Why the Math of Replacing You Doesn't Quite Add Up: A Filmmakers Attempt to Understand the Financials (and Bubble) of AI
This is not financial advice, nor should you take it from someone like me.
If you want to support my independent reporting, subscribe to Cinema Ex Machina for free. You can also become a supporting member or you can leave a tip. 🍻 Cheers!
I think like everyone else out there, I have my moments of dread where I'm terrified AI is going to take everything I love about my job away from me. And it might! That's the marketing pitch anyway.
With that said, I'd like to introduce you to a writer I've recently discovered named Ed Zitron. He helps with the dread because he's learned to speak a language I don't understand but is the crux of everything in our little capitalistic lives: finance. He's done PR for AI companies, and as of recent has taught himself how to understand and explain these gigantic deals the AI companies are doing.
I'll let Ed introduce himself:
I may not be a contrarian, but I am a hater. I hate the waste, the loss, the destruction, the theft, the damage to our planet and the sheer excitement that some executives and writers have that workers may be replaced by AI — and the bald-faced fucking lie that it’s happening, and that generative AI is capable of doing so. - Ed Zitron
I like Ed, and with any one source of information I encourage my dear readers to take Ed with a grain of salt and poke holes in his (and always my) arguments. Overall, I'm finding Ed, at the very least, to be a comfort in the AI hype machine that seems endless.
The problem is, if Ed's right, the American economy comes crashing down. It's a weird world we live in to find comfort in that.
The Bubble
To attempt a sum up of Ed's lengthy reporting and writing: he's basically saying the financials of the AI industry don't pencil.
Let me actually start this by saying I'm not a financial reporter. Not even close, I'm just a filmmaker trying to understand what's going on. Email me, hit the contact page or socials - I'd love to hear what you're seeing and what I'm not seeing. I digress.
The magnificent 7 (NVIDIA, Microsoft, Alphabet, Apple, Meta, Tesla and Amazon) make up about 35% of the value of the US stock market. Your retirement plan is certainly based on these companies continued growth. From there, Microsoft, Amazon, Meta, Alphabet, and Tesla themselves make up around 42.4% of NVIDIA's revenue through the sale of GPU's. And this is where things get scary.
The companies with AI products are not making money, and in fact, they're losing billions of dollars on whatever revenue they do make. The incredible stock growth that we see in the markets, as Ed says, is "vibes based":
So, really, everything comes down to NVIDIA's ability to sell GPUs, and this industry, if we're really honest, at this point only exists to do so. Generative AI products do not provide significant revenue growth, its products are not useful in the way that unlocks significant business value, and the products that have some adoption run at such a grotesque loss.
So, our current economy depends on 5 companies spending money on NVIDIA GPU's. If there is any wobble with these investments, and especially with NVIDIA itself, Americans are going to feel the brunt of a market downturn. All on a bet that AI will reshape the economy. How long can these companies invest in AI at a loss, therefore propping up the stock market? Will the demand for AI be there to make it profitable?
The AI Data Center Crisis
Since the beginning of 2023, big tech has spent over $814 billion in capital expenditures, with a large portion of that going towards meeting the demands of AI companies like OpenAI and Anthropic. - Ed Zitron
That's almost a trillion dollars folks. With a crazy dependence on debt to support companies without a lack of meaningful revenue and no proof that they'll start making meaningful revenue: the whole industry seems totally bonkers, according to Ed’s reporting.
To build a data center requires debt. To service that debt, you need paying tenants. Open AI has cut a deal with folks like Oracle to build these centers and run their compute.
Even if it can get the lousy 30.9% margins I believe it's making on Stargate Abilene, the costs associated with building, maintaining and running it are likely to swallow either Oracle or Crusoe whole far before the completion of the whole data center complex. Oracle can try and cover the losses by selling more shares and raising more debt, but at some point it'll come down to whether OpenAl can afford to pay its bills, which comes down to whether venture capital and private credit can afford to keep OpenAl afloat.
Relatedly, I also thought this interview with Ed was interesting:
And from 404 Media, surfacing Pew polling: People Hate Data Centers.
Most of the Americans surveyed believe that datacenters are bad for the environment, home energy costs, and the quality of life of people living nearby and the numbers aren’t close. Only four percent of people thought datacenters were good for the environment, six percent good for jobs, and six percent good for people’s quality of life. - 404 Media
Not only is the bet on AI gigantic with major risk in supporting these unprofitable companies, the cherry on top is: Americans have no love for the lowly data center.
This whole thing doesn't totally seem to be working exactly how the marketing is telling you it is.
These Are Not Profitable Companies and Ed's Not Sure They Ever Will Be
I'm going to rip this chunk from Ed's site directly as it is fascinating, and I quote:
Here are the projections that CFO Sarah Friar signed off on:
- In 2026, OpenAI will make $30 billion in revenue, or just under the $34 billion that Salesforce made in 2024.
- In 2027, OpenAI will make $60 billion, or $2.6 billion more than the $57.4 billion Oracle made in its Fiscal Year 2025, or $8.4 billion more than Broadcom made in 2024.
- In 2028, OpenAI will make $100 billion, or $2.31 billion more than Tesla made in 2024, or $11.66 billion more than TSMC — the single-largest chip manufacturer in the world — made in 2024.
- In 2029, OpenAI will make $145 billion, or $14.5 billion dollars more than the $130.5 billion that NVIDIA made in its Fiscal Year 2025**.
- In 2030, OpenAI will make $200 billion, or $35.5 billion more than the $164.50 billion that Meta made last year.
Just so we are clear, OpenAI intends to 10x its revenue in the space of four years, selling software and access to models in an industry with about $60 billion of revenue in 2025. How will it do this? It doesn’t say.
Endquote. There's plenty more reading to be done here but this all a little bonkers isn't it? With these kinds of numbers, and the impossible hill ahead to hit them, AI no longer necessarily feels absolutely inevitable right? Right?!
Edward Zitron
Edward Zitron
Edward Zitron
It All Comes down to Demand for AI
I'm going to steer away from Ed for a minute, although he makes this point repeatedly. Open AI and Anthropic HAVE to make billions to stay in business and frankly, to keep the US economy afloat at this point. What this means is they have to create demand and make their product an inextricable part of the enterprise economy. Overall, that's not happening. Right now, again, it all sounds like hype.
Let's start here, in Fortune, Thousands of CEO's just admitted AI had no impact on employment or productivity
A study published this month by the National Bureau of Economic Research found that among 6,000 CEOs, chief financial officers, and other executives from firms who responded to various business outlook surveys in the U.S., U.K., Germany, and Australia, the vast majority see little impact from AI on their operations. While about two-thirds of executives reported using AI, that usage amounted to only about 1.5 hours per week, and 25% of respondents reported not using AI in the workplace at all. Nearly 90% of firms said AI has had no impact on employment or productivity over the last three years, the research noted.
And on to Forbes:
The bottom line is stark: AI delivers measurable productivity gains of 14-55% at the task level—yet 95% of enterprise AI pilots fail, aggregate economic statistics show negligible impact, and a Nobel laureate projects only 0.5-0.7% total productivity growth over the next decade. This gap between micro-level wins and macro-level disappointment defines the central paradox facing investors, executives, and policymakers in 2026.
And from McKinsey with a subtle subtitle "Still more promise than profit"
While AI tools are now commonplace, most organizations have not yet embedded them deeply enough into their workflows and processes to realize material enterprise-level benefits... the transition from pilots to scaled impact remains a work in progress at most organizations
If You Wonder Why AI Is So Hyped and Why Everyone With Money Is Threatening to Take Your Job
It's because currently the American economy depends on it! The big AI companies need to find a use for their product that actually show gains for people's bottom lines to escape the amount of debt they've built. It's a scary situation to be in and it's something that'll affect all of us.
I'd like to thank Ed Zitron for keeping me grounded during any and all freak-outs I have about AI's destructive force in the creative industry. Thanks for tuning in, until next time.
