“Tokenomics” Equals Panic
It is possible to be a master in false philosophy, easier in fact, than to be a master in the truth, because a false philosophy can be made as simple and consistent as one pleases – George Santayana
Having lived through the insanity surrounding Y2K and valuations based on “eyeballs”, I can say with complete confidence that “Tokenomics” represents the same sort of false philosophy as the driver of the internet age. Tokenomics is an effort to justify an economic relationship that is meaningless when valuing companies, apart from the temporary beneficiaries such as Micron and Nvidia. So be it.
Tokenomics represents the panic borne out of the realization that you spent a fortune developing AI yet there is insufficient demand to generate a positive ROI. It’s an attempt to socialize the losses later this year when OpenAI and Anthropic come public combined with the need for the Trump Administration to show GDP expansion into the mid-term elections. It’s where political expediency meets financial necessity. I don’t plan to fight it.
Investments in Information Processing Equipment: Contribution to GDP Growth
Similar charts are available for software, data centers, and R&D. Machiavelli taught us that the end justifies the means. I suspect that we should be prepared for these market sectors to rally through year-end despite not making fundamental sense. Whether it’s stories of government investment in the frontier-model companies, mysterious studies that show potential positive ROI, or changing the accounting for model-training to capitalize costs as long-lived assets like electric utilities.
What is “Tokenomics”
I love the smell of napalm in the morning – Colonel Kilgore (Robert Duvall) Apocalypse Now
An AI token is a word and words are the basic building blocks of Large Language Models or LLM’s. “Tokenomics” is simply a top-down incentive system to drive software developers to maximize their use of tokens or units of LLM capacity. It’s an artificial attempt to force Jevon’s Paradox which simply states that increasing the efficiency of a new technology or resource often increases the overall consumption of that technology/resource.
The top tech companies are betting the ranch on forcing AI onto the software developers out of fear of being left behind. Companies such as Microsoft, Amazon, Uber, and Meta are creating internal incentive systems and outright directions for software developers to utilize internal AI tools.
It’s a last-ditch effort to make AI relevant and a “Hail Mary” pass to increase the revenues of frontier-model companies such as Anthropic and OpenAI ahead of IPO’s later this year. These companies are burning capital like it’s gasoline yet they need something to support the insane valuations “assigned” to them.
And it’s pretty clear that these “assigned” valuations are at the heart of the recent rally in AI stocks. One example is the 81% increase in Google’s earnings. But if you look closely, roughly half of Google’s earnings were non-operating, consisting of rising equity valuations on non-marketable equity securities such as Anthropic. Amazon reported similar results.
Is this a sign of success? Or is it a desperate attempt to create the impression that the frontier-model companies are proving to be good investments for the top tech companies instead of capital destroyers.
You can’t argue with the successful moves in the stock market but you can question their sustainability. My personal opinion is that we’re witnessing the “Swan Song” of the technology companies even as they hit absurd valuations in their last show.
I don’t believe it is sustainable because there is still no credible evidence that using AI results in a positive return on investment apart from some bottoms-up applications. When the fundamentals ultimately assert themselves, someone is going to be on the receiving end of a napalm bomb. The only question is whether Wall Street can socialize the losses ahead of time by selling OpenAI and Anthropic to the public. Regrettably, I believe OpenAI, Anthropic, and SpaceX will be jammed into index funds before the end of the year.
Damned If You Do…
The hasty stroke goes oft astray – JRR Tolkien
Major companies such as Meta and Amazon have created incentive systems for software developers to “token max” which basically incentivizes developers to waste as many “tokens” as possible while performing their jobs. Uber used up their 2026 budget for AI tokens within the first four months of this year. To date, there is zero evidence that maximizing tokens offers a positive return on investment despite numerous studies.
The problem for the tech industry is that in order to make Anthropic and OpenAI look good before the IPO’s, they have to embrace “Tokenomics”, pay a fortune in AI usage, which will negatively impact their own earnings. It’s not a coincidence that Microsoft, which owns 49% of OpenAI, told developers that they can no longer use Claude (Anthropic), which was formerly cheap, now that they are being billed for each token used. Those funds have to be kept in-house.
The tech industry is facing an open-ended cost problem with the adoption of token-based pricing models. But if the industry does not embrace “Tokenomics”, there is no way for OpenAI and Anthropic to report anything other than disastrous financials ahead of their IPO’s.
We would have to accept that the historical costs of training models represent a long-term asset on their balance sheets instead of a cost of doing business. Because Microsoft owns 49% of OpenAI, we know they have been consolidating their share of expenses to date so it would take a major change in accounting standards to reverse these losses and capitalize them into assets. Be prepared, it could happen.
Regardless, the cash flow returns are horrendous with no end in sight. Hyperscalers such as Amazon, Google, Microsoft, and Meta are looking at $700 billion in capital expenditures in 2026. Total worldwide spending is expected to reach $2.5 trillion.
As I predicted back in September 2025, these companies are using most of their internal cash flow to fund AI and even that isn’t enough. Google/Alphabet is discussing selling $80 billion in new equity shares to fund their commitments. Meta is also rumored to be considering equity sales. These companies have gone from being cash flow machines to cash beggars seemingly overnight.
It's why the market has to operate as a mania because by any objective analysis, we’re witnessing total insanity. Data centers, which are the base of AI, are so unprofitable that lenders are recording dramatic losses. Frontier AI models are insanely expensive to train and have not yet proven that there is a market for their products that will result in positive ROI’s.
The companies that sell the physical technology for data centers are making fortunes even as questions about whether the chips are presently being used in AI operations. Both Nvidia Blackwell GPU’s and Micron RAM memory may be sitting in storage even as data center buildouts slow due to energy/water/financing/commercial demand problems.
But none of that matters because the US economy currently depends on the AI mania and we have a highly contentious election coming up.
Mid-Term Elections
Men are so simple and yield so readily to the desires of the moment that he who will trick will always find another who will suffer to be tricked – Niccolo Machiavelli
Large Language Models are really good at one thing – aggregating information and drawing inferences from that information. They were never designed to be the Oracle on Mount Olympus of Greek mythology; they were designed for predicting outcomes on narrow questions. I suspect the ultimate purpose was Orwellian social control. Why else would the greatest concentration of data centers be in Loudon County, Virginia? It’s a suburb of Washington DC.
According to Marc Andreeson, a long-time venture capitalist in Silicon Valley and a life-long Democrat, he was told by the Biden Administration that AI would be limited to four or five companies – companies that would administer the system. We can readily guess these companies – Google, Microsoft, Meta, Amazon, and possibly Apple and Oracle.
But something funny happened on the way to the forum – the US electorate started to move in the opposite direction even as the magnates of Silicon Valley changed sides so they too could get a cut of the action. Presumably, major tech companies are betting that the original purpose can be revisited when Trump is out of office, otherwise, further investment is insane.
I understand the Administration’s angle because it’s twofold: make sure 401k’s are rising in value into election day and make sure new semiconductor plants go up all around the US. Trade barriers can be used to squeeze Taiwan and South Korea in the future to make domestic chip plants profitable. It’s why I’m not a fan of Taiwan Semiconductor.
Technology companies represent roughly 50% of the S&P500’s market capitalization when you include Amazon (Consumer Discretionary) and Alphabet (Communication Services) in the measure. The White House is going to place a heavy emphasis on maintaining stock market returns into the mid-term elections and that means supporting AI.
Conclusion
I am the spirit that negates!
And rightly so, for all that comes to be
Deserves to perish wretchedly – Mephistopheles from Faust - Goethe
Reported to be Karl Marx’s favorite quote, I view the rush to over-invest in AI as an omen of capital destruction. As I wrote earlier in this piece, I view AI as the Swan Song of the major technology companies. Nobody stays on-top forever and Microsoft, Google, Meta, Tesla, Amazon, and yes, Nvidia are all headed for a reckoning in the future.
We can already see the cracks – the need to sell equity, the rise of alternatives, and the decline of the utility of core products. Google searches have become crap, Microsoft Office is slow and intrusive, Musk can’t figure out what products he wants to sell or what planet he wants to operate on, and Amazon has to compete with Google and Microsoft in the wholesale market. Nvidia chips are like Range Rovers – cost a fortune but only functional for a few years. Micron has the right commodity product at the right time similar to Y2K.
The futures of OpenAI and Anthropic seem destined to be as regulated utilities but not at $1 trillion each. SpaceX seems destined to be merged into NASA in future years unless space becomes economically viable – SpaceX has the can-do attitude that once dominated NASA before it turned to bureaucracy. These predictions mean nothing over the next six months because the crazy daisies are in full-bloom.
Technology’s share of our nation’s market capitalization already exceeds Y2K levels by a wide margin but that doesn’t mean it can’t expand further. The Trump Administration needs mania and optimism into the mid-terms. Watch for extreme patriotism to be kindled into our 250th anniversary next month and combined with investing in technology, particularly domestic technology.
If Y2K taught us anything, it’s that companies can rapidly destroy themselves even as the market cheers them on. AOL, Worldcom, Enron, and Lucent all enjoyed a wave of optimism as the Grim Reaper was sharpening his scythe for them.
AI has been over-invested and the fallout is going to be brutal for investors probably starting next year. If OpenAI and Anthropic are successfully forced into index funds, the losses will be borne by everyone.
“Tokenomics” is going to prove to be a failure because the pricing isn’t worth the value AI creates. Over the longer term however, if ultimately priced at the margin, smart users will figure out ways to make it useful – perhaps even identifying the extraordinary waste in our government spending models. That alone would justify the investment!
There are other factors at play which could have a bigger impact on the markets, the main one being the Trump Administration versus global central banks but that’s a topic for another day. If you’re interested in learning more, visit us at https://geovestadvisors.com/ and drop us a line.
Philip M. Byrne, CFA