25 years later… software is dying.
Or maybe it’s growing.
Or maybe what’s actually dying is the seat-based model.
But then again, OpenAI and Anthropic still charge per seat and subscription.
Ok what’s the real story here… Lets go.
“Software is dead”: At least that is the tone. Growth has slowed in some areas. Hiring has flattened. AI is rewriting workflows in real time. Margins face scrutiny. Multiples have compressed. The subscription model that defined the last twenty five years feels suddenly fragile.
The argument goes something like this: if AI can replicate features at marginal cost, differentiation collapses. If differentiation collapses, pricing power erodes. If pricing erodes and seat counts stagnate, the economic engine that powered SaaS for two decades begins to break.
It is a coherent story. It feels logical. It feels inevitable really right?
But it may also be measuring the wrong thing.
Because most of the anxiety rests on one assumption that I think few are explicitly stating: that the seat is the business.
What if the seat was just a temporary pricing abstraction for an earlier phase of software? And what if that abstraction is now dissolving?
Lets dig in.
The Seat Was a Temporary Alignment.
The first generation of SaaS digitized work. That it replaced paper processes, manual coordination, and “dumb” spreadsheets with centralized systems. Every employee needed access. Every workflow required a login. The seat was a natural proxy for value.
It was easy, revenue = headcount.
As companies hired, software revenue grew. As organizations expanded globally, seat counts expanded in tandem. It was linear, predictable, and recurring. Investors could model growth off hiring plans. Vendors could forecast revenue with high confidence. CFO’s of the buyers loved it.
But this alignment worked only because software’s primary function was to assist human labor.
Seats monetize access to tools.
They did not monetize the work produced by those tools.
Therefore the structural ceiling of seat-based pricing is embedded in its logic. If output doubles without additional hiring, the vendor does not necessarily participate in that productivity gain.
Imagine a marketing agency uses a design tool. They pay for 10 seats (10 employees).
- Scenario A: The team produces 100 ads a month.
- Scenario B: The team uses AI to produce 1,000 ads a month.
In Scenario B, the agency is delivering 10x more value to their clients, but the software vendor is still only getting paid for 10 seats. The vendor is “locked out” of the extra profit because their price is tied to heads, not results.
If automation reduces the number of employees required to produce the same output, revenue growth stalls even if value delivered increases.
The seat was an elegant solution for a world in which software augmented labor. It is less elegant in a world where software begins to execute labor.
AI Breaks the Headcount Link.
Now AI changes the underlying economic relationship between software and production. It does not merely optimize workflows. It writes code. It drafts documents. It analyzes data sets. It synthesizes information. It executes tasks that previously required employees. It is pretty amazing really.
What is critical to lock in your brain is that when software produces work rather than enabling work, tying revenue to employee count becomes economically misaligned. The revenue driver shifts.
Revenue driver then becomes throughput.
API calls.
Compute cycles.
Transactions processed.
Claims adjudicated.
Decisions automated.
Models run.
This shift from access to activity is not cosmetic. It totally restructures the growth equation for these companies, and most importantly for this conversation, the software industry broadly. Under a consumption model, revenue scales with usage. Usage scales with value created. Value created scales with business output. Business output scales with economic expansion. This decouples revenue from payroll.
And that decoupling is what unsettles traditional SaaS analysis. It is also the biggest opportunity software investors and CEO’s have likely seen…ever.
When Pricing Aligns With Activity, Markets Tend to Expand.
Lets look back in time. History provides repeated examples of this dynamic.
Payments did not become a global toll road because merchants paid a flat monthly software fee. Payments scaled because processors monetized transactions. As commerce expanded across geographies, channels, and verticals, transaction volume increased. Revenue moved in tandem with economic activity. You see this clearly with Mastercard and Visa. Same with Apple’s App Store (you probably didn’t think of it like that). They don’t win because developers pay for access. They win because they participate in every transaction.
Toast is a more recent example. It started with subscription software tied to restaurant locations, similar to seats. But the real inflection came when it leaned into Toast Payments. As more restaurants adopted its integrated payments solution, gross payment volume accelerated across its installed base. Instead of just charging for access to software, Toast began participating directly in the success of each location.
Subscriptions tie revenue to the physical footprint in this case and payments tie revenue to performance. This move expanded their market opportunity. Again the opportunity.
How about Digital advertising, another good example followed a similar pattern. Platforms did not monetize access to dashboards or leads. They monetized attention. Engagement became inventory. More user time translated directly into more monetizable impressions. The digital ads market is multiples of many markets.
Cloud infrastructure scaled because it monetized compute. As businesses digitized more of their operations and pushed more workloads online, compute consumption naturally increased. Revenue grew with actual usage, not just with how many licenses were sold. That’s very different from the old model of selling servers. Companies like Dell and HPE built and sold the hardware. They owned the customer relationships. They had the technicians. But revenue was still tied to one-time equipment sales. The bigger opportunity wasn’t just selling the box. It was renting the compute and charging for usage over time. The cloud players captured that shift. Dell and HPE have since moved toward subscription and consumption models, but the core story remains. The real upside came from monetizing activity, not just shipping hardware.
How about a recent example. Netflix
I think this shows the principle in a more subtle way.
The original subscription model monetized access at a fixed monthly rate. You paid for the seat (household), and that was it. For years, Netflix dismissed ads entirely. It didn’t fit the walled garden model. Access was reserved for paying subscribers only. Then they introduced the ad-supported tier. What changed? They layered monetization on top of engagement (usage). Viewing time became a second revenue surface. Revenue no longer scaled only with subscriber count. It began to scale with usage intensity. The more people watched, the more valuable the platform became, not just to users but to advertisers as well. Usage created value for the customer and incremental revenue for Netflix.
That alignment is the shift.
The common thread above is consistent. When pricing aligns tightly with measurable economic activity, the addressable market expands. Not necessarily because prices increase. But because the base to which pricing is applied grows and aligns output to value.
Software’s Share of Spend Remains Small Today
Again, I think the common view is “software shrinks”. Our view “software will now be bigger than ever”. Here’s why:
Gartner estimates global IT spending at roughly $5 trillion annually, with SaaS near $250 billion, about 5% of IT budgets. Cool right?
Even if you flex those numbers, the direction is clear. Several CEOs have cited SaaS representing roughly 6-8% percent of their IT spend. That is opportunity, that SaaS is only a fraction of total enterprise expenditure.
For most companies, labor, logistics, compliance, procurement, customer acquisition, and outsourced services dwarf software spend. If SaaS captures only a small portion of IT, and IT captures only a slice of overall operating costs, then software today monetizes a narrow band of enterprise budgets. A significant portion of IT spend still sits outside SaaS, think hardware, infrastructure, managed services, legacy systems. As software becomes embedded in production workflows, it begins absorbing parts of that non-software IT spend. That is our view. Cloud already demonstrated this shift by monetizing compute instead of hardware ownership. That became a bigger market overall.
If pricing aligns with measurable output rather than employee access, software competes for operational and labor dollars, not just IT budgets. Some have turned that into a dooms day scenario, but diffusion in enterprises will take time. I talk about this next. But even a modest shift of labor spend toward production-based software meaningfully expands the revenue base.
TO BE CLEAR: Not All Vendors Survive the Transition
None of this implies universal upside.
Seat-dependent point solutions with narrow product offerings face legitimate pressure. If AI reduces the number of users required to produce equivalent output, seat counts may decline. If differentiation is built solely on a slick interface or basic automation, commoditization risk increases. I won’t name any names, but plenty come to mind both public and private.
The biggest factor is time.
Not every company gets the same time window to adapt. Some have years. Some have quarters. Some their days already seem numbered.
Time tends to come from a few places. Serving complex workflows buys time. Deep enterprise relationships buy time. High company specific switching costs buy time. Regulatory friction buys time. Proprietary data buys time. Embedded systems of record buy time. And strong margins and cash flow fuel the transition if you have that time.
Those characteristics slow disruption. They create inertia. They create room to transition from seat-based access to output and consumption based models.
As I mentioned, time alone is not enough. The companies most likely to navigate this shift, in our view, are the ones that have time and capital, and are still led by founders or product builders willing to cannibalize their own models before someone else does. Time gives you the opportunity and leadership determines whether you use it.
Historical enterprise behavior suggests that vendor plurality will remain. Organizations rarely commit fully to a single provider. Multi-cloud architectures and diversified software stacks reflect risk management and negotiation leverage. I think this approach repeats once again.
Moats: What Still Holds?
We all know the word moat. Durable advantage. Staying power. The ability to defend economics over time. That’s what we’re looking for, always. But a real moat shouldn’t evaporate overnight. If it does, it probably wasn’t a moat. It was an advantage that looked durable in a slower moving environment.
Some traditional edges get weaker. Like I mentioned earlier, a polished interface. Basic automation. Aggregating public data with no workflows attached to it. When intelligence and agents become widely accessible, those are easier to replicate. Other advantages get stronger. Proprietary data matters more. Models are only as good as the context feeding them. If a system has captured structured, domain specific data over years, that advantage compounds. The moat builds not shrinks.
Regulatory embedding becomes more defensible. Software sitting inside compliance-heavy workflows is harder to rip out. Always has been and will likely be the case once again.
Network effects can accelerate when activity flows through a centralized platform. More transactions mean more data. More data improves output. Better output attracts more activity. If your business extends beyond the screen into approvals, payments, compliance, physical workflows, or high context decision making, you likely have more protection.
We are still looking for moats. But a moat doesn’t mean immunity. It means time. And in this cycle, time is the asset that allows a company to adapt before the edge fully erodes.
The CFO in a World of Agents.
This is kind of a big deal and have not heard anyone else talk about it…
One legitimate concern around consumption models is volatility in expense line. Usage fluctuates. Customers optimize. Workloads spike and then normalize. Revenue visibility looks very different from clean, predictable subscription ARR.
Put your CFO hat on for a minute…
Traditionally, you ran your business with a high degree of expense visibility. Headcount is planned and budgeted. A portion of sales and marketing is fixed and brand-oriented. Performance spend flexes with demand. Cost of goods is relatively stable and modeled. Now imagine shifting from fixed annual or monthly subscription costs to usage-based expenses that move with activity, IE AI Agents.
Operating expenses and cost of goods become variable. Forecasting becomes harder. Budgeting becomes less clean. We saw a version of this during the shift from on-prem to cloud. On-prem required upfront capex that could be approved, depreciated, and mapped out over time. Cloud moved that spend into opex, tied to usage.
During COVID, some digital businesses saw cloud expenses spike quickly as demand surged. CFOs suddenly had to manage cost variability in real time. Cloud data lake and warehouse providers faced this directly. They had to work with customers to prevent cost sprawl and improve visibility because finance teams were uncomfortable with open-ended consumption. That probably matters in an AI-driven, agentic world.
From the buyer’s side of software, steady expense lines are easier to defend internally than highly variable ones. That likely slows adoption at the margin. This is another reason to believe that consumption models and AI credit systems will diffuse over time, but not instantly. Finance discipline acts as a natural governor here on how fast organizations are willing to embrace fully variable cost structures.
This is just starting.
So what we are witnessing is not the collapse of software as a category. You can make the case for specific companies. But it is the end of a pricing abstraction that defined a specific era. The seat made sense when software digitized work. Turning post-a-notes into digital post-a-notes. It aligned revenue with hiring. It created predictability. It powered two decades of expansion. But the seat was never the core economic engine. It was simply the meter used to measure value in a labor augmented world.
As software begins to produce work rather than merely enable it, that meter becomes misaligned. Pricing tied to headcount gives way to pricing tied to throughput. Access gives way to output. Tools become the infrastructure. Now while everyone is extrapolating linearly, this transition introduces volatility. It introduces competitive pressure. It will separate platforms from point solutions. Not every vendor will make the shift successfully.
But the broader implication is not contraction. It is expansion.
SaaS today remains a small share of IT. IT remains a modest share of total enterprise spending. If software begins capturing value from operational budgets, transaction layers, compliance infrastructure, and even portions of labor spend, the addressable pool grows meaningfully larger than what seat-based models ever implied. We have seen how opening up consumption led to expansion with specific companies examples above. It is that history that suggests pricing transitions often redefine the ceiling.
Software’s next decade will not look like its last. But if pricing aligns with measurable economic activity, software does not shrink, it likely compounds with the activity it powers. The seat is fading, probably not entirely to be honest as it introduces too much CFO noise, but the production layer is rising. If US is in front in AI, then while the rest of the world was the work layer of the world in prior decades, then the implications are that the US technology and software sector will become the work layer of the world over the next decade.
This is the beginning of the next phase.
Twitter: @_SeanDavid
The author and/or his firm have positions in the mentioned companies and underlying securities at the time of publication. Any opinions expressed herein are solely those of the author, and do not in any way represent the views or opinions of any other person or entity.
