By Ramakrishnan J

There seems to be a story that keeps repeating itself in the world of technology. A new invention arrives, everyone is convinced that this will change everything. Money pours in faster than the world can absorb it, demand skyrockets, and then everything crashes. 

We’ve seen it with the railroads during the 1840’s, the rise of the internet in the late 90’s and with the spike of Artificial Intelligence in 2026. It is fundamentally changing how we see and interact with the world. 

What makes this current moment so strange is that money is flowing in loops. Everyone seems to be selling something, to everyone else, using the same pool of money, and calling it growth. 

Where it all started

In 1999, a company called Cisco Systems was the most valuable company in the world. It made the routers and switches that the internet ran through. As the dot-com boom exploded, every startup, every telecom, every enterprise scrambled to buy Cisco’s hardware. 

Cisco was also lending money to those very same customers so they could afford to buy Cisco’s products. Telecoms were borrowing from Cisco, spending it on Cisco equipment, and booking it as revenue. When the bubble burst in 2001, Cisco wrote off $2.2 billion in inventory in a single quarter. 

Why does this matter? Because in 2025, Nvidia did something similar. The world’s most valuable semiconductor company — the maker of the GPUs that power every major AI system on the planet — agreed to invest up to $100 billion in OpenAI to help finance a massive buildout of AI data centers. In return, OpenAI committed to purchasing millions of Nvidia GPUs for those very same facilities. Nvidia was, in the most literal sense, financing its own future sales.

If that sentence reads strangely, you are reading it correctly.

Analysts have a name for what is happening right now in AI financing. They call it circular financing. 

Money, computer chips, and cloud credits are rotating in a closed loop among a small constellation of companies: Nvidia, OpenAI, Microsoft, Oracle, AMD, CoreWeave, and a handful of others. The arrangement is so circular that dollars spent by one player often returns as revenue for another. It gives the impression of growth when their actual numbers for their company’s growth is abysmal. 

OpenAI reported $4.3 billion in sales in the first half of 2025, which sounds gigantic. It burned through $2.5 billion in that same period and is not expected to be cash-flow positive until somewhere near the end of the decade. Meanwhile, Microsoft has invested over $13 billion into OpenAI. Microsoft is also one of OpenAI’s primary cloud customers. OpenAI’s revenue flows back to Microsoft as Azure compute costs.

Why Your GPU Costs So Much Now

As you may have noticed, if you have tried to buy a graphics card or upgrade your RAM recently, the prices are not what they used to be. The 32GB DDR5 RAM kit that cost around $90 in early 2025 now approaches $200. The GeForce RTX 5090 could hit $5,000 on the secondary market later in 2026. 

The reason comes down to memory.

Memory costs now account for nearly 80% of a GPU’s total manufacturing cost. AI workloads are memory-hungry in a way that consumer hardware simply cannot compete with. High-bandwidth memory used in AI servers consumes roughly three times the wafer capacity of standard DDR5. Samsung, SK Hynix, and Micron have pivoted their production lines toward the far more lucrative AI market.

AI is expected to consume 20% of total global DRAM production in 2026. Hyperscalers — Google, Microsoft, Amazon, Meta — have signed multi-year contracts reportedly locking up 40% of global DRAM supply. The CEO of Silicon Motion put it plainly: “We’re facing what has never happened before: HDD, DRAM, HBM, NAND… all in severe shortage in 2026.”

The consumer market is literally deprioritized. Nvidia has already announced plans to reduce RTX 50 series production by 30 to 40% compared to the first half of 2025 to cope with the shortage of memory components. 

Circular transactions create the appearance of demand that may not exist independently. When the companies investing in AI infrastructure are also the primary customers of that infrastructure, and the chipmakers funding them are also selling to them, the revenue numbers start to look less like evidence of a thriving market.

UBS analysts estimate the Nvidia-OpenAI circular arrangement represents only around 13% of Nvidia’s projected 2026 revenue. Amazon just announced an additional €18 billion investment in Spain alone, pushing its total committed European data center spending to €33.7 billion.

These are not investments made in response to consumer demand. They are bets placed in anticipation of demand that the same companies are also trying to manufacture.

Whether this ends in a crash or a slow deflation or something stranger is unpredictable right now. Stocks, shares, oil, silicon. Everything is unstable.

We have built a world where the same company that makes the chips is investing in the companies that buy the chips, which are being funded by cloud providers that also sell computers to those same companies, which are raising money from investors who are also customers.