I just watched a great interview on CNN by Madison Mills from Axios. Madison is reporting on the ballooning costs of AI at corporations across the world, with few of them having a clear understanding for the ROI of those very expenses.
Check out the CNN (4) minute interview - then come back for my thoughts!
AI Cost & Start-ups: What Caught My Attention
As it relates to AI costing more than humans -- in the short term this was bound to happen in most cases. When you are moving at break-neck speed, despite your best efforts, you're inevitably going to pass right over a great many things. Inefficiencies will abound and corners will be cut.
However, companies who will succeed in integrating AI correctly, will take the time to create accurate KPIs and trusted methods to measure ROI to avoid this issue continuing over the long term. And this is actually, already happening -- more on this below.
As for Start-ups trying to brag about their AI expenditures - this is probably attention seeking behavior (notorious habit with start-ups) but it also could be cover for executives who have overinvested in a technology and have no clear ROI metrics with which to validate those same investments. This is simply trying to obfuscate (my favorite word) failure with distracting PR or industry accolades.
Eventually however, the chickens will come home to roost.
A Few More Things
Madison also mentioned some shocking data about token use:
- VP @ Nvidia - costs for AI are outpacing human costs
-
CTO @ Uber - already blew their entire 2026 budget for the year via AI expenditures
Add this to Nvidia CEO Jensen Huang's potentially self-serving take on token use:
"If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed." - Jensen Huang, CEO Nvidia
The AI Spending Flywheel
So what to make of all this?
Well at some point, the hyperbole and public positioning are going to give way to market and financial realties.
I believe in the AI Spending Flywheel that we find ourselves in, we're currently at the tail end of Stage 3 and on our way to Stage 4.
Precursor | Step -- AI SPENDING FLYWHEEL
What Triggers Stage 4?
In my opinion, Stage 4 will be in full swing when we see the largest investors into AI start to pull back on their token use and leverage that pull back to negotiate better costed deals with AI firms.
This has 100% started already, and will simply increase over time.
And in part, this is inevitable with the continual increase in cost of RAM and electrical power for data centers. Perhaps Elon can help with Terafab on the RAM side of things -- SpaceX + Google with space based Data Centers.
But in the short term, cost for AI will continue to increase as the AI Firms have to continually pass on their ever-increasing delivery costs to consumers.
FillmOracle Prediction One: Corporate AI Budget Management Changing
As the euphoria of AI investment naturally gives way to rational economic pressures, more companies and executives will stop blindly spending on AI and require their leadership to apply normal budgeting restrictions on technology investments, but -- NOW include AI.
They'll use a decision tree, something like this:
AI Budget Allocation Decision Tree
FillmOracle Prediction two: Cost Pressures Could Result in AI Restrictions
Something I would watch out for is how executives at AI firms get creative with costs. Could we see a day soon when AI platforms implement wider content restrictions on memes, re-generations and upscaling's for categories of content they deem "not worth wasting tokens on"?
If you're shaking your head and saying "that could never happen", I would just say in response that you shouldn't underestimate the corporate classes creativity to implement consumer unfriendly changes to benefit their bottom line.
Sources
