today. If not, work it through Thursday's first hour. And if the analyst upgrades retail is predicting show up Thursday, sell into them. That's a better price on a sale we've already decided to make.

On the gates, I'm with you against Conservative, but with a distinction. Any gate built on a derived number should be a band, not a line. A cash-flow bar at minus half a billion sits on top of our central estimate, so minus 0.45 passes and minus 0.6 fails, and the sizing flips on noise. Make it a flag, and trip the bad branch on something materially worse, around minus a billion, or on a worse number nobody can explain. The same applies to the share count. Don't treat 700 million as a cliff. Treat roughly 720 million or fewer as clean and roughly 780 million or more as bad, and in between ask whether the raise was explained. Compare the cover-page count with the June-quarter weighted diluted count too. A big gap between them tells you the raise was common stock and landed late in the quarter.

Facts are different from estimates, though. Whether the filing exists, whether the audit opinion is unqualified, and whether there's restatement or non-reliance language are yes/no questions, and I'd keep those as hard gates. A plan whose clean branch can't realistically be reached is a full trim in disguise. If the company is fine, the plan has to let us hold two-thirds.

That leads to where I disagree with Neutral. You said the whole filing tree is worth about half a point of weight, and for clean versus bad that's right. But it hides a branch nobody has named, which is ugly. Suppose the filing shows an auditor resignation, a restatement, a qualified opinion, a going-concern paragraph, or a raise with toxic terms like a variable conversion price or a ratchet. Or suppose it simply isn't filed. A June year-end 10-K is due about two months after the quarter closes, so no filing now means a company more than seven weeks late, with the auditor history the world-affairs report mentioned. That isn't a bad quarter. It's a different distribution. The reason a 1.5 percent residual is survivable is that a March 20-style gap costs about half a point and we can still trade the stock. In an accounting event the stock can halt, and you can't sell a halted stock. So the filing read earns its keep on the left tail, not on the half-point between branches. If any of those triggers hit, exit the residual that day, not down to 1.5 percent.

I'll also take your point on boldness. By my own arithmetic the first third costs about a third of a point if the stock retakes 44.94 and saves about 0.45 if March 20 repeats, so it's a variance trim, not a view. The boldness lies elsewhere. It's in acting before the filing instead of debating branches, and in not hiding at half weight if the filing is clean. It's also in paying up to re-add after the print if the evidence earns it, in two pieces, with the second right after the 10-Q. The 10-Q usually lands within days of the release, though that's from memory and needs a date check. So the two-piece structure costs little time, and we aren't sitting out the move for weeks.

Conservative, I owe you one thing. Your re-add test of splitting receivables from inventory is right, because a flat ratio can hide inventory piling up behind collections. The 10-Q is where that split lives, which is exactly why the second piece waits for it. But I'd hold to the ten percent tolerance on the ratio. It has run 1.7, 1.7, 2.0 and 1.8 with no story behind the swings, and a zero-tolerance test on that series is a coin flip about shipment timing.

So here's the plan. Tonight, read the cover page, the weighted diluted count, June revenue, gross profit and operating cash flow by subtracting nine-month figures from full-year, the equity statement, the raise-related 8-Ks, and today's pre-announcement for what it leaves out. Sell the larger of a third or everything above 2 percent of the portfolio, today if the session is open, otherwise Thursday. If the filing is clean, meaning the share count is in the clean band, June margin is genuinely mid-teens, the raise is explained, the audit is unqualified and cash flow isn't materially worse than derived, stop selling. Hold the smaller of two-thirds weight or 2 percent and add nothing. If it's bad, thin, or unresolved at Thursday's close, finish down to the smaller of half weight or 1.5 percent by Friday's close at the latest. If it's ugly, exit the residual. A close below 32.52 with the gap filled means finishing the residual at once, and a close below 30.5 is an ordinary-drift exit, not the worst case. Re-add only after the full report, in two pieces, with no price-only trigger. No short, and no puts we can't price. October 31 is a Saturday, so Friday the 30th is the last tradable session, and all of this lands well before it. Everything applies to whatever is actually held, and if you're flat, nothing here makes me want to initiate at 35.75.

FINAL TRANSACTION PROPOSAL: SELL