Buyers Gain Leverage as AI Rewrites Enterprise Software

The piece from Goldman Sachs Asset Management argues that artificial intelligence is changing how enterprise software is built and bought. Yeah, no: that observation is obvious; the harder, more consequential claim is unstated — that growth equity will fund a very different set of bets than it did in the last decade. The money is going to have to follow the plumbing, not the sheen.

Start with the habit investors picked up in the SaaS boom. They learned to read product-market fit off a dashboard: top-line growth, net retention, new modules shipped, logos added. It was a world where better interfaces, tighter integrations, and more sales reps were the levers that mattered.

That world is quietly ending.

The new competitive moat in enterprise software is data and the compute stack that runs models. Here’s the thing: building those moats looks less like traditional SaaS scaling and more like industrial capital expenditure. Data pooling, labeling, model ops, private clusters — these are long, messy, expensive projects that don’t show up as immediate ARR expansion but sit under whatever AI magic the customer thinks they’re buying.

So growth equity funds, which historically entered once product-market fit was obvious and go-to-market could be poured on like gasoline, now face a weirder proposition. They’ll still be invited in once there’s traction, but the most important checks may be earmarked for infrastructure projects that feel suspiciously like utilities.

That flips the old thesis from “buy more customers” to “buy infrastructure that keeps customers from leaving.” It also forces some unglamorous changes in deal structure. Holding periods stretch as data and model investments take time to pay off. Capital gets explicitly ring-fenced for migration and refactoring work instead of just headcount. Diligence expands from pipeline and churn metrics to questions about data provenance, model lifecycle management, and who actually owns which bits of the stack.

If the Goldman Sachs Asset Management piece sketches a high-level rewiring of the enterprise stack, the downstream story is about where growth dollars get routed: less to aggressive sales hiring, more to GPUs and data contracts that nobody puts in a keynote slide.

On the buyer side, the shift is just as real. Vendors can’t simply slap an “AI” label on existing modules and hike prices. Enterprise customers are already asking what exactly they’re paying for: unique data access, predictable outcomes, or a new vector for operational risk. Vendors that can anchor their offering in proprietary customer data — and prove they handle it safely — will have a stronger negotiating position. Vendors that lean too heavily on commodity third-party models will feel pressure on both margins and trust.

Pricing models will follow that gravity. Contracts that once revolved around seats and feature tiers start to mutate into outcome-based structures and consumption tied to inference or data usage. Finance and procurement teams will poke hard at any “model access” surcharge that isn’t tied to measurable business value. Growth investors that ignore those conversations because they seem “post-investment” are setting themselves up for awkward surprises.

There’s a bigger system-level concern the original article only glances at: concentration. If growth capital disproportionately backs firms with the largest datasets and deepest compute stacks, the market tilts toward a small set of heavyweight platforms that become gatekeepers. That’s efficient from a capital perspective; it’s also a magnet for regulators.

Once a software company turns into a de facto data custodian, it starts inheriting scrutiny that used to live in banks and healthcare systems: consent, portability, auditability, sector-specific rules. Growth equity funds that take big positions in these businesses can’t treat compliance as a back-office function — it becomes central to the investment case. You’re not just underwriting product and go-to-market anymore; you’re underwriting whether the company can survive a regulatory audit without pausing feature development for a year.

I keep thinking of Neuromancer: the protagonist is forever negotiating access to systems owned by someone else, paying premiums to get closer to rare, powerful nodes. Today’s CIOs may wear less leather, but the architecture rhymes — gatekeepers charging for proximity to scarce data and compute, while everyone else rents passage.

There is a thoughtful counter-argument: if AI capabilities commoditize quickly, differentiation collapses, margins fall, and data moats erode. In that world, why pour growth capital into heavy infrastructure instead of waiting for cheaper, off-the-shelf tools and slimmer businesses?

Because commoditization doesn’t eliminate capital needs; it reshuffles them. When every vendor can call the same base models, the edge shifts to who can migrate legacy workflows without breaking them, who can wrap models in reliable guardrails, and who can prove outputs are safe and auditable. Those are capital-intensive projects too — but the money goes to engineering, security, and compliance, not just feature sprawl.

That has implications for how growth equity evaluates teams. The classic pattern — charismatic founder, seasoned CRO, proven playbook for scaling sales — still matters, but it’s not sufficient. Funds will need to favor product-engineering DNA, insist on real data contracts rather than just big logo slides, and interrogate compute and model-governance roadmaps as core risk factors, not technical footnotes.

And talent plus compute costs don’t magically normalize because models get cheaper. The most attractive companies may turn out to be the ones that build clean abstractions over expensive hardware and scarce specialists, making sophisticated AI feel boringly reliable to customers. Funds that assume any developer can be dropped into that environment will learn the hard way what “non-fungible talent” really means.

The Goldman Sachs Asset Management argument that AI is rewiring the enterprise software stack is right as far as it goes; the next chapter is whether growth equity chooses to fund durable plumbing or just higher-gloss interfaces on someone else’s infrastructure.