The Retreat From Software And Why It’s the Wrong Call

The bear case on software is understandable, and still wrong.

I’ve been getting a version of the same conversation a lot lately. Sophisticated investors (GPs and LPs), thoughtful in their reasoning, tell me they’re pulling back from net new software commitments for the next year. Their logic: AI forces investors to make a wider range of assumptions, the risk profile has fundamentally shifted, and private market pricing isn’t reflecting that. They can’t see how they get compensated for the added uncertainty.

I understand exactly why they’re saying this. And I think they’re making a mistake.

The Bear Case for Software

The concern isn’t irrational. The questions are real: Will AI-native products displace point solutions? Will copilots erode seat-count economics? Does AI compress time-to-build so much that it also compresses time-to-competition?

When you’re underwriting a software investment at a valuation set in calmer times, and you don’t know which of these scenarios play out, you’re carrying more assumption risk than you were in the SaaS heyday of 2019.

And the private market hasn’t fully repriced to reflect it. Public SaaS multiples have settled into a 2-7x ARR range (median ~5.4x), roughly back to 2016-2017 norms. But the uncertainty profile is decidedly not 2016. A buyer in 2016 knew what they were buying. Today, they’re buying a company that may be augmented or disrupted by something that didn’t exist eighteen months ago. Investors aren’t wrong that the spread between risk and price is uncomfortable.

But here’s where I disagree with the conclusion they’re drawing from it.

The Category Confusion

The broadest problem with a blanket “no new software” posture is that it treats “software” as a monolithic category when it isn’t. It never has been. AI is doing very different things to very different parts of the stack.

For light-weight SMB SaaS built on thin differentiation and frictionless user experience, there’s genuine pressure. Products competing on UI/UX convenience for small buyers with short payback cycles are the most substitutable. The concern applies there.

But for enterprise software, infrastructure, and DevTools, the story is almost the opposite.

Enterprise software buying is not spontaneous. It involves procurement cycles, security reviews, compliance requirements, multi-year contracts, and deep integration with existing systems. No enterprise is buying from a vendor that may wipe their entire codebase. These are not markets that flip overnight. The companies in this segment that are winning are the ones embedding AI into products that enterprises already depend on, not getting replaced by it.

Enterprise software spending globally hit $341 billion in 2024, growing at roughly 11% annually, with projections to $466 billion in SaaS alone by 2029. Gartner is forecasting 14.7% software spending growth in 2026, the highest numbers they’ve ever recorded in enterprise tech. The AI spending wave isn’t eating software budgets. It is the software budget. Enterprises gave up trying to build their own AI in 2024. They’re buying from vendors instead. Just select vendors with unique capabilities that can’t be replicated. Technical differentiation matters. And customers are paying more for it.

This nuance matters enormously. The question isn’t whether AI disrupts software. It’s whose software it disrupts. High-quality, deeply embedded enterprise products are showing the opposite of displacement risk. They’re showing AI as leverage.

AI Compresses Timeline, Not Opportunity

The argument that AI increases investment risk assumes that AI is primarily a competitive threat. But for companies primed to thrive, AI is primarily an operational multiplier.

12-person companies running Cursor, Cognition, Codex, and/or Claude internally is shipping features at a pace that used to require +40 engineers. High Alpha’s 2025 SaaS Benchmarks Report found that SaaS companies with AI deeply embedded in their products grow twice as fast as those where AI is a secondary feature. At the <$10m ARR cohort specifically, the gap is even more dramatic.

The efficiency and scalability implications are the parts that get under-discussed. When a company can do more with fewer people, the total capital required to reach a given scale drops. That directly reduces the risk of dilution-at-exit and improves return profiles for early growth investors. It also opens options for founders without limiting their potential (which is the argument against efficiency in years past). Scalability is no longer inversely correlated to profitability. It also means the companies that will win are not the ones that raise the most, but the ones that deploy AI most intelligently against their existing moat. As an investor, that’s a selection problem, which is exactly what underwriting is for.

Where The Real Risk-Return Mismatch Is

Complaints about pricing not reflecting risk is accurate. But the direction is wrong on where that mismatch cuts against investors.

The private market IS mispricing risk - at late stage, not early stage. Where consensus is driving inflated valuations as capital crowds in.

Late-stage software valuations still carry significant pre-correction expectations baked in. The 2021 vintage hangover hasn’t fully cleared. There are hundreds of software unicorns that haven’t raised since 2022 and will eventually need to reprice against fundamentals that don’t support their last marks. That’s where the asymmetric risk is sitting. When those companies try to come back to market in 2026 or beyond, the reckoning happens.

The entry point where we invest ($2-7m ARR) is structurally different. At that scale, you’re paying for what exists and a bit of what’s to come, without the projection of hypergrowth (if that happens, great!). Multiples at the early growth stage are grounded in current revenue with a modest premium for trajectory. SaaS Capital’s index pegs private company multiples at roughly 4.8-5.3x ARR for well-performing companies at this stage, a rational multiple for a real, profitable business - and in line with most public tech companies. Investors thrive with company success, and alignment is locked in. Cambridge Associates data shows growth equity as the most consistent, high-performing private asset class, precisely because entry discipline at that stage is easier to maintain.

LPs are worried about paying 2021 prices for 2025 risk. So are we. Having a blank slate now is an advantage when violent market shifts occur. But the answer is to be selective about stage, valuation, and sub-category, not to exit the software asset class as a whole.

Why Overlooked Companies Specifically De-Risk This

The companies most exposed to investor concerns are the ones the capital machine already discovered. Heavily funded, multi-round companies burning toward growth targets with inflated last-round valuations are the assets I’m most worried about. AI serves as a threat, leads to pricing dislocation, and creates exit uncertainty. Stacking these risks on top of each other makes investing here nearly untenable.

The companies we target aren’t on most people’s radars. They’ve been building efficiently, profitably, in markets overlooked by the crowds. They don’t carry the stacked risk of inflated prior marks. They don’t have warped cap table dynamics. They’re not in a race to IPO before the window closes. They’re quiet compounders, building what customers need.

A profitable $4M ARR company with a clear technical differentiation, no capitalization issues, and a founder who sees over the horizon is not carrying the AI disruption risk that some investors are worried about. It’s carrying the risk that its current success becomes too significant to risk an even larger future.

The True Consideration

Here’s the honest version of what I’d ask investors to consider.

Software is not a category in decline. It’s a category in flux. The market is punishing assets that shouldn’t have been priced the way they were, while the underlying demand for enterprise software - deeply embedded, workflow-critical, compliance-laden - keeps compounding. Personally, it points me toward enterprise and infrastructure and away from vertical software (you can read more on my views here or watch a conversation I had about this here)

The risk-adjusted opportunity in that environment isn’t to sit out. It’s to be precise about where you enter, how much you pay, and what you’re actually buying. Late stage, broadly distributed, index-like exposure to software is probably getting riskier. Early stage, concentrated, high-conviction growth equity in capital-efficient companies operating in durable enterprise categories is probably getting more interesting, not less.

Supercruise doesn’t run on the excess fuel that powers the boom cycle. It runs on the companies that found efficient flight before anyone showed up to help them. In a market that just discovered it overbuilt its assumptions, those companies look increasingly attractive.

The investors that pauses here will likely come back in a few years, when the pricing has adjusted and the consensus has reformed around the winners. That’s fine. That’s how this works. But the window to enter at the current basis, in the current market, with the current degree of access to overlooked companies… that’s not permanent.

---