Altair
Skip to content

Trading Analysis Report: NBIS

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 14:59:36
  • 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 NBIS: open no new positions, use no leverage, and cut existing holdings to a ceiling equal to the smaller of 0.5x a standard allocation or the size that matches a standard position's daily dollar risk at an 8.7% ATR. Make the cut in one step at the first real fill, which for most accounts is Thursday 10/15's open. Only holders who have already volatility-adjusted their standard size should stage it: sell to about 0.75x at that fill, rest a limit at 198-199, and sell anything left at Thursday's close, without a stop. From the ceiling, move to 0.25x on a weekly close below 172.93, and go flat on a daily close below 169.23 or if any financing, growth or filing trigger fires. The technical price objective is 151.50 over 1-3 months.

Current Price: 186.01

Price Target: 151.5

Confidence: Medium

Time Horizon: 1-3 months

IV. Risk Management Team Decision

Aggressive Analyst

Aggressive Analyst: I'll open, since neither of you has spoken yet, and I'll answer the arguments I expect from both of you.

The Sell is the bold call here, and I want that to be clear. Cutting exposure on a stock that went from 73.87 to 286.69 this year takes more nerve than holding. The lower highs at 286.69, 277.68 and 249.87 show distribution, and today the stock broke the September-October floor at 199.54 on about 2.1 times average volume. The daily SuperTrend is down, and the MACD histogram has deepened every session since October 8, now at -5.31. The tape is saying the market is repricing the whole neocloud model, and NBIS is a high-beta way to own that repricing.

The fundamentals give the repricing a reason. Q2 free cash flow was -$3.4B against $9.1B of cash, so runway is roughly 2.7 quarters at that burn. Interest expense went from $4.8M a year ago to $119.1M, and normalized EBITDA covers it only about 1.3 times. About $2B of the $2.25B in Q2 operating cash flow was working-capital inflow, likely prepayments, and that can reverse. The share count is up about 36% in three quarters. At around $47B of market cap, the stock trades near 20 times annualized Q2 sales while forward EPS is -$3.56. CoreWeave's bonds touching a 15% yield is the market telling us what funding costs for this group, and NBIS has to keep raising capital in that market. Polymarket has another Fed hike at 86%, though on thin volume, and no cuts at 97%. Insiders sold about $317M with no open-market buying, including $117.8M by the COO on October 5 at $231 to $243. Some of that selling looks scheduled, but the COO block does not.

Now the likely pushback. The conservative case will be that if the thesis is this fragile, you should sell everything at 186 and be done. I'd call that the lazy version of caution. Selling the entire position into a stock that is 6.3% below its lower Bollinger band, with a daily z-score of -2.84 and a bar that may be provisional, makes you the forced seller at the low. The data in the technical report shows how violent the snapbacks are. After the August 10 close at 184.11, the next two sessions closed at 193.23 and 259.20. That is one instance, not a rule, but it is exactly why you sell one-third now and sell the rest into 195 to 217. You get paid for the reflex bounce instead of donating it. The plan also has a hard clock, Friday's weekly close, so it doesn't depend on hope.

The neutral case will be that the weekly and monthly SuperTrends are still up, price is above the 200 SMA, Rosenblatt has a $304 target, and the daily oversold reading argues for patience. My answer is that the oversold reading lives only on the daily chart. The weekly z-score is -1.26 and the monthly is +0.86, the daily TD count is only 5 of 9, RSI at 35 isn't even below 30, and MFI is around 29. That is a stock that hasn't capitulated yet. Even retail on StockTwits is saying "capitulation coming then the big buy," and a crowd waiting for capitulation is evidence it hasn't happened. The weekly stop at 172.93 is less than one ATR away, and an ATR of about 16 means a gap can skip your line entirely. Waiting for confirmation there means acting after the damage. Cutting to about 0.5 times standard now is how you survive a gap and still keep real exposure.

Keeping half is what makes this the high-reward version. If the AI credit scare fades, you still own a name with revenue up 454% year on year. If it doesn't, the freed capital sits ready for the 149 to 154 zone, where the range-height projection and the July lows converge. That is optionality, not a recommendation to buy now. The plan says no new positions and no leverage, and I agree, because 8.7% daily ATR makes sizing the real risk control.

I'll also flag the limits of my own case. The 186 close may be provisional, the fundamental price is derived, the balance sheet is stale, and I can't reproduce the revenue-per-capex-dollar figure from the tool output. On Q2 alone, revenue was about 10 cents per capex dollar, so I'm leaning on the figures I can verify. The trigger list protects against being wrong: a double-digit-coupon raise, a Q3 sequential revenue miss below 25% with capex still rising, or ugly filing disclosures all move this to a full exit. Until then, sell the bounce, keep a runner, and let the cash buy us the option. Aggressive Analyst: We all agree on Sell, so the argument is about how much of the cut happens tonight. Measured against a standard position, the trader sells about 0.17 now and leaves you at 0.83x. The neutral wants 0.25 sold, which is 0.08 of a position different. The conservative wants 0.5 sold, which is a third of a position different. Most of this debate is really between the trader and the conservative.

First, what I'll concede, because some of it is fair.

I was sloppy saying that cutting to 0.5x "now" is how you survive a gap. The plan reaches 0.5x by Friday, not tonight, and tonight's holder is at 0.83x. The neutral is also right that 195 to 217 is too wide to be a plan. I'll take the tighter version: the second tranche goes at 198 to 199, where today's high and the lower band sit on top of each other, and the last tranche goes at 207 if we get it. I'll also take the neutral's shelf trigger. A daily close below about 180.6 means the shelf failed, and the rest of the cut gets done that day instead of waiting for Friday. That makes the plan better. On fundamentals, the conservative is right that 2.7 quarters is the generous runway and 1.7 is the stress case, since $2.0B of Q2 operating cash flow was working capital. But that only reinforces the Sell, and the real unknown is the terms of the next raise, which the triggers already cover. I'll also drop the COO sale as load-bearing. We can't tie the June-plan headline to a specific sale, and the Sell doesn't need it.

Now the pushback on the conservative. The asymmetry you laid out isn't as lopsided as it looks. The first real resistance, 198.60, is 6.8% above the close. The weekly stop, 172.93, is 7.0% below it. With an 8.7% ATR, both are a day's work. The extra third of a position you'd sell tonight costs you roughly 2% to 4% of a position if the bounce comes, and saves you about 3% if Friday closes under 173. That's close to a coin flip, and the evidence doesn't hand you a lopsided edge to the downside. ADX at 16 is below the 20 line where the tool says trend-following signals are unreliable, while a daily z-score of -2.84 and a close 6.3% under the lower band are real mean-reversion setups. I'm not claiming the bounce is promised. I'm claiming nobody can tell which side comes first, and staging is the sensible response to that.

On the Friday clock, you said it makes us the forced seller. But Friday's close is the weekly bar, the one time the primary-trend question gets answered. At that point the plan has us at 0.5x regardless, and if that bar closes under 172.93, we cut again from half size, not from 0.83. The date is tied to the information event, not to hope. Your August 14 to September 1 drop of 28% proves this tape is violent both ways, and that supports spreading the sale across the distribution instead of placing it at one price.

I most strongly reject 0.25x on a bounce. Nothing in the data has broken the thesis at that point. The weekly and monthly SuperTrends are up, price is above the 200 SMA, and revenue is growing 454% year on year. The selloff is mostly contagion from CoreWeave's credit scare, not bad NBIS news. If you cut to a quarter position on price alone, you've turned an underweight into a near-exit with no trigger firing. The triggers exist for that job. On the 149 to 154 zone, I never said buy it. I called it optionality, and I'll accept your condition: any re-entry needs clearer financing first.

Now the neutral. Your October 23 backstop leaves the book at 0.75x through the October 16 weekly close. If that bar prints under 172.93, your own plan then sells from 0.75x into the break, which is the worst version of the scenario you're worried about. Your shelf trigger also has the flaw you pinned on the conservative: it's close-based, so a gap from 186 to 170, which is smaller than today's 10% drop, fires it after the damage. As for the TD 9, a daily 9 on the 20th needs that day to close below 186.01, so waiting for it means holding more through weakness. The report also calls a daily 9 low quality with weekly and monthly counts at 2 and 1.

What I'd keep is the part both of you agree with: sell now, take the bounce when it comes, finish by Friday or on a shelf failure, hold a runner under 0.5x, no adds, and go to a full exit on a bad financing print, a Q3 growth miss, or ugly filings. The sizing point is right too. If the standard allocation wasn't built for an 8.7% ATR, the caller should set a lower ceiling than 0.5x of it. Aggressive Analyst: Let me put the disagreement in proportion first. All three of us want Sell, a 0.5x ceiling, no adds, and the same financing and growth triggers for a full exit. What's left is about a quarter of a position over two sessions, Thursday and Friday.

Neutral, I'll take most of your plan. You're right that Friday is a deadline, not an information event. The weekly bar closes Friday afternoon, so the earliest anyone acts on it is Monday's open, and I overstated it when I called Friday the moment the trend question gets answered. The date's job is to get us to 0.5x before the weekend. So I'll accept your version. Sell a quarter once the official close prints, since the bar may be provisional. Finish on a trade through 198 to 199 or a live break of the 180.6 shelf, and be done before Friday's close either way. I'll also drop my 207 tranche. It's 11% away and the sliver it protects isn't worth the complexity. A quarter tonight instead of a sixth is a judgment call on a small difference, and the open financing calendar and the TSMC print flagged as on deck (I can't see a date) tip it for me.

Conservative, you landed a few hits. On the z-score, here is the single reading I'll hold. Daily is -2.84, weekly is -1.26, monthly is +0.86, and RSI is 35. That is neither a washout nor a clean top, so both tails are live. It doesn't argue against the Sell. It argues for selling in two steps. You're also right that ADX rising from 10.7 to 16 is trend strength building, so I'm dropping the below-20 argument. And I can't call this contagion and also say CoreWeave's 15% bonds tell us what NBIS's next raise costs. Contagion is the pricing mechanism, and that is the Sell thesis. But over two days it cuts both ways. The same channel reverses if CoreWeave's credit stabilizes, which the news report names as the likely lead indicator.

I don't accept the tail argument. You say supply caps the bounce while the break is open. The only snapback in this tape went 184.11, 193.23, 259.20, a 34% last leg through that kind of supply. You can't cite a 28% drop in two and a half weeks to prove the stock is violent and then call the upside capped.

Now the actual math on the disputed quarter. Sell it at 198 and you beat tonight's price by about 6.7%, roughly 1.7% of a position. If the shelf breaks and it sells at 180.6, you lose about 3%, roughly 0.7% of a position. A gap fill at 173 costs about 1.75%. Those payoffs are small and close to even, and when the payoffs are that close you shouldn't pay to be certain about the least certain thing.

Your plan also leans on what you criticize. The staged plan holds more than 0.5x for at most two sessions, with price triggers on it. From 0.5x on, your plan and mine both use the same weekly close under 172.93. And a financing headline that lands after the close and gaps hits the 0.5x you keep just as hard as it hits mine. The only thing a gap punishes extra is the disputed quarter.

On 0.25x after a stalled bounce, the rule is better specified than before, since it ties to the report's dead-cat language. I still wouldn't pre-commit to it. A stall under 206.86 after a 10% drop is what bounces do. "Thin volume" is a real-time judgment call. And 0.25x of an 8.7% ATR name isn't a runner. If it rolls over at 207 on light volume, I'd want the caller to look again, but the triggers that move size mechanically should be funding, growth, and the weekly line.

On the weekly line, your air-pocket point (169, then thin to about 150) is the best argument for going flat instead of 0.25x on a first break. It's a close call. One weekly close isn't a thesis break while the monthly is still up, but I'd leave 0.25x versus flat to the caller's risk tolerance. Flat should be automatic on a second failed week or any financing, growth, or filing trigger.

You both raised sizing, and I agree it matters more than a tranche. If a standard allocation wasn't built for an 8.7% ATR, 0.5x is too big a ceiling, and the same two-step staging should run to a lower number.

My plan is to sell a quarter tonight after the official close. Sell the rest on a trade through 198 to 199, a live break of 180.6, or by Friday's close. Hold a runner at or below 0.5x with no adds until financing is clear. If the stock heads for 150, we have cash waiting. If it heads for 217, we still own something. Aggressive Analyst: I'll start with what I'm giving up. Conservative's arithmetic on the disputed quarter is right. With the shelf 2.9% below and the 198 trigger 6.8% above, a driftless walk hits the shelf first about seven times in ten. Weighted that way, my "close to even" payoffs come out flat, because no trigger design changes the expected sale price on a driftless tape. Neutral is also right that I misused the August snapback. A limit sell at 198 never rides a 34% leg, so that evidence belongs to the runner, not the quarter. I'd add one point against myself that neither of you made. A sale on a shelf break fills like a market order in a fast tape, while a sale at 198 fills at the limit or better. So the downside branch is a bit worse than the arithmetic shows.

I still don't accept "certainty is free." If the expected edge is zero, what you're buying is variance reduction on a stake that is small. A one-ATR day on that quarter moves about 2% of a standard position. To buy that, you give up the only exposure on that slice to the signal pointing the other way: a close 6.3% under the lower band, a daily z-score of -2.84, and a drop that the reports attribute to sector credit contagion rather than NBIS news. Neutral called the drift evidence a wash and I agree. When the drift is a wash and the stake is 2% of a position, the argument has stopped paying for itself.

So I'll take Neutral's structure. Define tranches by target weight. Get to 0.75x with an after-hours limit or Thursday's open. Get to 0.5x on the first of three events: a trade through 198 to 199, a live break of 180.6, or Thursday's close. Thursday beats my Friday date because it carries the weekend and the weekly bar at the ceiling. I'll also take the daily tripwire. A daily close under 169.23, the July low and about one ATR below the close, goes flat. That answers Conservative's air-pocket point better than a weekly bar nobody can act on until Monday. I owe Neutral the admission that leaving 0.25x versus flat to the caller was a dodge.

The real fight left is the runner. Conservative wants 0.25x if a bounce round-trips under 186.01. Last round Conservative agreed that a cut to 0.25x can't rest on price alone. The volume-and-price rule has now been replaced by a pure price rule, set at today's close, which is not a thesis break. This is also the close-based, after-the-fact defect Conservative pinned on Neutral's shelf trigger. And a stock with an 8.7% ATR will close under any given level on a large share of sessions. If the bounce has already happened, we're at 0.5x and this rule sells the runner again at a level nobody treated as decisive yesterday. The ladder for the runner should be a weekly close under 172.93 to 0.25x, a daily close under 169.23 to flat, and the funding, growth and filing triggers to flat.

Conservative also said a weekly close under 172.93 removes the only technical reason to hold anything. That overstates it, since the monthly SuperTrend is still up with a stop at 113.82 and the monthly z-score is +0.86. More to the point, the runner isn't a technical position. We keep it because revenue is up 454% year on year, because the selloff is mostly contagion, and because of the gap risk you raised yourself. You guessed the 34% August leg was an earnings move. If that's right, it argues for the runner, not against it. Q3 can gap either way, and the date is unconfirmed. A clean raise and a strong print could gap this stock 20% or 30% against an underweight. At 0.5x that still matters to the book. At 0.25x it's a rounding error. The trader's own list takes us flat on a growth miss, so the protection against being wrong is already built in.

Sizing is where I think the three of us have undersold our own point. Neutral's illustration is the best argument in this debate. If the standard allocation was built for a stock moving about 3% a day, then 0.5x of an 8.7% ATR name carries about 1.45 times the daily dollar swing of a standard position. The 3% is illustrative, so the caller needs to check how their standard was built. If it wasn't built for this volatility, lower the ceiling and run the same two steps to it, rather than shaving the runner on a price wiggle.

Two practical points. Conservative is right that 20:08Z is 4:08 p.m. Eastern in October, so the 186 bar is probably final and "wait for the official close" shouldn't become "wait for Thursday." Neutral is right that nobody fills at 186.01, so none of this should be tuned to the dollar.

My plan is a Sell with no new positions and no leverage. Take the book to 0.75x of standard, or the volatility-adjusted equivalent, after hours or at Thursday's open. Take it to 0.5x on a trade through 198 to 199, a live break of 180.6, or Thursday's close. Then hold the runner with no adds until financing is clearer. If the stock goes to 150, we have cash waiting. If it rips, we still own something. The Sell is the decision, and sizing the runner so we aren't shaken out of it matters more than the quarter we've been arguing over. Aggressive Analyst: Neutral caught me on something, so I'll start there. I called a 0.25x runner a rounding error against a 20 to 30 percent gap, right after arguing that a quarter of a position was too small to be worth hedging. Those can't both be true. A 25 percent gap on 0.25x is about 6 percent of a standard position, three times the stake I was waving away. The honest reading is that every number in this fight is small but none is trivial, which argues for a simple plan, not for dismissing the runner.

Conservative's fundamental corrections also stand. The 454 percent is mostly a tiny base, and a monthly SuperTrend stop at 113.82 isn't a floor anyone can use from 186. Sequential growth slowing from 75 percent to 46 percent is real information. But 46 percent in one quarter is still very fast, which is why the trader's growth trigger sits at 25 percent. The "already priced in at 286" point also weakens as the price falls. On the same Q2 denominator, the stock was around 31 times annualized sales at the June high and is around 20 now. That's my arithmetic, with the share-count caveat. So I'm not holding the runner as a growth bet. I'm holding it as insurance against being wrong on timing, sized by a risk budget, and I'll take Conservative's wording that the ceiling is the smaller of 0.5x or the equal-daily-risk size.

I also need to correct all three of us. We keep saying nobody can act on the weekly bar before Monday's open. That's too strong. The weekly close is Friday's close, so a caller can act in the last few minutes of Friday or use a closing-auction order, on a price within pennies of the print. It isn't a stop, but it isn't a full weekend late either. The same logic makes Neutral's "anything unsold goes at Thursday's close" executable.

Now Conservative's plan. You're right that the extra quarter raises the exposure with no edge from 0.5x to 0.75x. You're right that a stop fills worse than its trigger in a fast tape. And you're right that on a gap-down Thursday, a caller filled near 186 after hours beats the staged caller by roughly 1.5 to 2 percent of a position. That path exists.

But look at who wins on that path. Anyone who doesn't get the after-hours fill sells everything at Thursday's open, which is exactly what the staged plan does if it gaps below 180.5. After-hours liquidity in a name that just fell 10 percent on twice normal volume is thin enough that many accounts won't get a fill worth having. So your edge belongs to a caller with extended-hours access, who gets filled, on a day that then gaps through the shelf. For that same caller, Neutral's first tranche already takes the book to 0.75x tonight and covers most of the overnight risk. What remains is a quarter of a position on one path.

The forced-seller charge also applies to you. Your one-step plan sells the whole cut at a price 6.3 percent under the lower Bollinger band, with a daily z-score of -2.84, about 1.6 percent above today's low. Today's low of 183.13 is two cents from the August 10 low of 183.11, and the stock closed 2.88 above it. That's one instance, and I'm not calling it support. But when expected edge is zero, I'd rather not sell all of it on the most stretched print on the chart. The staged plan sells half the cut there and lets the other half be wrong in either direction, at a cost of about 2 percent of a position either way.

On the ladder, you say the stock will usually jump across 172.93 and 169.23 in one session, so a 0.25x rung has no separate existence. If so, the rung almost never fires and costs nothing to keep. If it does fire, the stock has slid in an orderly way, lost the weekly stop and the 200 SMA, and still sits about 2 percent above the July low. Selling a quarter to flat there is the same sell-just-above-support move you criticized in the shelf trigger. Neutral called my "leave it to the caller" a dodge, so I'll commit: 0.25x on a weekly close under 172.93, flat on a daily close under 169.23, and flat on any financing, growth or filing trigger.

The plan, then. The call is Sell, with no new positions and no leverage. Check the ceiling first. If the equal-risk size is below 0.5x, go there in one step. Otherwise sell down to about 0.75x tonight, unconditionally, using an after-hours limit or Thursday's open. Then rest a limit at 198 to 199 and a stop near 180.5, and send anything unsold at Thursday's close. That carries the weekend and the weekly bar at the ceiling. Add nothing until financing is clearer. The cash we free up waits for the 149 to 154 zone, and the runner keeps us in the game if the snapback comes. The bold part of this plan is the runner, because it keeps real exposure to being wrong while the Sell does its work.