That flip is the story I keep coming back to. Not because every software name is suddenly cheap, and not because chips are finished. Because leadership changed hands in a way that felt almost personal to anyone who sat through the first-half chase. The big averages only managed modest gains over those three months. Under the surface, the rotation was loud. A broad software basket climbed roughly 17 percent. A broad semiconductor basket slipped about 11 percent. Same quarter. Same macro backdrop. Completely different mood.
The Quarter That Quietly Rewrote The Market Script
I’ve found that the quarters people remember are rarely the ones with the prettiest index close. They are the ones where a crowded belief breaks. Heading into summer, the crowded belief was simple. Own the picks and shovels of the data-center buildout, and treat traditional software as a show-me story at best. By the time September wrapped, that belief had a crack in it wide enough to drive a portfolio through.
A veteran market host put it in a way that stuck with me. We may remember this stretch as the moment software bounced and some of the astonishing data-center plays finally met profit taking. That is not a eulogy for infrastructure. It is a reminder that even good stories get over-owned. When everyone is already in the room, the next buyer has to pay up for a reason that is fresher than last quarter’s headline.
Perhaps the most interesting aspect is how fast the fear itself aged. Disruption talk did not vanish. It got more specific. Investors stopped treating AI as a single wrecking ball and started asking a better question. Which companies can sell the tool, sit next to the workflow, and charge for the outcome? That question is kinder to software than the spring version was.
A rally built on fading fear is real, but it is not the same thing as a rally built on fresh demand. The next few months have to prove which one this is.
Market notebook, end of quarter
Why The Averages Hid The Real Move
Modest index gains can feel like a yawn. They are not. They are camouflage. When the headline averages drift higher while one sector rips and another gives back a chunk of a historic run, the index is averaging two arguments. One argument said software had been left for dead too early. The other said data-center winners had borrowed too much from the future.
I like to sketch it this way, because tables sometimes say what a paragraph talks around.
| Pocket of the market | Quarter mood | What changed in the story |
| Broad software | Strong rebound, about 17 percent | AI seen more as a feature than a funeral |
| Broad semiconductors | Pullback, about 11 percent | Profit taking after a huge first-half run |
| Large enterprise platforms | Outsized gains | Workflow tools and cloud demand back in favor |
| Prior data-center winners | Sharp cooling in spots | Great businesses, stretched expectations |
None of those rows is a destiny. They are a snapshot. Snapshots matter because they show you where the crowd just moved, which is usually where the next argument starts.
Fear Had A Half-Life, And It Ran Out
The spring scare was not stupid. Seat-based software really does face a question when a model can draft, sort, and route work that used to need a licensed user. Budgets really did get scrutinized. A few management teams really did sound vague when asked how AI would show up in the bill.
What aged poorly was the all-or-nothing version. Companies do not rip out the system of record because a chatbot got better at email. They bolt the chatbot onto the system of record, then argue about who gets paid. That argument can be won by the incumbent if the incumbent owns the data, the permissions, and the habit. Incumbency is not a moat by itself. It is a head start. Head starts counted again this quarter.
Salesforce And The Case That AI Can Sell, Not Just Threaten
Salesforce was the cleanest illustration of the mood swing. The shares rallied about 46 percent over the quarter. That is not a drift. That is a crowd changing its mind in public.
Two things happened at once, which is usually how these reversals work. The company showed a product path that treats outside models as something customers can plug into the data they already keep there. One version of that idea pairs a widely watched assistant with the records, the tasks, the emails, the updates. Call it distribution more than invention. If your customer already lives in the system, you do not have to win the model war. You have to win the doorway.
The other thing was simpler and, in my experience, more persuasive to skeptics. The quarter itself was solid, and management used a depressed share price to retire stock. Buybacks are not a strategy. They are a tell. They say the people closest to the numbers would rather own the business at this price than let the share count sit there. When that tell lines up with a product story investors can repeat in one sentence, the multiple can re-rate faster than the revenue line.
Can the shares still go higher? The bullish case says yes, because the re-rating started from a place of disbelief, not from euphoria. I would not treat that as a promise. A 46 percent move spends a lot of the easy disbelief. What is left has to be earned in renewal rates, in attach of the new tools, and in the unglamorous question of whether customers pay for AI features or merely applaud them on a slide.
Microsoft And The Argument That The Rally Is Young
Microsoft gained about 37 percent in the same window, which would be a year for most companies and was merely a quarter here. The optimistic read, and it is one I find hard to dismiss, is that the move is still early relative to the product cycle underneath it.
Three threads keep getting pulled.
- Demand for the assistant layered across the office suite looks less like a demo and more like a procurement line.
- Cloud growth has been described as accelerating, which matters because that is the meter the market actually trusts.
- The enormous data-center spend is starting to show signs of payback, the part skeptics said would stay a cost forever.
Spend without payback is a story that ends. Spend with a rising utilization curve is a story that compounds. I do not need the curve to be perfect. I need it to bend the right way for more than one quarter. That is the bar, and it is a fair one.
There is a subtler point. Microsoft sits on both sides of the debate that split the market this year. It sells software. It also builds the infrastructure software runs on. When software was hated and infrastructure was loved, the stock was a tug-of-war. When both sides can be defended in the same breath, the tug-of-war gets easier. That does not make the valuation gentle. It makes the argument cleaner.
The rally is just beginning if the assistant keeps getting bought and the cloud meter keeps speeding up. Those are observable. Everything else is commentary.
Workday, Veeva, And The Relief Trade In Specialized Software
Workday climbed about 55 percent. Veeva climbed about 60 percent. Numbers like that usually mean one of two things. Either the business inflected hard, or the stock had been priced for a problem that did not arrive on schedule. In both cases, the second explanation does a lot of the work.
Specialized software has a strange advantage when disruption talk gets loud. The workflows are narrow, regulated, or painfully specific. Human resources systems and life-sciences platforms are not casual apps you swap over a weekend. The switching cost is not a slogan. It is a project plan, a validation cycle, and a career risk for the person who signs the change order. AI can sit on top of that stack. Ripping the stack out is a different sport.
I have a bias here, and I should admit it. I tend to trust vertical software more than horizontal software when the scare is vague. Vague scares punish anything with a high multiple. Specific workflows punish only the vendors who cannot show a reason to stay. Workday and Veeva spent the quarter being reclassified from “maybe obsolete” to “probably still the system.” That reclassification is worth a lot of percentage points. It is not worth infinite ones.
Cybersecurity Refused To Be A Sideshow
CrowdStrike gained about 39 percent, and the logic was almost the inverse of the spring panic. More powerful models do not reduce the need to lock the doors. They increase the number of doors, the speed of the intruder, and the cost of being casual. Security spend has always had a fear bid in it. This cycle handed that bid a fresh script.
There is a catch, because there is always a catch. Security stocks can trade like growth stocks on the way up and like event stocks on the way down. A single incident, a guidance nick, or a renewal scare can erase a month. The quarter rewarded the idea that AI makes protection more urgent. The next quarter will ask whether urgency converts into durable bookings, or whether customers keep saying yes in surveys and maybe in contracts.
Still, if you are building a mental map of what worked, cybersecurity belongs in the bright column, not the footnote. It was not a sympathy bounce. It was a thesis that got louder as the tools got stronger.
When Last Year’s Heroes Took A Breath
Every comeback has a mirror image. Corning fell nearly 40 percent after a massive run. Caterpillar dropped about 24 percent. If you only read the percentage, you might file both under broken. That would be lazy. A stock can fall hard because the business cracked, or because the price had sprinted ahead of even a good business. These looked, to more than one seasoned watcher, like the second kind.
Corning’s decline was framed as profit taking, the sort of dip a buyer would happily revisit if it showed up again. Caterpillar’s drop was framed as a possible opening, on the idea that engines are becoming a real source of power for data centers, not just a side note in a machinery cycle. I can hold both thoughts. Great end-markets do not immunize a stock from a crowded exit. They do give you a reason to look again after the exit is done.
This is the part of the quarter I do not want to romanticize. Profit taking is not a moral event. It is arithmetic. When a name doubles and then some, the holder who bought the story early has a number in mind. Some of them hit it. The stock does not owe the late arrival a smooth path. If you are hunting dips in former winners, write down what would make the dip a gift rather than a warning. “It fell” is not a thesis. “The end market is intact and the multiple finally exhaled” can be.
Data Centers Did Not Die. Expectations Did Some Of The Work.
I keep hearing a sloppy version of this quarter, the one that says the data-center trade is over. It is not over. Power, cooling, connectivity, and accelerators are still the physical layer of the whole boom. What ended, at least for a few weeks, was the assumption that every related stock deserved a higher price every month.
Think of it like a restaurant that became famous. The food can still be good while the line gets shorter. Shorter lines are not a review. They are a crowd adjusting. Software getting loved again does not require chips to be hated. It requires the incremental dollar to have somewhere else to go. For one quarter, software was that somewhere.
A useful split when a theme cools: Business trend: still up, flat, or cracking? Expectation trend: still rising, stalling, or resetting? Ownership: crowded, balanced, or abandoned? Most "broken" trades are only broken on the third line.
If you separate those three lines, the quarter gets less dramatic and more useful. Software moved from abandoned toward balanced. Some infrastructure names moved from crowded toward merely owned. That is a rotation, not a regime change, until the business lines confirm it.
The Rate Shadow Over An Otherwise Happy Tape
Here is the part that keeps the celebration honest. The biggest worry heading into the new quarter is not another AI headline. It is the cost of money. The central bank lifted its benchmark rate by a quarter point in September. Rate-sensitive shares felt it. Home improvement retail was a clean example of a stock that does not need a software narrative to feel a tighter policy screw.
Why should a software investor care about a hardware-store multiple? Because higher rates do not stay in their lane. They change the discount rate on every long-duration cash flow. Software, for all the relief this quarter, is still a long-duration asset. You are paying today for renewals that arrive over years. When the discount rate rises, the same renewal is worth less, even if the customer never cancels.
My own bias is that people underestimate this channel until it shows up in the multiple, then they overestimate it for a month. A single quarter-point move is not a recession. It is a reminder that the path of policy is not a one-way street toward easier money. If the next meetings lean tighter, the stocks that just re-rated will have to justify the new price with numbers, not with narrative relief.
- Higher rates raise the hurdle for any stock priced on distant cash flows.
- They also cool the parts of the economy that buy big software projects on credit-sensitive confidence.
- They can push investors back toward shorter-duration cash flows if the scare returns.
- Earnings season is the first real test of whether customers are slowing or merely negotiating.
The line I keep hearing, and it matches how I am framing the next month, is that the impact of higher rates is the fear that matters now. Not as a slogan. As a filter. If a company can grow through a firmer rate backdrop, the comeback has a second chapter. If guidance starts to wobble, the 17 percent sector bounce will look like a trade, not a turn.
What Earnings Season Has To Prove
Earnings season is where stories go to get a receipt. I want receipts on a short list, and I want them in plain language.
First, are customers paying for AI features, or sampling them? A pilot is a press release. A price increase tied to usage is a business. Second, are renewal cohorts holding, or is the logo count flattered by discounts? Third, is cloud consumption actually accelerating, or was one noisy month promoted to a trend? Fourth, are data-center investments showing up as margin progress, or only as capex guidance that keeps getting larger?
I have sat through enough of these seasons to know the script. The first week produces a hero and a villain. The hero gets extrapolated. The villain gets buried. By week three, the aggregate looks nothing like week one. If you are positioning off a single print, you are trading a headline. If you are listening for whether the software rebound has a demand spine, you can afford to let week one be noisy.
A Practical Way To Sort The Winners From The Bounce
Not every stock that rose deserves the same sentence. Some rose because the business improved. Some rose because the punishment had gone too far. Both can be buyable. They are not the same hold.
Here is the sort I actually use when a group rips and I feel late.
- System of record. Does the product sit where the data already lives, or is it a feature that can be copied?
- Paid attachment. Is the new AI layer showing up in dollars, not demos?
- Balance-sheet posture. Are buybacks a complement to growth, or a substitute for it?
- Customer concentration. A vertical winner can be wonderful and still fragile if a few buyers set the tone.
- Multiple versus memory. A stock can be up 50 percent and still below the price the business earned two years ago. Context is not an excuse. It is a coordinate.
Salesforce checks several of those boxes in the optimistic version of the story, with the buyback as supporting evidence rather than the whole case. Microsoft checks the cloud meter and the distribution advantage. Workday and Veeva check the switching-cost box, which is boring and powerful. CrowdStrike checks the urgency box, with the usual event risk attached. None of this is a recommendation to chase a print. It is a way to stop treating “software” as one ticker.
The Portfolio Angle Without The Fantasy
A charitable trust tied to that same market commentary owns shares of the customer-platform name, the software giant, and the cybersecurity name. I mention that only because ownership changes how a person talks. When you hold something, you hear the bull case in your own voice. That is fine if you also write down the exit. It is a problem if the hold becomes the argument.
My preference, when a group has already moved this far, is to separate core from trade. Core is the business you would still want if the next quarter is messy. Trade is the re-rating you are renting. Mixing them is how people turn a good quarter into a stubborn position. If rates keep leaning higher, the trade portion should be the first thing you interrogate. The core portion should be interrogated on renewals, not on the last 30 days of price.
Simple hold test: if the price fell 15 percent on no news, would the thesis still read cleanly? If the answer needs the chart, it is a trade.What Could Go Wrong From Here
I would rather list the breaks than pretend the path is smooth. The comeback can stall in ordinary ways.
Guidance can undershoot because buyers paused to see what the new tools actually replace. That pause can look like weakness even when it is digestion. A hotter policy path can compress multiples before a single contract is lost. A handful of infrastructure names can catch a second wind and pull attention, and capital, back to the first-half winners. One messy security incident, or one sloppy integration of an outside model, can revive the disruption scare in a single afternoon.
There is also the quieter risk. The market may have pulled forward a year of relief into one quarter. Relief is not repeatable. You only get to stop being terrified once. After that, you need growth. If the next prints are merely fine, stocks that rose 40 to 60 percent can go sideways for longer than feels fair. Sideways is not failure. It just feels like it when you arrived late.
How I Would Watch The Next Ninety Days
Ninety days is a human unit. It is long enough for earnings, a policy meeting or two, and at least one narrative scare. It is short enough that you can remember why you cared.
I would watch four dials, and I would try not to invent a fifth.
- Whether large platforms talk about AI in revenue terms, not feature terms.
- Whether cloud growth comments stay firm after the victory lap.
- Whether former data-center leaders stabilize because buyers returned, or only because sellers got tired.
- Whether rate expectations ease, stick, or tighten again, because that dial sits under all three.
If the first two stay constructive and the fourth does not worsen, I can live with a choppy third. If the fourth worsens and the first two turn vague, the quarter we just lived through will read as a bear-market rally inside a longer argument. I do not know which script we get. I know which evidence would change my mind, and that is enough to stay in the conversation without marrying the chart.
A Note On Chasing What Already Worked
There is a social tax on missing a move. You feel it when a name you dismissed is up by half and your group chat will not let it go. The tax is emotional. Paying it with a market order is optional.
I have made both mistakes. I have waited for a dip that never came, and I have paid up for a story that had already spent its surprise. The second mistake is more expensive in software, because the multiple can give back the surprise in a week. The first mistake is more expensive in a genuine regime shift, because the dip is a rumor. This quarter had elements of both. The fear was real, and so was the snapback. That combination tempts people to treat the snapback as proof they should never wait. It is not proof. It is one outcome.
A calmer approach is almost dull. Decide what you want to own if the stock does nothing for six months. Size it so a rate scare does not force your hand. Let the next earnings call be a source of information rather than a verdict on your intelligence. Dull is allowed to make money. Exciting is not required to.
The Bigger Picture Under The Percentages
Step back from the tickers and the quarter looks like a negotiation between two eras of the same boom. One era is physical. Power, fiber, accelerators, the loud buildout. The other era is the layer where work actually happens. Records, workflows, permissions, the quiet systems companies complain about and cannot leave. For a while the market paid almost any price for the physical era and discounted the quiet systems as if they were about to be optional. This quarter pushed back.
Pushback is not the final word. The physical buildout can reassert itself the moment a supply number or a power contract surprises. The quiet systems can disappoint the moment a renewal cohort slips. What I do not expect is a return to the spring cartoon, the one where software is obsolete by default and infrastructure is infallible by default. Cartoons are useful for a month. They are expensive for a year.
If I had to leave a single sentence on the desk, it would be this. Software earned back the benefit of the doubt, and a few data-center darlings were asked to earn their prices again. Both of those are healthy. Neither is finished.
Questions Worth Asking Before The Next Headline
Before the next hot take lands, a few questions are worth keeping on a card.
Is the product a doorway to data the customer already trusts? Is the AI feature billed, bundled, or still a slide? Did the company buy back stock because the price was wrong, or because growth needed a costume? Did the fallen infrastructure name break, or did the holders simply collect? And, less comfortably, what happens to the multiple if the next policy move is another tighten rather than a gift?
I do not have tidy answers. I have a bias toward companies that own the workflow, a respect for security spend that rises with capability, and a refusal to treat a 40 percent drawdown in a prior winner as automatic damage. That mix will be wrong in places. It is at least a mix, not a slogan.
The market did not fall in love with software again because the risk disappeared. It fell back in because the risk had been priced like a certainty, and certainty was the thing that broke.
Where That Leaves A Patient Buyer
A patient buyer does not need the quarter to repeat. A patient buyer needs the thesis to survive contact with the next set of numbers and the next policy headline. On that standard, the leaders that paired a real distribution advantage with evidence of demand still deserve a seat in the conversation. The names that only rose because they had been punished deserve a watchlist, not a victory lap. The infrastructure stocks that cooled deserve a second look if the business case is intact and the crowd has thinned.
Rates sit over all of it. Ignore them and the rest of the analysis is a hobby. Respect them and you can still own growth, just without pretending the discount rate is a footnote. That is the grown-up version of this comeback. Less thrilling than the percentage gains. More likely to still make sense in the spring.
I closed those old April notes. I did not throw them out. They are a record of a fear that felt complete and was not. The next set of notes will probably feel complete too. Markets have a talent for that. The useful habit is not certainty. It is keeping a line in the margin for the thing you might be early, or late, or simply wrong about. This quarter, that line belonged to software. Next quarter, it might belong to the cost of money. Either way, the tape already showed its hand once. It can do it again.
]]>
