DeepSeek and the falling cost of AI: good news for small businesses
A Chinese start-up's reasoning model triggered the largest one-day loss of market value in stock market history. Behind the headlines is a simpler trend: useful AI is getting dramatically cheaper.
By the Xuru Technologies team · 28 January 2025
Key takeaways
- On 27 January 2025 Nvidia lost about US$589 billion in market value in a single day after DeepSeek's R1 model drew attention to how cheaply strong AI can be built.
- The model's weights were released openly, so other companies can run and adapt it.
- Competition is pushing AI prices down, which makes chat assistants, document processing and automation affordable for small teams.
- Cheaper is not the same as safe: where your customer data is processed still matters.
A week ago, most business owners had never heard of DeepSeek. By Monday its app had topped download charts and investors were asking whether the huge sums being spent on AI chips were all necessary.
What happened
DeepSeek released its R1 reasoning model in January with performance close to leading Western models on many tests and a reported training cost far below what US labs spend. On 27 January, chip and AI stocks fell sharply. Bloomberg reported that Nvidia's US$589 billion drop was the largest single-day loss of market value in history. The selloff came days after the US announced Stargate, a plan to invest up to US$500 billion in AI infrastructure — a reminder that the race is happening at both ends: huge investment and rapidly falling prices.
What it means for businesses
- AI features are no longer a luxury. Summarising enquiries, drafting replies, translating content and answering customer questions now cost a fraction of what they did a year ago.
- More choice of models. Open-weight models can be hosted by cloud providers in the region you choose, which gives businesses options on cost and data location.
- The value moves to how AI is used. When the model is cheap, the advantage comes from good data, clear workflows and a website that turns answers into enquiries.
- Data rules still apply. Before sending customer information to any AI service, check where it is processed and whether it is used for training. This matters under GDPR, UK GDPR and India's privacy rules.
What to do now
| Step | Why |
|---|---|
| List three repetitive tasks (enquiry replies, quotes, FAQs) | These are the cheapest wins for AI help |
| Write down what data each task touches | Decides which AI services you can use safely |
| Pilot one workflow with a person approving the output | Gets value quickly while keeping quality under control |
| Measure time saved after 30 days | Shows whether to expand or stop |
If you are unsure where to start, our guide to seven AI workflows that save hours covers practical examples.
Frequently asked questions
Is it safe for a business to use DeepSeek?
It depends on how you access it. Using a hosted app means your data is processed on that provider's servers under its terms. Running an open-weight model through a cloud provider in your own region gives more control. Avoid sending personal or confidential data to any service until you have checked its terms.
Will AI get cheaper still?
Prices for a given level of capability have been falling quickly as competition increases. Plan for costs to keep dropping, but budget for the work of setting AI up properly.
Do I need my own AI model?
Almost never. Small businesses get the most value by connecting existing models to their own information and workflows.
Put affordable AI to work on your enquiries.
We set up AI assistants and workflows for small businesses, with your data handled carefully and a person approving key steps.
Websites, payments, digital marketing, AI automation, consultancy and practical technology training — for businesses and aspiring professionals worldwide.
Saturday – Sunday: closed