DeepSeek has reignited the artificial intelligence race, but this time the battle isn’t over which model is smartest. It’s about which one offers the best value.
The Chinese AI company has shaken the global industry by dramatically lowering the cost of AI models, forcing businesses to rethink how they invest in artificial intelligence. While cheaper AI opens the door for more companies to experiment, experts say cost alone won’t determine success.
DeepSeek changes the AI conversation
For much of the past two years, AI developers have competed to build the smartest models.
DeepSeek introduced a different question: which model is the cheapest?
The company attracted global attention after releasing its R1 reasoning model, claiming it cost around $5.6 million to train while delivering performance close to leading American AI models.
The announcement sent shockwaves through financial markets. Nvidia’s share price dropped 17% in a single day, wiping $589 billion off the company’s market value.
DeepSeek originated from High-Flyer, a Chinese hedge fund founded by Liang Wenfeng, who built his fortune using AI-powered trading models before moving into AI development.
Since then, the company has focused on open-source models and aggressive pricing.
Lower prices are changing the economics
DeepSeek launched V4-Flash on 31 July, with the model currently available in public beta.
Although it does not outperform the strongest AI models from companies such as OpenAI and Anthropic, its significantly lower pricing changes the equation for businesses that do not require cutting-edge performance.
For many organisations, “good enough” AI at a much lower cost makes commercial sense.
Independent testing is still ongoing, but one trend is already clear.
Blended AI inference costs across major providers fell by roughly two-thirds between early 2025 and early 2026, with DeepSeek playing a major role in driving those reductions.
What it means for South African businesses
For South African businesses, the biggest challenge has rarely been interest in AI.
Instead, companies have struggled to justify the investment financially.
Lower AI costs improve that business case, but they don’t guarantee a return on investment.
Price is only one part of the equation.
Lower prices don’t always mean lower costs
One misconception is that cheaper AI automatically leads to lower spending.
The opposite is often happening.
Enterprise AI spending continues to increase because businesses are using AI more frequently.
As organisations adopt AI agents capable of completing complex tasks through dozens of model requests, overall usage continues to grow faster than prices are falling.
The result is that many businesses are still spending more on AI, despite lower costs per request.
Cheaper AI also makes failure cheaper
Lower costs reduce the financial risk of experimentation, but they do not guarantee successful outcomes.
Research into enterprise AI projects continues to show that many pilots fail to generate measurable business value.
A cheaper model simply reduces the cost of unsuccessful experiments.
DeepSeek itself also appears to be adjusting its pricing strategy.
The company has announced plans to introduce peak-hour pricing that will increase rates during business hours in China, signalling that even the industry’s most aggressive price competitor may not remain the cheapest indefinitely.
Success depends on business strategy
The falling cost of AI gives more organisations access to powerful technology than ever before.
But affordability is only the starting point.
The businesses that benefit most will be those that clearly measure how AI improves efficiency, customer experience or revenue once deployed at scale.
As AI becomes more affordable, the real competitive advantage will come from using it effectively rather than simply using it cheaply.
Read SA Business Integrator online
