Over the last few weeks, and probably over any CNY gatherings for our friends who were celebrating, AI was definately one of the highlight topics.
And one AI company that took centerstage this time was China’s Deepseek.
In case you haven’t caught up in the latest gossip yet, here’s a quick few pointers on it:
-
Significantly Lower Costs: DeepSeek’s V3 model was trained over approximately 55 days using 2,048 Nvidia H800 GPUs, totaling around 2.8 million GPU hours. At an estimated cost of $2 per GPU hour, the total training expense was about $5.6 million—significantly lower than the hundreds of millions typically spent by competitors.
-
Efficient Training: By utilizing only about 2,048 GPUs, DeepSeek achieved a 95% reduction in GPU usage compared to companies like Meta, which often employ up to 16,000 GPUs for similar tasks.
-
Stock Market Impact: Following DeepSeek’s announcement, Nvidia’s stock experienced a significant decline, dropping by 17% and resulting in a loss of nearly $600 billion in market value—the largest single-day loss for any company on record.
-
Token Cost Comparison: DeepSeek-V3 offers a significant cost advantage over other AI models, namely OpenAI’s GPT-4. DeepSeek-V3 is nearly 55x cheaper for input tokens and over 27x cheaper for output tokens! (If “tokens” are new to you, think of them as LEGO bricks. Just like how we can use plastic pieces to bring our imagination to life, tokens are used for back-and-forth communications with the AI. For example, if you ask ChatGPT, “How’s the weather in Singapore today?” it will cost between 20 and 25 tokens.)
So what does all this mean for end users like you and me?
Simple. AI is about to go from a luxury to a necessity.
AI is only getting cheaper, faster, and more powerful.
“Wait Mav… if AI models are getting cheaper, does it mean this is an AI bubble? Is this the end of the AI hype?”
Actually, it’s the opposite. And I’m not saying that because I’m a maverick (bad pun, I know).
I believe the drop in cost will simply get even more people using AI.
Ok, let me explain.
While these AI developments are new, the phenomenon is not. And for that, we look at history again, at Jevons Paradox.
And by the end of this article, not only will you see proof that AI will spread like wildfire, but you will also learn some strategies to ride the wave.

Can’t argue with history.
What is the Jevons Paradox?
You’re probably thinking, “Wait, shouldn’t efficiency mean we use less of something?”
That’s what common sense says, but history says otherwise. Let’s go back in time.
-
Coal & Steam Engines (In the 19th Century) – More efficient steam engines should have reduced coal consumption. Instead, demand for coal skyrocketed because industries found more ways to use it.
-
Fuel-Efficient Cars – Making cars fuel-efficient should reduce gasoline use, right? Nope—people just drove more miles.
-
Cloud Computing – Storage costs plummeted, so instead of using less, we started hoarding data.
This is known as the Jevons Paradox, coined by a British economist and logician, William Stanley Jevons (1835–1882). In his groundbreaking book, The Coal Question (1865), Jevons argued that Britain’s increasing reliance on coal efficiency would not reduce consumption but instead accelerate it.
And as you can see from the examples highlighted above, his theory remains highly relevant today, especially in the context of AI and energy.
What’s the lesson here? The more efficient a resource gets, the more demand explodes. AI is following the same trajectory, and businesses that ignore this shift risk falling behind.

What has this got to do with me?
Jevons Paradox in AI
AI isn’t just improving—it’s becoming radically cheaper and more accessible.
This means:
-
Lower AI costs = Mass adoption across industries. Even small businesses can afford AI that once required enterprise budgets.
-
Smaller startups now have access to AI tools that once gave tech giants a competitive edge.
-
Open-source AI models like DeepSeek are democratizing AI, making it easier than ever for companies to integrate AI into everything.
Token Cost Trends
Here’s a quick look at the cost of tokens from one of the top AI labs, OpenAI, compared to Deepseek.

Note: The prices for are based on available data as of January 2025.
DeepSeek-V3 isn’t just cheaper—it’s disrupting the entire AI pricing model!
With or without DeepSeek’s existence, the fact remains true:
AI costs have nosedived by 100x in just a few years, making AI the biggest business enabler since the internet.
I mean, how often do you see the price of raw materials drop at such an alarming rate? We’re talking about a 90% drop year on year.
Plus, it’s not due to an oversupply: AI enablement is more in demand year after year. A quick look at AI spending over the last few years will confirm that (I wrote about it a few weeks ago in this article)

This move will change EVERYTHING!
The AI Revolution
We’ve checked the prices, and we’ve confirmed with history.
While it’s never safe to predict the future, there’s nothing wrong with being prepared for it.
Let’s lay out all the puzzle pieces here before we put them together.
-
Lower Costs = More Adoption – AI is no longer an elite tool for the big boys. It’s accessible to anyone with a phone (and on some AI models, you don’t even need the internet!). Businesses that don’t adopt it will be left so far behind it will take years to even catch up!
-
Easier Access = More Innovation – With AI available to everyone, expect entirely new industries to emerge. This means that even the most unlikely industries will be disrupted by the smallest players.
-
More Compute Power = Greater Demand – As businesses integrate AI, data center demand is exploding. You can call this the “network” effect we once see with fax machines and social media. In fact, verticals like data and energy are probably where the next wave of opportunities will come from.
-
Regulatory Lag = Unchecked Growth – Governments can’t keep up with AI’s rapid evolution, leading to an unregulated gold rush. It’s like the time your parents are out of town and you got the house to yourself for the entire weekend. Only this time, instead of scoring street cred with your friends, you can use AI to gift you a competitive advantage others would dream of.
Like the steam engine, internet and social media that came before us, AI will start a revolution, if it hasn’t already. Some day in the future, you and I will look back at this period of time, and realized we were at “ground zero”.
The question is, how will you and your organization take advantage of this?

Radiant Institute’s AI Advantage Framework
The AI Advantage Framework – Where Businesses Can Apply AI for Maximum Impact
“Where to start, Mav?”
Yes, that’s the million-dollar question I get a lot, especially after I’ve explained to clients the impact of AI.
The knee-jerk approach is perhaps signing up for a free AI tool and start using it. While I appreciate you doing that (because going from 0 to 1 is hard!), that’s like getting a random seed and throwing it on the ground, hoping that it will bear fruit.
That’s AI Adoption. It only works if you already have a good foundation within your organization.
Imagine having a garden with well-prepared soil, great sunlight, access to water etc. Your seed is going to grow up well!
That’s AI Enablement. Similar words, but big difference. I just wrote about it in last week’s #bebrilliantwithAI. Check it out here.
Now, let’s shift our perspective from ground-level (AI Adoption) to eye-level (AI Enablement). I’m currently working on a deepdive article on this framework, but let me give you a quick overview first:
The AI Advantage Framework highlights six ways businesses can turn AI into a competitive advantage.
AI Advantage #1 – Cost Reduction
💡 What it’s all about: AI streamlines operations, eliminates inefficiencies, and slashes costs—without cutting value.
🌟 Example: AI-powered logistics optimization cuts fuel and shipping costs with real-time route adjustments.
🧭 Keyword: Optimization
AI Advantage #2 – Revenue Generation
💡 What it’s all about: AI isn’t just about saving money—it’s about driving more revenue.
🌟 Example: E-commerce platforms use AI-driven recommendation engines to increase customer spending.
🧭 Keyword: Acceleration
AI Advantage #3 – Position Solidification
💡 What it’s all about: AI isn’t just a backend tool—it’s a frontline differentiator.
🌟 Example: Microsoft’s Copilot AI in Windows boosts user retention by embedding AI into daily workflows.
🧭 Keyword: Fortification
AI Advantage #4 – Market Expansion
💡 What it’s all about: AI enables businesses to analyze trends, predict demand, and expand into new markets.
🌟 Example: Nestlé uses AI-driven trend forecasting to accelerate product development.
🧭 Keyword: Opportunity
AI Advantage #5 – Cultural Transformation
💡 What it’s all about: AI eliminates busywork, freeing employees for high-value, creative work.
🌟 Example: Atlassian’s ShipIt Days encourage employees to experiment with AI, driving innovation.
🧭 Keyword: Renaissance
AI Advantage #6 – Future-Ready Resilience
💡 What it’s all about: AI strengthens decision-making, helping businesses stay ahead of disruptions.
🌟 Example: AI-powered financial modeling helps companies anticipate market shifts with confidence.
🧭 Keyword: Confidence
🚀 Want to go deeper? Stay tuned for our full breakdown of the AI Advantage Framework next week.

Great news! But at what cost?
The Hidden Cost of AI’s Growth
Nothing in life is truly free, and even “cheapness” comes with a cost.
Same goes with AI development. Yes, as an AI enabler, I am excited about how these tech advancements can mean for the community, my clients, or even selfishly, for Radiant Institute.
But the world works in balance, so let’s look at the setbacks of AI’s rapid growth over the last few years.
Infrastructure & Energy Costs
-
AI requires massive compute power, leading to soaring electricity usage and straining power grids.
-
Data centers consume enormous amounts of energy, sometimes rivaling small cities.
-
AI training and inference workloads increase cooling and water consumption in data centers.
Hardware & Supply Chain Strain
-
The demand for GPUs, specialized chips (e.g., Nvidia H100, TPUs), and servers drives up hardware costs.
-
AI chip shortages cause supply chain disruptions for other industries reliant on semiconductors.
-
E-waste concerns rise as AI models become obsolete quickly, leading to discarded hardware.
Economic & Market Disruptions
-
Job displacement: Automation replaces traditional roles faster than new AI-driven jobs emerge.
-
Market consolidation: The AI industry is dominated by a few major players, limiting accessibility for smaller companies.
-
Rising cloud service fees: While AI tools are cheap, the cost of high-scale deployment can still be a barrier.
Ethical & Regulatory Costs
-
AI bias and misinformation: Cheap, widespread AI increases the risk of deepfakes and misinformation.
-
Data privacy concerns: AI requires vast amounts of training data, raising ethical concerns over user consent and data ownership.
-
Regulatory fines and compliance risks: As governments step in with AI laws, companies must spend resources on compliance.
Security & Cyber Risks
-
AI-powered cyber threats: Hackers are using AI to create more sophisticated attacks.
-
Higher security costs: Organizations must invest more in AI security to prevent misuse and vulnerabilities.
-
AI hallucinations & reliability risks: Even the best AI models can generate misleading or outright false information, requiring human oversight.
While it’s too much for us to cover each of these “costs” in an article, it’s always good to bring these issues up early so they stay on our radar. Just like how being able to better navigate the potholes on the road once we know them, awareness is key to strategic business growth.

Wait for the right moment? Or make this the right moment?
Look Back As We Go Forward
We’ve looked at facts, figures and frameworks.
And if AI is not on your priority list for 2025, I can only hope this article will change your mind. After all, you’ve read this far.
At the start of this piece, we look back at history, at Jevons Paradox. We used that theory to review the structural changes in the past.
The good thing about history though? We know how it turns out.
The story is done, the outcome is fixed, the path is set.
The question is, while the story, outcome and path of AI are only in the early chapters, at the epilogue of this “book”, do you want to emerge the victim of the circumstances, or the victor of your choosing?
Or, as my Master would say, you can sit on the bench and watch life go by, or you can get off and be part of it.

Maverick Foo
Lead Consultant, AI-Enabler, Sales & Marketing Strategist
Why Most AI Training Fails (And What to Do Instead)
Feb 15, 2025
Is it rewarding to have employees taking such initiatives to become more productive… or is the risk of security something to worry about?
The AI Advantage Framework
Feb 8, 2025
Is it rewarding to have employees taking such initiatives to become more productive… or is the risk of security something to worry about?
AI Adoption X AI Enablement
Jan 26, 2025
Is it rewarding to have employees taking such initiatives to become more productive… or is the risk of security something to worry about?
REACH OUT TO US
Fill up your details below, and we will get back to you A.S.A.P.!
0 Comments