AGI's Impact on Tech, SaaS Valuations

Thought experiments on how AGI affects SaaS companies of all shapes and sizes

Back in March 2023, I wrote that the 1) rapid advancement of AI can be bearish for the median tech company, and 2) investors should bet on greater dispersion of fates among tech companies.

The core argument was this. A stock’s value ultimately rests on 1) discount rate, 2) projection of future free cashflow, and 3) multiple. Discount rate is a function of risk premium (RP), which should go higher as AI advances, thus bringing valuations lower.

RP increases because very advanced AI increases disruption and execution risk for most tech companies. Even if companies manage to avoid disruption, it becomes harder to project cashflows, because it’s harder to predict future product and revenue mix. This should lower the multiple. (more nuanced discussion here)

Fast forward 6 months, those predictions have mostly been on the right track. $META, $NVDA, $GOOG, $MSFT have rallied ~15%-90%+ further since, while Cathie Wood’s innovation fund ($ARKK), $WCLD, etc, are unchanged since March.

But it’s time to revisit those predictions again, given how much has happened since:

• Massive AI enablement efforts: a much wider awareness among Enterprise customers in every sector about how to apply LLMs (and soon multi-modal).

• OSS LLMs: the birth and flourishing of OSS LLM ecosystem, especially Llama2

• Democratization (?) of RAG, vector search techniques

• Maturing and widening of AI offerings by Azure, GCP (VertexAI), AWS
  
  

In this post, I will do some thought experiments as to how "really” advanced AI - AGI - impacts SaaS (public and private), and make a few additional predictions about SaaS and DB companies specifically.

Assumptions

As for the definition of “AGI”, let’s just assume something a few generations better than current GPT-4, is completely multi-modal (works on spatial 3D video, even), with low latency, can be retrained and re-parameterized effortlessly, hallucinates minimally, has near infinite context length, is perfect at agentic task execution and long term problem solving, and not need RAG, etc. Oh, and it can click buttons on your existing UIs.

One relevant test of AGI would be that it allows 1 non-technical person faithfully replicate an app as complex as Airtable or Salesforce via 1) taking a video of the app, 2) stating the requirements, 3) and having the AI produce and customize the app, while 4) migrating all existing data. All within say 20 minutes.

Let’s also assume that such AGI is owned by at most 1-2 companies, and trade secrets will be guarded even more fiercely through multi-year non-competes (kind of like high frequency trading industry, but even worse). Also assume that even if secrets leak, OSS community won’t have the right talent and capital combination to close the gap materially, at least for ~1 year.

Lastly, let’s assume all this happens within 5 years - which is realistically the time horizon that the stock market should be discounting, as it tends to be forward looking.

Here’s AGI’s Impact on SaaS, Tech Valuations

To brainstorm how AGI impacts SaaS / Tech valuations, let’s first observe the following effects of AI:

• the advance of AI both increases the demand and the supply of AI-powered solutions - whether they are made by professional developers, amateurs with self-serve tools, or your grandma talking to ChatGPT.

  • We will be in some type of a “pervasive AI” situation.

• consequently, AGI will initially accelerate both the demand and supply side of AI-powered software.

• due to supply side expansion of really good AI, it will become harder to enjoy sustained excess profits from AI-powered experiences.

• innovation cycles get shorter, boom and bust cycles also. This is inherently bearish as it increases risk premium.

• the monetizable apps and experiences will use the very best AGI, but they will be beholden to the AGI provider. This increases the supplier’s power for everyone, and hence lower overall moat, too.

• AGI will compress most “BS” work into simply a prompt, and this will lead to redistribution of knowledge workers to other sectors.

So how do these axioms affect your average public SaaS company’s position. There are really 3 competitive threats:

• AGI acquiring some emergent ability that essentially “removes” the raison d'etre of your SaaS. For example, if AGI can do taxes perfectly in 5 minutes, why need Intuit ($INTU)? If AGI can effortlessly create AR experiences of new furnishing concepts for your home, why need Matterport ($MTTR)?

• AGI creating some new category of app that existing SaaS incumbents can’t enter, and threatens to eat into their lunch (due to talent, research gaps). For example, say AGI enables personalized, real-time spatial content creation to satisfy every fantasy. What does not mean for Netflix ($NFLX)?

• Extreme margin compression due to more supply of software. AGI simply enables vastly more creation of software, which hurts the pricing power of incumbents. Every company will do some double take on their IT spending, and try to rebundle.

Deriving from the above axioms, AGI’s impact on SaaS and tech company valuations could be the following:

• AGI is especially bearish for project management software or any CRUD app. Many of the past SaaS winners were essentially CRUD apps that facilitated workplace collaboration and productivity (Airtable, ClickUp, Asana, Monday, etc).

  • They were accordingly priced with a per-seats model, and benefitted from the rise of other SaaS startups. With AGI, it’s hard to justify paying a fortune for a project management software, when there’s just fewer "steps” or “workflows” to most tasks. Perhaps these companies will become even thinner wrappers around the AGI model, in which case they will lose their differentiation and defensibility. Per-seats pricing will be at risk if there are simply fewer seats per company.

• Lower margins due to supplier pressure (from the AGI provider): If AGI were to play the role of “operating system” for the next generation of apps, then incumbent SaaS companies will lose bargaining power. Value will aggregate to the AGI provider, similar to how value accrued to Microsoft in the 90’s due to its Windows operating system.

• Lower margins overall due to competitive pressure from 1) AI-native startups, and 2) self-serve app creation: The vast, vast majority of AI startups won’t work out, but with AGI, it’s plausible to see some of them take 80% of functionality of, say Airtable, and undercut them to win customers (even if they weren’t doing anything mindblowing).

  • And this will be possible because the new AI startups are extremely light on headcount, and with AI assisted development, AI support, and managed services, it’s theoretically feasible to scale to millions of users with just a 5 person team. Either case, incumbent SaaS companies will be forced to think much harder about both vertical and horizontal expansion (even international), and that only makes competition worse for everyone.

• Note, the impact of AGI has little to do with whether the SaaS company sells vertically or horizontally. AGI, if developed, will be a key part of any software’s value chain, and become a big chunk of the whole value, as software reorients itself from being a CRUD app layer to an intelligence layer. If much of your SaaS company’s value prop is from being a workflow management tool, then it’s more vulnerable to AGI.

• Basically, every SaaS company that facilitates knowledge work is at danger: The foundational assumption of all business software is productivity gain. If AGI simply compresses and subsumes a N-step workflow into a single step one, then that base assumption is violated. Of course, new workflows may spring out after AGI, but those will probably not knowledge-based but human service based, and not facilitated by software.

Defensibility in AGI Era

At this point, you may disagree with this post by citing the following advantages of incumbent SaaS companies, such as 1) brand and trust, 2) user base, 3) distribution, 4) capital, 5) existing contracts, 6) user data, 7) proprietary data, and so on.

The above factors will give incumbents a slight edge over AI in the short run, and I’m sure smart execs can figure out a way to elongate their lifespan, but none of that matters with AGI.

Case in point - Oracle migration. For decades, people thought it was impossible to migrate out of Oracle. But eventually many companies ended up migrating out to new systems. All it took was Oracle’s customers deciding to do it, and having the economic incentives to switch. And this didn’t even require AGI.

If something as difficult as Oracle migration can be done at the enterprise level, then it’s not that hard to imagine companies switching out of Asana, Salesforce, Monday.com, etc - especially with the help of AGI.

Something to think about.