Steve Hsu, cofounder of AI startup SuperFocus and professor at Michigan State University, is leading his company’s transition from closed-source AI models like GPT-4o to DeepSeek’s open-source alternatives, citing three compelling reasons that could reshape the AI industry landscape.
SuperFocus specializes in creating superhuman AI systems for businesses, handling use cases including customer service, ordering, delivery, document analysis, and travel scheduling. After extensive testing of DeepSeek-V3 models, Hsu’s team has confirmed the technology is robust enough to replace their current OpenAI-powered systems.
Price represents the most dramatic advantage: DeepSeek models cost approximately 30 times less to run than comparable OpenAI models. For customer service applications, this translates to extraordinary savings. While US customer service employees cost around $25 per hour and Philippine workers cost $5-10 per hour, SuperFocus’s OpenAI-powered models already achieved costs 10 times lower than Philippine labor. With DeepSeek-V3, costs could drop another 30-fold, making AI customer service exponentially more affordable.
Speed improvements matter significantly for SuperFocus’s voice interaction applications. The DeepSeek team implemented clever optimizations and modified model architecture to reduce memory usage and computational requirements, resulting in faster token generation. This reduced latency creates more natural conversations, as humans expect minimal delays between speaking and receiving responses.
Privacy concerns drive the third major advantage. Open-source DeepSeek models can run entirely on customer hardware or rented cloud servers without communicating back to DeepSeek. This allows SuperFocus to build AI platforms fully within clients’ existing cloud instances, keeping sensitive data—like private equity documents—completely secure. Closed-source models from OpenAI or Chinese companies require API communication through company-controlled hardware, creating potential security vulnerabilities.
Hsu acknowledges some customers may prefer avoiding Chinese models for political reasons, but emphasizes that for narrow AI applications serving specific business purposes, such concerns rarely arise. He praises DeepSeek as impressively open and transparent, noting the company produces detailed research papers and encourages validation of their work—ironically more open than most top US labs, including OpenAI.
The founder dismisses accusations that DeepSeek stole information or misrepresented their $5.5 million training costs, viewing such claims as denial about China’s AI progress. Hsu predicts intense US-China competition will benefit consumers through lower prices and faster AGI development, while enabling researchers to fine-tune open-source reasoning models for scientific breakthroughs.
Key Quotes
DeepSeek models are about 30 times cheaper to run than comparable OpenAI models. That offers significant cost savings for the customer service industry.
Steve Hsu explains the dramatic cost advantage that’s driving his startup’s switch to DeepSeek. This price differential could fundamentally reshape the economics of AI deployment across industries, making sophisticated AI systems accessible to far more businesses.
I think DeepSeek is a very impressive company that produces really good research papers; you can tell that they’re trying to be very clear about what they did and their results, and are encouraging others to reimplement or validate what they did.
Hsu praises DeepSeek’s transparency and openness, contrasting it with American AI labs. This assessment challenges narratives about Chinese technology companies being secretive and highlights an ironic reversal where Chinese firms are more open than US counterparts like OpenAI.
Our American self-esteem is so high that people are in denial and pretending that we’re still ahead in AI and China hasn’t caught up.
Hsu directly addresses American skepticism about DeepSeek’s capabilities, suggesting that accusations of theft or fabricated costs reflect denial rather than evidence. This observation speaks to broader geopolitical tensions surrounding AI leadership and technological competition.
It’s going to be a great competition, and consumers will win because the price goes down. People will get more and more inference and applied intelligence at a lower price in their daily lives.
Despite geopolitical tensions, Hsu emphasizes that US-China AI competition will ultimately benefit end users through lower costs and faster innovation. This optimistic view suggests that competitive pressure, rather than monopolistic control, drives better outcomes for AI adoption.
Our Take
Hsu’s perspective represents a pragmatic, technically-grounded view that contrasts sharply with the nationalist rhetoric dominating AI discourse. His willingness to embrace Chinese open-source models based purely on technical and economic merit—speed, cost, and privacy—demonstrates how business realities may override geopolitical concerns.
The 30-fold cost reduction isn’t incremental improvement; it’s transformative disruption that could trigger mass AI adoption across sectors currently hesitant due to expense. If DeepSeek’s performance claims hold under broader scrutiny, we’re witnessing a potential inflection point where open-source models become the default choice for most applications.
Hsu’s observation about American denial is particularly noteworthy. The AI community’s dismissive response to DeepSeek—claiming theft or fabrication—mirrors historical patterns where established leaders underestimate emerging competitors. Whether DeepSeek’s claims fully withstand scrutiny, the competitive pressure will undoubtedly accelerate innovation and reduce costs industry-wide, ultimately benefiting the entire AI ecosystem.
Why This Matters
This development signals a potential paradigm shift in the AI industry from expensive closed-source models to cost-effective open-source alternatives. DeepSeek’s emergence challenges the dominance of American AI giants like OpenAI and Anthropic, intensifying the US-China AI race that will shape technological leadership for decades.
For businesses, the 30-fold cost reduction could democratize AI adoption across industries, making sophisticated AI systems accessible to smaller companies previously priced out of the market. The customer service sector alone could see massive disruption, with AI agents becoming economically superior to human workers globally.
The privacy advantages of open-source models address critical enterprise concerns about data security, potentially accelerating AI adoption in regulated industries like finance, healthcare, and legal services. Meanwhile, researchers gaining access to powerful open-source reasoning models could accelerate scientific discovery across physics, mathematics, and other fields.
This story reflects broader tensions about technological competition, national security, and the future of AI development—whether it will be controlled by a few proprietary labs or distributed through open-source collaboration. The outcome will fundamentally impact innovation speed, cost accessibility, and global AI leadership.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources:
Recommended Reading
Related Stories
- OpenAI CEO Sam Altman’s Predictions on How AI Could Change the World by 2025
- Mistral AI’s Consumer and Enterprise Chatbot Strategy
- How Companies Can Use AI to Meet Their Operational and Financial Goals
- The AI Hype Cycle: Reality Check and Future Expectations
- OpenAI’s Valuation Soars as AI Race Heats Up
Source: https://www.businessinsider.com/ai-startup-founder-3-reasons-to-switch-to-deepseek-2025-1