UGC Created Social Media; Might AI Create "Social Software?"

Are investors and financial markets consistently rational or likely to be correct where it comes to the valuation of high-performance computing infrastructure investments and the likely impact on enterprise software as a service?

Investors seem to be simultaneously betting that AI capex is "too high" and won't produce big benefits, while simultaneously betting that software as a service will be disrupted successfully by AI. 

Can both outcomes occur at once? 

• If AI is ineffective, then the high capex won’t affect SaaS

• If SaaS can be disrupted, then the AI is effective. 

So markets are betting that an “ineffective means” (high AI infrastructure capex) produces an effective outcome (SaaS is disrupted). 

In that scenario, high capex “fails” but then SaaS disruption “succeeds.”

The scenario the market appears not to believe is that high AI capex will prove effective, even in the short term, but that SaaS providers are able to avoid disruption because “write your own app” is not a long-term source of value. 

The business moats for enterprise SaaS lie elsewhere. 

Under what conditions might both be right; both be wrong?

Some might argue the answer is that, “yes, both arguments can be correct, at the same time.”

In this scenario, the "capex is too high" argument could be right if hyperscalers and suppliers of neocloud services are overspending on hardware that they cannot monetize profitably at current margins.

Simultaneously, the "SaaS is disrupted" argument might also be right because AI replaces relatively more expensive software as a service with "service-as-software" in the form of a cheap, autonomous agent.

But it is possible both theses are wrong. 

Perhaps AI increases the value of SaaS. Perhaps:

• The demand for AI turns out to be even larger than anticipated (every dollar of GPU spend generates five dollars of high-margin revenue)

• Incumbent SaaS companies (Salesforce, ServiceNow, Adobe) successfully "wrap" AI into their existing workflows. 

• Users prove they do not want to use 10,000 separate AI agents

• The SaaS "moat" proves to be the proprietary data and the user workflow

• Hyperscalers cut headcount to fund AI capex, with relatively-neutral cash flow implications medium term. 

Outcome	Condition for Capex Validity	Condition for SaaS Survival
Optimism is Rational	AI leads to "Artificial General Intelligence" (AGI) levels of productivity.	SaaS companies fail to innovate, and "Agentic" workflows replace seat-based licenses.
Optimism is Irrational	We are in an "AI infrastructure bubble” and financial returns will not emerge.	Investors overestimate AI's ability to handle complex, m.essy, human-centered business logic.

The market might be irrational, in other words, creating a valuation opportunity:

• Yes, capex is high, but warranted: “market leaders” are being born

• Other means will be found to reduce pressure on free cash flow

• SaaS valuations already have reset, and suppliers will successfully respond

• SaaS value is not based on code, but other variables. 

The analogy might be user-generated content. It might once have been feared it would be a substitute for “professional” content. But that did not happen. Social media emerged as a distinct experience. 

In the mid-2000s, the debate was: Will blogs replace newspapers? or Will YouTube replace TV?

When the cost of "publishing" went to zero, we didn't just get the same news in a different format; we got a massive explosion of new content types (influencers, viral memes, and real-time citizen journalism) that professional studios could never have produced.

Feature	The Social Media Shift (Web 2.0)	The AI Shift (Generative Era)
The "Threat"	UGC replacing Professional Media.	AI replacing SaaS/Labor.
The Reality	Expanded the total volume of content by orders of magnitude.	Expands the total volume of "work/code" by orders of magnitude.
The New Form	The Feed (Algorithmically curated attention).	The Agent (Autonomous execution of intent).

If SaaS is about providing a tool for a human to do work (Salesforce is a tool for a salesperson), the new form of AI is about providing the outcome itself.

Instead of primarily displacing SaaS, AI is enabling a "service-as-software" or maybe “social software” model:

• Hyper-personalized software: AI might "spin up" a bespoke, ephemeral interface tailored to a specific project’s needs, then dissolve it when the task is done

• Shadow workers: In social media, everyone became a creator. In the AI era, everyone becomes a "manager" or "orchestrator." The "something else" is a world where a single person can run a complex operation that previously required a department of 50

• Micro-services: SaaS was always limited by "Total Addressable Market" (TAM). If a problem was too small, no one built software for it. AI can make it profitable to solve problems that were previously too "minor" to automate.

So we might be looking for signs that “social media” is developing. Perhaps it is not so much that SaaS is disrupted as it is that “social software” is developing.