An Unusual Phenomenon
Imagine you bought a Tesla Model 3 three years ago for 300,000. Now a more advanced Model S is on the market—your Model 3 should have depreciated significantly, right? That's how the used-car market works.
But in the AI data center world, the opposite is happening.
In 2024, rental prices for Nvidia H100 GPUs fell steadily, bottoming out around the time DeepSeek R1 made its stunning debut. Then, after December 2025, the trend suddenly reversed—four-year-old H100 cards are now worth more than they were three years ago.
This isn't financial market manipulation. It's a textbook case of scarcity meeting exploding demand.
Chip Shortage: A Perfect Storm
Supply side: Nvidia's next-generation chips are more powerful, but production can't keep up. TSMC's advanced-process fabs are already running at full capacity, yet global demand for AI chips is a black hole that never fills.
Demand side: The rise of DeepSeek R1 and other inference-focused models changed everything. Previously, we mostly used AI for training—one-off, large capital expenditures. Now, inference (actually having AI answer questions and generate content) has taken center stage. That means data centers need to run these GPUs 24/7, instead of shutting down after a training run.
Picture a gym that used to serve members occasionally now operating as a nonstop factory assembly line. Equipment demand is on a completely different order of magnitude.
Why Are Old Cards Worth More?
A counterintuitive economic principle is at work. Traditional hardware depreciation logic has been upended. Usually, electronics lose value fast—last year's iPhone feels outdated this year. But in the current AI boom:
Good-enough performance > newest model
The H100 launched in 2022, but it still excels at inference workloads. For running large language models to answer user queries, the gap between the H100 and newer chips is far smaller than the price gap between them.
Available now > waiting for new hardware
If you want to build a data center today, ordering the latest B200 or next-gen chips could mean waiting months or even a year. But if you can get H100s now, you can start earning revenue tomorrow. In business, time is money, and 'usable right now' often beats 'better but later.'
It's like choosing between a high-speed train departing in an hour (older model, available now) and a maglev leaving next week (newer, but you must wait). If you must arrive tomorrow, which do you pick?
A Fundamental Shift in Business Models
This phenomenon is reshaping how the entire data center industry thinks about asset valuation.
Historically, data center assets declined in a straight line: buy servers, depreciate 20% per year, and after three years the book value is less than half. That's how investors and CFOs did the math.
But the H100's 'reverse appreciation' suggests that in the AI era, hardware asset value curves may be U-shaped—or even continuously rising.
What does this mean?
For data center operators, holding GPUs is no longer 'burning money on equipment'—it may be 'value investing.' Your old cards may not depreciate; they may appreciate due to supply-demand imbalance.
For investors, an AI data center asset revaluation is underway. Companies holding large H100 inventories may be undervalued on their books.
The Deeper Lesson
This case reveals a broader truth: in eras of rapid technological iteration, we tend to overvalue 'new' and undervalue 'good-enough and available.'
The H100 is like a Swiss Army knife—not the latest model, but it can already do plenty, and it's in your hand right now. When everyone urgently needs tools, whoever holds tools is the richest person.
It also reminds us: when forecasting technology trends, don't look only at performance specs. Supply-demand dynamics, timing windows, and ecosystem maturity often determine value more than raw compute numbers.
What Happens Next?
No one can say whether H100 prices will keep climbing. If Nvidia solves its capacity problems, or competitors (AMD, or even cloud providers building their own chips) ship good-enough alternatives, the supply-demand balance could break.
Until that day arrives, the legend of the four-year-old card continues. It's a vivid lesson in scarcity, timing, and technology adoption curves—unfolding right before our eyes.
If you're considering entering AI infrastructure, remember this lesson: sometimes the most valuable technology isn't the newest, but the one that works at the right time and in the right place.
---
Original post tags: easy-learn-ai, daily updates, memory, Xiao Kai, AI infrastructure, GPU, data centers