This week, socials are in a tizzy about a scenario (not a prediction!) that Citrini released outlining a possible 2028 “global intelligence apocalypse.” It arrives on the heels of, and is a possible contributor to, a global meltdown in tech stocks and another thinkpiece arguing that we’re all, as the kids would say, cooked. 

The premise is clean and unsettling: AI agents become ubiquitous. White-collar work is automated at scale. The professional class is displaced. The intellectual aristocracy — long shielded from the impacts of globalization and technological efficiency gains — gets hit by a tidal wave. Unemployment spikes. And consumer spending collapses because, unlike humans, agents don’t eat sandwiches or buy the latest Nikes.

If income disappears for a large swath of high earners, demand must fall. If demand falls, growth falls. The U.S. economy in particular is disproportionately driven by elite consumer spending. When those consumers are wiped out, markets unravel.

That’s the argument, at least. It’s coherent. It makes sense.  

It also has no basis in history.

The “agents don’t eat sandwiches” line assumes that demand is mechanically tied to payroll, and the economy is propped up by inefficient transaction costs. Without that inefficiency, demand might somehow collapse. Agents are infrastructure; if they aren’t people, they don’t have preferences and they don’t spend. From there, a second Great Depression is all but inevitable.  

But that ignores what agents actually do: collapse the cost of deciding, coordinating, and executing. When the cost of something falls dramatically, usage expands. This isn’t blind techno-optimism; it’s observable behavior seen throughout history.  

W.S. Jevons’s The Coal Question (1865) demonstrated that improvements in coal efficiency led to more coal consumption, not less. (This later became known as the “Jevons Paradox.”)

Similarly, W.D. Nordhaus’s paper on the price of light demonstrated that usage exploded as lighting became exponentially cheaper over centuries. A fact even more profound: Most economic indicators significantly underestimate real income growth driven by technological change.  

Realistically, if cognition becomes cheaper, we’ll use more cognition. If decisions become cheaper, we’ll make more decisions. If execution becomes cheaper, we’ll execute more. Agents are extensions of human intent, and will “consume” constantly on our behalf: compute cycles, tokens, API calls, background tasks, triggered services, automated purchases.

And humans have never historically responded to increased capacity by wanting less.

Long-run data makes this painfully clear: Global GDP per capita has increased more than 10X since 1820. Energy consumption per capita has multiplied as societies industrialized.

Each time productivity expands, so too does consumption. What counted as “enough” in AD1500 bears no resemblance to “enough” today. Upper limits on human ambition are consistently underestimated.

If each individual has a small team of agents acting on their behalf — researching, negotiating, building, transacting — their effective economic reach expands. When reach expands, activity expands.

There’s no observable reason to believe that human appetites — for status, resources, experiences — contract in any way as they become more powerful. History has only shown the opposite. As technology makes more things possible, more things are done. And consumed. 

The most serious counterargument is a temporal one.  

Yes, Jevons.
Yes, Nordhaus. 
Yes, centuries of expansion.

But never at this speed: AI capability advances in months, not decades, while inference costs fall quarter over quarter. Entire workflows that felt stable five years ago now feel brittle. 

Thus, the real question is not whether markets will reallocate labor and capital. It’s whether they can match this velocity.

Research suggests two things simultaneously: First, automation tends to displace tasks more than entire occupations. David Autor’s “Why Are There Still So Many Jobs?” argues that tech reshapes task composition and often increases demand for complementary work.

Second, technology diffusion itself has accelerated over time. Comin and Mestieri demonstrate materially shorter adoption lags for major technologies. The same structural forces that enable rapid AI adoption — deep capital markets, global information networks, modular production, high human capital — are also the mechanisms that allow modern economies to reallocate labor and capital more quickly than at any point in history. 

In other words: Disruption accelerates. But so does adaptation.

AI productivity research reinforces this: Brynjolfsson, Li, and Raymond found that genAI increased productivity of customer support agents by ~14%, with even larger gains for lower-skilled workers. The same models that displace tasks also lower the cost of entrepreneurship, coordination, and firm formation.

It might be possible that this time, the dislocation shock of AI is more than the global economy can handle. It’s also possible that the same destabilizing forces might create enough adaptability to forestall global economic catastrophe. We’d need data to answer this question. But for now, the data we do have points more clearly to a broad techno-optimism than its opposite. 

Before moving from the Citrini scenario to today’s markets, let’s address the bigger picture for a moment. 

Citrini’s is fundamentally a demand-collapse thesis: If AI agents displace a significant portion of white-collar work, income and consumption fall, and growth slows. 

That’s a macroeconomic contraction story. And what we’re seeing in public markets right now is not that.

Rather, today’s markets are repricing something much more specific and immediate. On Topline this week, Asad Zaman noted that “over the course of a couple of days, $300B worth of market cap was wiped out from SaaS companies.” Salesforce and HubSpot trade well off prior highs. Forward revenue multiples across the sector have compressed significantly. It looks apocalyptic if you glance at the chart without context.

But when I responded that “the straw man argument is Salesforce is going away. That’s not what anybody’s saying. What everybody’s saying is the stability and predictability of those cash flows are not as stable or predictable as they were yesterday,” I was clarifying precisely this point. 

Markets aren’t declaring that software demand disappears. They’re questioning whether incumbents retain the same pricing power, switching friction, and expansion math in an agentic world. If agents reduce the cost of execution and shift workflows upward, the economic pie may very well expand. What’s uncertain is who captures it.

The SaaSpocalypse, then, is not a demand collapse. It’s a leverage reset. The repricing of control over expanding activity. Expansion and repricing can occur at the same time, and that’s the tension we’re living inside of right now.

The consequences of this repricing are mechanical rather than philosophical. First, whether you’ve changed anything operationally or not, your company is worth less. This is because private SaaS valuations don’t exist in isolation; they anchor, implicitly or explicitly, to public comparables. When forward revenue multiples compress in public markets, private expectations follow. The spreadsheet resets.

So, a business growing 25% at breakeven might have commanded a premium multiple in 2021 because public comps supported that math. But if those comps trade at materially lower multiples today, the ceiling on what someone is willing to pay for your business moves with them. This isn’t judgment; it’s simple arithmetic.

Second (more subtle but arguably more important): Your likely acquirer has less currency. Strategic buyers frequently use their stock as acquisition consideration, and research has shown that acquisition waves are closely tied to equity market conditions. As Asad put it, “strategic outcomes are now more limited because the people that might buy you have less money.” Optionality narrows, not because your product stopped working but because the ecosystem’s purchasing power has contracted. That’s the short-term impact of the SaaSpocalypse.

If you’re operating a SaaS company right now, you have two legitimate paths.  

You can attempt to re-accelerate and prove that you’re positioned to capture the expanding economic surface area. Or, you can become meaningfully profitable.

As I commented on the Topline podcast, “If you’re a $50M business growing 25%, I need you to be generating $10M EBITDA.”

A business that generates real EBITDA has time, resilience, and negotiating power. It compounds independently of market mood.

The Citrini scenario is gripping, but it lacks support in economic history. 

Jevons was right about energy and Nordhaus was right about light: Every general-purpose technology over the last two centuries has automated tasks, disrupted incumbents, and unsettled identities… and then expanded the surface area of the economy.  

That being said, I’m not suggesting complacency. The SaaSpocalypse is real in a different way: Multiples have compressed. Acquisition currency has weakened. Incumbent leverage has eroded. Markets aren’t saying demand disappears; they’re saying that capture is up for grabs.

The path forward is not dramatic. Listen to customers. Build durable value. Focus on long-term economics instead of narrative. Operators move up the stack from execution to decision advantage. CEOs defend real wedges around workflow ownership, governance, distribution, and use AI to expand margin and time-to-value.

Citrini imagines collapse. Markets are repricing leverage. Neither changes the work. 

Expansion will come. The question is whether you’ve built something that earns its place inside it.

Become a Topline insider by joining our Slack channel.

We want to hear your feedback! Let us know your thoughts.