AI Innovates Less Than You Think: C3.AI (AI)
Image - Dall.E 3. Prompt: “Artificial intelligence being confused with question marks.”
AI wanders down the long road to valuation nowhere
“Software is eating the world” was the 2011 line that made tech investor Marc Andreessen infamous. The man’s probably still correct. Except today, AI-enabled software is eating itself.
Artificial intelligence is driving markets into dangerous terrain. Web lists of AI stocks worth selling draw serious audience engagement. Estimates are 40% of globel fund managers sense AI stocks are “in a bubble.” Equity valuations tied to irrational AI exuberance are setting new highs. “Little wonder unease is settling over markets,” was The Economist’s pithy summary.
All of which is leading AI down a long and winding road to valuation nowhere. Take C3.ai (AI), the Redwood City, Calif.-based enterprise AI company that features heavyweights like Condoleezza Rice on its board.
Despite growing sales, operating cash flow and cash flow margins have steadily declined. Based on the company guidance, revenues would need to grow at 20% annually through 2028, just to keep the implied stock value inline with today’s share prices. That has been enough for many to become bearish on the company, with sellers outpacing buyers for a little over a year.
But such AI hand-wringing overlooks that disruptive digital markets are now more than two decades old. The base valuation narrative is the Recording Industry Association of America’s U.S. Music Revenue Database which shows, in full economic color, the grisly story of wave after wave of seemingly innovative bits of software consuming previous iterations of the same idea.
It is visualization worth memorizing.
The RIAA’s U.S. Music Revenues by Format, dataset is the benchmark reference for new digital technologies replacing legacy tools. Sales of blue LPs are to the left; orange CDs are in the middle. green digital platforms are to the right.
Revenues have been adjusted for inflation.
The blue columns on the left are the sales from lucrative physical record albums from the 1970’s and 80’s. The orange bars are the ensuing waves of yet-more-profitable CDs, that powered growth through the 1990’s. The wee purple bits were the short-lived paid-downloads, mostly on Apple iPods. That is followed by the various green waves digitally delivered music platforms like ad-supported and subscription-based streaming services Pandora and Spotify.
Just as important as the colored bars, are the empty white spaces on the right of the graphic, which indicate the opportunity cost of lost sales. If the music industry had grown as it had, estimates were it would have been a $50 billion business by 2024. Instead, the mismanaged transition to digital tools mired the music business in a mere $16 billion in total inflation-adjusted revenues.
As lean as digital revenues might be, margins are even thinner. When Sony Music Entertaiment Group posted record revenues, the gross profits from those sales were a fraction of what they were back in 1990. Same goes for streaming giant Spotify, which does show reasonable sales growth over the years. But its margins on those sales averages out to about 2.5%. That’s about even with inflation.
A CD from a savings and loan would have performed better.
AI Hears the Music
Given what investors know about the music industry, AI becomes nothing more than one low-margin digital technology replacing another, say like streaming music replacing paid downloads. AI-powered tools promise improved efficiencies and greater audience engagement. Sure enough, the charts in this story were not developed with the automated tools we used to create ourselves. Instead, these visuals were created from an intriguing Canadian AI financial analyst tool called FinChat AI.
Charting and analysis that used to take a few hours now takes a few minutes.
FinChat is but one of an entire stack of media-based AI tools that essentially automate the newsroom process. The list is so long now, it can’t be listed here. The dozens of titles simply can’t fit.
However, the history of digital music tells us that the efficiencies captured will be minimal and profit margins will be thin. Sure enough, that is exactly our experience. None of these tools, at least in our testing, were perfect. Most were excellent for developing ideas. But for actual fact-checked, ready-to-publish material, each chart had to be confirmed by hand. As easy and fun as it is to have FinChat chart-up the return-on-equity versus operating cash flow margins for Spotify, unless one understands what both mean, terrible mistakes await.
FinChat.io does an excellent job of automating financial analysis and charting, It essentially the automated tools we would create in-house.
“If it’s anything important, it’s the same process of validation and testing,” said one AI developer. “No matter how efficient the AI is.”
AI Devalues AI
AI is becoming a powerful tool for quantifying how deeply AI will devalue itself. Take something called Exploding Topics. This San Francisco-based AI startup tracks what’s trending online. The tool is fun and insightful on a full spectrum of topics, including AI itself. Exploding Topics currently estimates there are 70,000 companies involved in AI globally. That number seems impossibly large, but it is also impossible to confirm. Could tens of thousands of operations be tinkering with AI globally? Possibly. But assuming just 3 people are working on each idea – which is what the least number we know of needed to develop an AI tool – the world is dealing with an AI hive-mind workforce of about 210,000.
That’s about the size of Bank of America.
Let’s assume that estimate is half wrong, and only 100,000 AI practitioners are developing machine automation. Even that number represents a simply massive overcapacity for the artificial intelligence labor force. Anecdotal reports over the past 6 months of too many AI teams chasing too little work have been increasing.
AI tools do a great job documenting how hard it is for AI companies to gain attention. Exploding Topics creates a public list of the 24 Fastest Growing Companies. AI firms are rare. The truly popular companies offer network security, sell baby bassinets with built-in monitors, and operate in the “Not Suitable for Work” media market, which is what today’s adults call “Adult Entertainment.”
The fictional all-ruling, all-powerful AI-driven robots of Skynet and Black Mirror do not seem near at hand.
Unproductive AI
AI’s similarity to the music industry is revealed in how it is duplicating the music industry’s deteriorating productivity. Recent data from the Office of Productivity and Technology reported that, despite more than 3 years of AI deployment, total factor productivity in 2023 grew at about the same rates as it did back in 2007.
Total factor productivity has remained in line with figures from 2007. That’s surprising considering artificial intelligence has been quietly deploying in business settings for that entire period.
Source: Office of Productivity
And that when those productivity figures are walked back in time, basic productivity has remained at the same levels as it was from 1947. That could change. But as of now, there is little macroeconomic evidence that machine automation is making society more productive.
The total change in productivity for the current cycle starting 2019 is well below the average from 1947 to 2023. The economy would need to be twice as productive to match gains in 1947 to 1973 and 2001 to 2007. | Source: Office of Productivity
Image: Output of Dall.E 3 via Microsoft CoPilot from the prompt: “C.3 being confused by what it is worth with question marks.”
The Long and Short of AI
But yet again, AI is proving to be just like in the music industry: Operational woes will do nothing to dampen user interest and engagement. C3 was proclaimed one of America’s fastest-growing companies. The market share for the verticals it serves, like healthcare, are entering exponential growth. It is set to sell AI systems in America’s security industry. This is an operation with the right clients and access to capability..
How many investors will notice how much smaller and nastier the AI market becomes as machine automation tackles glamorous problems like gene sequencing and molecular chemistry? The AI market is already quietly consolidating. News startup Artifact recently sold itself to Yahoo!, despite 150 million users and a tech braintrust that included the founders of Instagram.
In the end, AI companies like C3 will neither be a boom nor a bust. The upsides and the downsides will balance themselves out. There will be none of the dramatic narrative swings that capturing value requires. AI will be the same old song about one digital tool making a meal of another.
It’s just these days, there is less analog gristle on the bone to sing about.