I don’t know about you, but this year started with a hiss and a roar, and suddenly, here we are—mid-February! Time flies when you’re juggling running a business, moving house, and keeping up with life.
But back to work, and one of our favourite topics, AI, just when Silicon Valley thought it had won the game, a new player has entered the table—and it might just be calling OpenAI’s bluff. DeepSeek, the Chinese AI startup that’s sending shockwaves through the industry. With a model that reportedly rivals OpenAI and Google’s best, but at a fraction of the cost and energy consumption, DeepSeek is making us all rethink what’s possible in AI.
But is this truly the start of an AI revolution, or just the latest twist in the ongoing battle for dominance? And what does this mean for AI’s future - both in terms of accessibility and ethical concerns?
Let’s dive into the hype, the reality, and the sudden rise of DeepSeek.
Is DeepSeek Calling OpenAI’s Bluff?
In poker, it’s not just about the cards you hold—it’s about convincing your opponents you have the winning hand. OpenAI’s Sam Altman, a former poker player, seems to have played this game well in AI. With ChatGPT’s meteoric rise, OpenAI successfully convinced the world that bigger is always better - more data, more processing power, more expensive GPUs.
But DeepSeek just surprised everyone, they upended the table.
In comes DeepSeek R1, a model reportedly as powerful as OpenAI’s best model but trained at a fraction of the cost. While OpenAI is betting on $500 billion AI infrastructure investments, DeepSeek took a different - one that challenges Silicon Valley’s assumption that AI dominance requires unlimited resources.
DeepSeek’s entrance raises an important question: Was the AI industry overspending on scale all along? If DeepSeek can achieve OpenAI-level performance with a budget of just $6 million, does this mean that the arms race for ever-larger AI models was more about corporate hype than actual necessity?
What Makes DeepSeek Different?
DeepSeek’s success wasn’t magic—it was engineering. Here’s what sets it apart from the compute-hungry models developed by OpenAI, Google, and Meta:
Reinforcement Learning 2.0: Teaching AI to Reason
Unlike traditional AI models that rely heavily on human-labeled training data, DeepSeek R1 trains itself through reinforcement learning. This method - once considered too unpredictable for language models—has allowed DeepSeek to achieve cutting-edge results while using significantly less compute power.
By focusing on structured problems like math and coding, DeepSeek optimized its model to become more efficient at reasoning, rather than blindly predicting text based on probability.
Model Distillation: Bigger Isn’t Always Better
Instead of continuously scaling up, DeepSeek developed a method to "distill" the intelligence of large models into smaller, more efficient ones. This means that less powerful chips can run AI models that once required supercomputing clusters.
In contrast, OpenAI and its rivals have largely followed the "bigger is better" philosophy - spending billions on massive AI clusters with increasing power demands.
Cost-Efficient AI: Democratizing Access
DeepSeek’s models are 20 to 50 times cheaper to run than OpenAI’s equivalent. This means that small businesses, startups, and even individuals could afford high-performance AI without needing expensive infrastructure.
DeepSeek’s pricing model is a potential game-changer. It could shift AI from an elite, billion-dollar industry into a tool accessible to everyone.
The Environmental and Economic Implications of AI Scaling
With AI's rapid expansion, there’s a growing concern about its energy consumption and economic feasibility. If the future of AI requires mega data centers consuming as much power as small countries, then we may be headed for an unsustainable model.
Is DeepSeek the Green Alternative?
AI companies have been racing to expand their energy-intensive AI clusters, with OpenAI, Google, and Microsoft investing in new data centers that consume massive amounts of electricity.
DeepSeek’s energy-efficient training offers an alternative:
Less computing power required to train AI models
Lower operational costs for running AI applications
Reduced environmental impact compared to traditional AI models
If DeepSeek’s methods become the industry standard, it could delay or even eliminate the need for trillion-dollar investments in AI infrastructure.
Are Big AI Investments Overkill?
OpenAI, Microsoft, and Google have spent billions on high-end GPUs, power-hungry servers, and data centers. The justification? AI dominance requires massive financial backing.
But DeepSeek’s breakthrough suggests otherwise.
If AI models can be trained and run on far less hardware, are companies like OpenAI overspending?
Will businesses start looking for cheaper alternatives, rather than relying on Big Tech’s expensive AI services?
With DeepSeek R1 delivering results at a fraction of the cost, we may see a shift away from Silicon Valley’s monopoly.
The US-China Chip War, Did Washington’s Plan Backfire?
For years, the United States has controlled the semiconductor industry, particularly the production of high-end AI chips. Recognising that China’s AI ambitions depended on access to cutting-edge chips, the US government imposed export bans on advanced GPUs—particularly NVIDIA’s A100 and H100, which are crucial for training large AI models.
The goal? Slow down China’s AI progress and maintain US dominance in artificial intelligence.
DeepSeek’s rapid rise suggests that this strategy may have backfired.
The US Thought Chips Were the Key
When Washington tightened export controls, many expected China’s AI development to stall. Instead, DeepSeek found ways to work around the restrictions—demonstrating that AI innovation isn’t just about having the best hardware.
What did they do?
Instead of relying on the latest NVIDIA GPUs, DeepSeek:
Stockpiled older NVIDIA A100 chips before the export ban
Optimized its AI models to run on less powerful hardware
Used a "Mixture of Experts" approach, activating only parts of the model when needed
DeepSeek R1 achieved OpenAI-level performance without access to cutting-edge chips.
This raises an uncomfortable question for US policymakers: Did the US just spend years stockpiling high-end AI chips, only to learn they weren’t necessary?
The "Chokepoint" That Wasn’t
The US believed that cutting off China’s access to top-tier chips would limit its AI capabilities. However, Chinese AI companies:
Developed new optimization techniques that reduced dependency on high-end chips
Invested in domestic chip manufacturing to produce alternatives
Focused on software-based efficiency improvements rather than brute-force hardware scaling
By focusing on efficiency instead of sheer processing power, DeepSeek made the AI arms race less about hardware and more about smart engineering. Have a look at the number of engineers in China compared to the US.
A Wake-Up Call for Silicon Valley?
For years, companies like OpenAI, Google, and Meta followed the "bigger is better" philosophy—scaling up AI models with exponentially larger datasets and computational power.
DeepSeek just proved that this approach may not be the only path forward.
If smaller, more efficient AI models can achieve the same results, then:
Will Big Tech rethink its reliance on trillion-dollar AI infrastructure?
Will investors start questioning whether the AI industry’s obsession with size was justified?
Could AI become more decentralized, reducing reliance on just a few dominant players?
This isn’t just about AI—it’s about who controls the future of technology.
What This Means for Businesses and Consumers
DeepSeek’s breakthrough isn’t just an academic or geopolitical discussion - it has real-world implications for businesses, tech companies, and everyday consumers.
For years, AI development has been dominated by US companies, requiring massive investments in hardware and cloud infrastructure. DeepSeek’s rise challenges this model, making AI cheaper, more efficient, and more accessible. But what does this mean for the rest of us?
A More Competitive AI Market
For businesses, DeepSeek’s advancements mean more choices and lower costs when adopting AI solutions.
Lower Prices: If DeepSeek’s AI models are 20-50x cheaper to run than OpenAI’s models, companies will spend less on AI-powered tools and services.
More AI Providers: Competition forces companies like OpenAI, Google, and Anthropic to innovate faster and lower their prices.
Better Accessibility: Smaller businesses that couldn’t afford high-end AI tools may now have access to models that previously required deep pockets.
Will companies stick with expensive Western AI models, or will DeepSeek force the industry to rethink pricing?
The End of the "Bigger is Better" AI Model?
DeepSeek’s success has disrupted Silicon Valley’s playbook. The AI industry has operated under the assumption that more data, more GPUs, and more money automatically result in better AI models.
DeepSeek challenges this idea:
Efficiency over scale: Instead of throwing billions at computing power, the future may be about "smart scaling"—making models smaller, cheaper, and more efficient.
More companies will experiment with smaller models: If DeepSeek can match OpenAI without massive supercomputers, expect more startups to challenge the AI giants.
Investment strategies may shift: Investors might start questioning whether AI companies need billion-dollar budgets to remain competitive.
Cheaper and More Ethical AI?
For consumers, the biggest impact will be on:
Lower subscription costs: If DeepSeek’s AI is cheaper to run, companies offering AI-powered services (e.g., ChatGPT, Claude, Gemini) will be pressured to lower their prices.
Less environmental impact: AI models trained using less energy-intensive methods could reduce the carbon footprint of AI-driven industries.
More AI options, but with risks:
Western AI models (like ChatGPT) have been trained with strict ethical guidelines, reducing bias and avoiding political manipulation ( This one is controversial as we have seen misinformation spread across the internet).
DeepSeek, self-censors and follows the Chinese government’s restrictions - raising concerns about AI bias and global digital influence.
Will Western consumers trust AI models with government-imposed censorship, or will regulation shape the AI landscape?
AI's Future Just Became Unpredictable
DeepSeek’s rise has changed assumptions about AI development.
AI doesn’t need to be expensive to be powerful.
The US no longer has a monopoly on AI innovation.
Businesses and consumers will soon have more choices than ever before.
The next few years will determine whether Big Tech adapts to this new AI era - or whether DeepSeek truly changes the game.
Would you switch to a cheaper AI model, even if it came with political restrictions?