Happy Saturday, one and all!
Brian McGleenon here,
If you are worried about your investment portfolio and how it would fare against an increasingly precarious AI bubble, let’s look at the raw data. Since the launch of ChatGPT in 2022, combined AI-driven capital expenditure by the sector’s largest players has grown at a staggering average of 72% per year. Their collective annual infrastructure spend is projected to top a historic $725 billion this year alone.
Right now, a massive amount of hope lies on a future where “agentic payments” become the norm, with institutions from investment banks to major retailers rushing to upgrade their back-end systems to fit this new AI paradigm.
But there is a staggering investment imbalance hiding in plain sight.
This blind spot was brought to light for me this week during a conversation with Sandy Kaul, Head of Innovation at Franklin Templeton. She explained how the investment imbalance between Artificial Intelligence and blockchain infrastructure is sitting at a massive 10:1 ratio. While crypto and blockchain startups brought in a highly respectable $20 billion over the last year, Wall Street and Silicon Valley went absolutely ballistic on the other side of the fence, pouring over $200 billion into AI.
Capital is being thrown at anything with an .ai extension. Yet, the foundational architecture actually needed to run agentic payments, such as smart contracts, tokenization, and blockchain rails, is sitting quietly in the corner, severely overlooked and undervalued.
Speaking on our latest BeInCrypto Expert Council podcast, Sandy Kaul pointed out this exact blind spot—one born from pure AI FOMO. As institutional investors pour billions into agentic transaction startups, they rarely stop to ask a deceptively basic question: How are these autonomous agents actually going to pay for things?
An AI agent isn’t going to open a traditional bank account, sign a paper check, or wait three business days for an ACH transfer to clear. They require 24/7, instant, programmable value. That means these new payment rails will be fundamentally built on public blockchains, utilizing smart contracts, stablecoins, and tokenized real-world assets.
However, the message is finally starting to click for institutional investors: to power this agentic payment paradigm shift, the financial world is going to need open, public networks like Ethereum. As Sandy put it:
“Most people think they hear private networks permissioned, that they think safe, right? Safe because it’s permissioned, it’s private... But if you compare that to a public blockchain where you might have 3,000 or 4,000 developers working on the chain, and where the chain itself is paying bug bounties for any of those developers to find any flaw in the system, you get a much more scrutinized platform from which to operate.”
She compared this shift to the early days of cloud computing, noting that financial institutions swore they would never put data on the public cloud, only to eventually move everything there.
The real magic happens when you couple that security with the 24/7 nature of smart contracts. For instance, Sandy explained how their Benji fund breaks the traditional “close of the trading session” model to cater to institutional capital moving at the speed of modern tech:
“Today if I sell out of a fund at midnight and I sell out of a fund at noon, I would get none of that day’s interest in the other tokenized money market funds or in a traditional money market fund because I did not own that fund at the close of the trading session. Whereas with our system, because there is no close of a trading session, you owned it for twelve of the 24 hours, so you would get half the day’s interest. This is a very important offering if you’re a corporate treasurer or if you’re a hedge fund who is moving dry powder capital intra-day.”
You can catch my entire, wide-ranging conversation with Sandy Kaul on the BeInCrypto Expert Council podcast right here:
While blockchain is quietly building these hyper-efficient, fractionalized yield systems, what is happening over in the AI sector?
BeInCrypto’s Lockridge Okoth published a brilliant, sobering investigative piece this week that completely pulls back the curtain on Big Tech’s $2 trillion AI gold rush.
Read the full news: https://beincrypto.com/ai-boom-fake-revenue-bubble/
Lockridge argues that the entire boom is hiding a severe structural flaw, pointing to a “round-trip funding loop” that feels dangerously reminiscent of the 2001 dot-com crash.
Think about it: who is actually buying AI en masse right now? Over 90% of OpenAI’s user base is on the free app. Instead, as Lockridge exposes, the tech giants are essentially quietly paying themselves through their own cloud bills.
Here is how the loop works:
The Investment: A tech giant writes a billion-dollar check to an AI startup.
The Catch: The investment lands largely as cloud infrastructure credits.
The Revenue: The startup feeds those credits straight back to train its models, and the tech giant books that consumption as fresh commercial revenue.
Lockridge highlighted that OpenAI’s annual cloud bill has ballooned past $60 billion, while its actual revenue sits closer to $25 billion. Anthropic is running the exact same play with Amazon, spending $2.66 billion on AWS in nine months, essentially every dollar it earned.
This creates massive paper profits on income statements. In Q1, Alphabet posted a record $62.6 billion profit, but $28.7 billion of that was just a paper markup on its Anthropic stake. Amazon did the same trick, marking up $16.8 billion, even though its actual free cash flow cratered 95% to $1.2 billion because they are pouring billions of real cash into physical data centers.
The wall comes when AI hits the budgets of ordinary companies who can’t recycle infrastructure spend into revenue. Lockridge uncovered some staggering enterprise data:
Uber completely torched its entire AI coding budget by April because engineers using Claude Code and Cursor were burning up to $2,000 each per month in API charges.
Microsoft literally ordered its own employees to stop using Claude Code internally because the token consumption became unsustainable.
Even Nvidia’s VP of Applied Deep Learning, Bryan Catanzaro, admitted his team now spends far more on compute costs than on human salaries.
Have a great weekend,
Brian McGleenon,
Global Head of News, BeInCrypto
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.