Software’s Good Story – and Why It’s Not Good Enough
The darling sector of the last decade is caught between an attractive narrative and a deteriorating fundamental picture. The message is clear. The risk-reward is not in your favor.
«It ain’t what you don’t know that gets you into trouble. It’s what you know for sure that just ain’t so.»
Mark Twain, American Writer (1835–1910)
Software stocks are bouncing.
After one of the sharpest sector sell-offs in recent memory, the iShares Expanded Tech-Software ETF (IGV) has staged what looks like a relief rally. The narrative is seductive: valuations have compressed, free cash flow yields have expanded, and these are still high-quality businesses with recurring revenue.
What’s not to like?
Quite a lot, actually. The fundamental picture is deteriorating in ways that a bounce cannot fix. And the chart, read correctly, is telling a story. One that most software investors do not want to hear.
I say this not as an outsider looking in. As quality investors, we have been invested in software for years – and we liked it. The sector ticked every box we care about: wide moats, high visibility, low leverage, exceptional profitability, and solid structural growth. For a long time, software was a natural hunting ground for our approach. The kind of businesses you buy and hold for decades, not quarters.
We have turned away.
When quality investors walk away from a quality sector, it is worth understanding why.
The «Good Story»: 20x Earnings, Great Margins, Buy the Dip?
Let us start with what bulls see. Because they are not wrong about the facts. They are wrong about what the facts mean. I have had this conversation many times over the past few weeks – with clients, with peers, even with myself. The pitch is good. Almost too good.
Software companies have, for good reason, been among the most admired businesses of the past fifteen years. Recurring revenue models. High gross margins, often above 75%. Capital-light balance sheets. Enormous returns on invested capital. Predictable, subscription-based cash flows that lend themselves beautifully to a discounted cash flow framework. For a long time, these were the highest-visibility earnings streams in the equity market.
And valuations have come down. The software sector’s forward P/E ratio, measured by the IGV ETF, has compressed from north of 40x at the 2021 peak to roughly 20x today. That is a significant de-rating. At 20x forward earnings, the premium to the equal-weight S&P 500 has collapsed to around 19% – down from over 150% at the peak. On the surface, this looks like the kind of reset that creates opportunity.
Chart 1: Software forward P/E (IGV) — absolute and relative to equal-weight S&P 500
Source: FactSet, Goldman Sachs
The «Good Story» is simple: the market overreacted. These are wonderful businesses at now-reasonable multiples. The AI disruption narrative is overdone. Incumbents will adapt, integrate AI into their platforms, and emerge stronger. Buy the dip.
It is a compelling pitch. And here is what makes it so seductive: the numbers have not deteriorated yet. Adobe’s revenues keep growing. Salesforce keeps delivering. SAP’s order book is intact. If you look only at the latest earnings reports, you would see businesses firing on all cylinders. Nothing is broken – on paper.
But I think that is precisely what makes this moment dangerous. Because if there is one lesson I have learned over a decade of watching markets, it is this: Mr. Market prices the future, not the present. He moves before the spreadsheets catch up. And right now, the benefit of the doubt belongs to him – not to the sell-side consensus.
The Uncomfortable Truth: When Uncertainty Becomes the Moat’s Kryptonite
So what is Mr. Market seeing that the earnings reports are not yet showing?
One word: uncertainty.
Not bad news – uncertainty. And uncertainty is the kryptonite of the stock markets. Why? Markets can price bad news. They discount it, adjust, and move on. What they cannot price is a future they cannot model. And that is exactly where software companies find themselves today.
The arrival of large language models (LLMs) has introduced a structural uncertainty into the software business model that is unlike anything the sector has faced before. This is not a cyclical downturn. It is not a temporary compression in IT budgets. It is a fundamental question about the durability of the moat itself.
And that – the erosion of the moat – is the worst thing that can happen to a high-multiple stock. Worse than a revenue miss. Worse than margin compression. Because when the moat erodes, everything in the valuation framework changes: the terminal growth rate, the discount rate, the length of the high-growth runway, and the confidence interval around all of them.
Software companies used to command 30x or even 40x earnings because investors believed in three things: the stickiness of their product, the defensibility of their switching costs, and the predictability of their long-duration cash flows. All three are now under pressure.
LLMs are collapsing the cost of capability at a speed that has no precedent. What once required a team of forty engineers and two years of runway can now be assembled by a single developer in a weekend. The replacement cost of a SaaS product that would have been valued at $20–100 million in ARR a few years ago is approaching the cost of a cloud compute bill and a long weekend.
When the replacement cost of capability approaches zero, the scarcity premium that justified sky-high multiples evaporates.
Now, the pushback I hear most often goes something like this: «Sure, you can build a product in a weekend. But try selling it to a Fortune 500 company with SOC 2 compliance, 18-month procurement cycles, and a legal department that needs to approve every vendor.»
Fair point.
The code can be replicated in a weekend. The enterprise sales motion cannot. But when code becomes commoditized, the pricing power that funded that sales motion erodes too. The moat shifts from product to distribution – and distribution moats are narrower and command lower multiples.
But it gets even worse. And that’s the part that keeps me up at night.
The real structural damage is the erosion of pricing power within existing customer relationships. Most enterprise software is priced per seat. When AI agents do the work of five employees: fewer seats, lower revenue per account. And when customers renegotiate armed with AI-native alternatives, the pricing per remaining seat compresses too. A double squeeze. That is a structural repricing of the entire SaaS business model.
This is why 30x earnings is very likely not coming back for this sector. I wrote in «AI Eats Software» that Adobe’s P/E had compressed from 27 to 13 while earnings kept accelerating – the textbook definition of a market that sees something the income statement does not yet reflect. The old multiple assumed high visibility, high stickiness, and high pricing power. All three are deteriorating. At below 20x, the market is acknowledging the new reality. But whether even that is generous depends on how fast the erosion plays out.
This does not mean every software company will disappear. Adobe will still exist in five years. Salesforce will still exist. SAP will still exist. But the question is not whether they survive. The question is what they will earn, at what margin, and with what degree of predictability. Survival and prosperity are two very different things.
We have covered Adobe extensively at arvy – it was a portfolio holding for years and a stock I know inside out. Today, it trades at a forward free cash flow yield of over 10%. That looks cheap – until you see what Mr. Market is actually telling you. Adobe’s FCF yield expanded from 2% to 10% in two years, not because profitability improved, but because the market is repricing the durability of those cash flows.
Generative AI is attacking the core of what Adobe sells: the creation of visual and written content. The adoption curve is steeper than the internet’s. And the market started pricing this in long before the consensus caught on.
Chart 2: Adobe (ADBE) — Forward free cash flow yield. From 2% to over 10%. Mr. Market started pricing this long ago
Source: Fiscal.ai
A 10% FCF yield on a business facing structural competitive pressure is not cheap. It is a market reflecting genuine doubt about the durability of those cash flows. Mr. Market did not just wake up yesterday. He started pricing this deterioration long before the consensus caught on.
Of course, not all software is equal.
Adobe is probably the most directly disrupted large-cap name: generative AI attacks the core of what it sells. Companies like ServiceNow, where AI is additive, tell a different story. But betting on the exceptions while the sector tide is going out requires conviction most investors overestimate. When the broad sector is in a primary downtrend, even the good names get dragged down.
LLMs are not a one-off disruption. They will continue to erode moats and compress margins for years. In my humble opinion, the old norm of 30–40x P/E won’t return, at least not in the coming quarters, if not years. Perhaps 20x is the new normal – perhaps even generous.
I am not saying software is uninvestable at any price. At 10–12x earnings with stabilizing margins and a return of visibility, the sector could become attractive again. But we are not there. And buying at 20x hoping fundamentals catch up is the kind of bet Mark Twain was warning about.
Enough about the «Good Story» and the corresponding fundamentals.
Let’s now turn to Mr. Market’s objective assessment.
The «Good Chart.»
Interested in arvy’s Database and Portfolio (30 Holdings)?
The Chart Speaks: A Textbook Washout – and a Trap for the Impatient
Looking at the price action, it is not just confirming the fundamental story – it is telling us exactly where we are in the psychological cycle.
Elliott Wave theory, at its core, is not about drawing lines on charts. It is about crowd psychology. Ralph Nelson Elliott observed that markets move in recognizable patterns driven by the collective emotional cycle of participants: from hope to euphoria to denial to panic to resignation – and finally, to a recovery attempt that tests whether the damage was temporary or structural.
The basic structure is elegant. An impulsive move unfolds in five waves: wave 1 begins the trend, wave 2 corrects but does not make a new extreme, wave 3 is the most powerful and typically the longest, wave 4 provides a counter-trend bounce, and wave 5 completes the move – often with waning momentum and divergences that signal exhaustion. After the five-wave impulse is complete, a three-wave corrective structure – labeled A, B, C – follows.
Fun fact: 70% of all corrections and pullbacks follow an ABC pattern.
Chart 3: The classic Elliott Wave structure — five impulsive waves followed by an A-B-C correction
Source: equity, Yazeed Abu Summaga
As for the software industry, let’s apply this theory in reverse.
What matters here is not the mechanics. It is the psychology. Each wave maps to a stage of the crowd’s emotional cycle.
Wave 1: disbelief – the decline begins but is dismissed as a pullback.
Wave 2: hope and buy the dip – a bounce convinces people the worst is over.
Wave 3: panic – reality hits, forced selling accelerates, and the damage becomes undeniable.
Wave 4: exhaustion – a weaker bounce, the crowd is tired.
Wave 5: capitulation – the final leg down, often on diverging momentum, where the last holdouts give up.
And then the A-B-C correction: the market catches its breath. It is not a reversal. It is a pause.
Now look at the iShares Expanded Tech-Software Sector ETF (IGV) since its October 2025 high.
Chart 4: IGV daily – five-wave decline complete, now in an A-B-C corrective bounce
Source: TradingView
The pattern is textbook. A five-wave decline from the October high, with a clear wave 3 panic that took the sector to 52-week lows by late January while the S&P 500 traded near all-time highs. Then a wave 4-5 washout with visible momentum divergence: price made lower lows, but selling pressure was waning. Classic exhaustion.
We are now in an A-B-C corrective bounce. Wave A rallied sharply from the March low. Wave B is forming – a higher low that gives bulls hope. Wave C, if it plays out, could take IGV meaningfully higher from here.
This is where most investors get excited. «The bottom is in! Time to buy!»
I get it. I have felt that pull myself. After a 30% decline, the contrarian instinct kicks in. But here is where discipline matters more than instinct.
Not so fast.
An A-B-C correction is, by definition, a counter-trend move. It does not reverse the primary trend. It does not resolve the structural questions. It means the market is catching its breath before deciding what comes next.
And what comes next depends entirely on the fundamentals. If LLM disruption turns out to be overstated, if moats prove more durable than feared, if margins stabilize – then the correction becomes a base and a new bull trend can emerge. But if pricing power keeps eroding, if free cash flow keeps compressing, if uncertainty remains the dominant force – then the A-B-C bounce is just the counter-trend before the next leg down.
The Asymmetry Is Wrong: Limited Upside, Significant Downside
Buying software here might work out. It might not. But that is not a bet worth taking.
If you are right, and the bounce extends into a new uptrend, your upside is capped by the structural headwinds. These stocks are not going back to 40x earnings. The best case is that they stabilize around 20–25x, margins hold, and you earn a market-like return. That is the bull case.
If you are wrong, and the fundamental deterioration continues, you are holding a former darling in a sector where visibility is collapsing, pricing power is eroding, and the market is still in the early stages of re-rating what these businesses are worth.
The downside from a failed A-B-C correction is another impulsive decline — wave 1 of a new, larger-degree move lower. That is the bear case.
The asymmetry is the wrong way around. Limited upside, significant downside, and a fundamental backdrop where the margin of error is razor-thin. That is not a risk-reward profile that serious investors should be attracted to.
«Good company» and «good stock» are two entirely different things. The Nifty Fifty of the early 1970s were wonderful businesses. But investors still lost 50 to 90 percent – in companies that continued to operate profitably. Software in 2026 faces the same problem.
Mr. Market Knew First – As He Always Does
There is a pattern I have seen again and again over ten years of watching markets. The price moves first. The narrative follows.
Software stocks began underperforming long before the consensus recognized the scale of the LLM threat. Adobe’s free cash flow yield started expanding – meaning the stock was falling relative to its earnings power – well before analysts started writing about AI disruption. Mr. Market, as always, was early.
Five-wave declines do not happen in healthy sectors. They happen when something structural has changed. And in software, something structural has changed.
Simply put: this is not a bet worth betting on.
From Software to HALO: Where Clarity Lives Now
And here is the final irony. The numbers in software have not broken yet. Revenues are still growing. Margins are still high. The businesses are still executing. But that is exactly how it looked for the Nifty Fifty in 1972, for Cisco in early 2000, for the banks in mid-2007: the fundamentals always look fine right up until they don’t. Mr. Market does not wait for confirmation. He prices the trajectory.
For a decade, the market worshipped asset-light businesses – scalable, code-based, capital-efficient. The lighter, the better. But when disruption turns against code itself, when the thing you built your moat from becomes the thing that erodes it, the logic inverts.
The businesses that suddenly look attractive are the ones the market spent fifteen years ignoring: «HALO» stocks – Heavy Assets, Low Obsolescence. Mines. Pipelines. Grids. Infrastructure. Defense. The things that cannot be replicated by a founder with a laptop and an API key. The things that take permits, capital, and decades to build.
At arvy, we find ourselves in exactly this situation after selling our software companies – which have helped us generate attractive returns with low risk but ultimately cost us dearly – that we owned for many years over the course of the past year.
We now own ExxonMobil and Chevron – businesses with irreplaceable physical assets and pricing power that strengthens when the world gets messier.
We own Waste Management – try disrupting landfill permits and collection routes with an LLM. We own Safran and GE Aerospace –a duopoly in jet engine manufacturing where the barriers to entry are not switching costs or network effects, but decades of engineering, certification, and installed-base lock-in.
These are businesses where the moat is not code. It is physics, regulation, and scarcity. As I described in «The Return of the Old Economy», the very constraint that made these businesses unfashionable for a decade – their physical, capital-intensive nature – is now their greatest source of pricing power.
In the past, the competitive advantage of software lied in its code and in the fact that it was integrated into customers’ daily lives. Code is increasingly becoming a commodity. Integration is losing its importance due to the new circumstances.
A pipeline’s moat is geography, regulation, and physics. Good luck commoditizing that.
Markets do not reward you for catching falling knives. They reward you for clarity. And right now, the clarity is not in software. It is in the heavy, the physical, the irreplaceable.
Remember Twain’s warning: it is not what you don’t know that gets you into trouble – it is what you know for sure that just ain’t so. What most investors «know for sure» about software – that these are unassailable franchises, that the moats will hold, that 20x is cheap – may turn out to be exactly the kind of certainty that gets them into trouble.
You can ignore Mr. Market’s message. But his signal is not ambiguous.
Legal Notice: At the time of publication of this article, the companies mentioned may or may not be portfolio holdings of arvy. Such securities may or may not be included in the portfolio at any point in the future. This document has been prepared solely for informational and marketing purposes and does not constitute an invitation, offer, or recommendation to acquire or sell any financial instruments or to engage in any other transactions. The relevant prospectus and key investor information documents (e.g. PRIIPs KID / Key Information Document) are available free of charge from arvy AG.
Original article here.