DeepSeek R1: A Game-Changer for Agentic Companies

On January 20, 2025, a seismic shift occurred in the artificial intelligence (AI) landscape: DeepSeek R1 was released. This open-source AI model, developed as a side project by a small Chinese firm, has sent shockwaves through the tech industry and the stock market. Why? Because DeepSeek R1 delivers performance on par with OpenAI’s o1 at maybe 2% of the cost.

A week later, the broader market took notice when the Apple App Store version of DeepSeek R1 was released. Nvidia’s stock dropped 17%, and Oracle was down 8%, underscoring the disruptive potential of this new model. While much of the conversation has revolved around its impact on AI giants like OpenAI, Meta, and Microsoft, the release of DeepSeek R1 is perhaps even more significant for Agentic Companies—organizations developing compute-intensive AI solutions, such as streaming avatars and edge computing applications. These companies now have a tremendous opportunity to reduce costs, increase scalability, and deliver more powerful solutions to their customers. Both President Trump and Nvidia had positive comments, calling this "a wake-up call" and "an excellent advancement for AI."

DeepSeek R1: Redefining AI Economics

DeepSeek R1’s release has disrupted long-standing assumptions about the costs of building and running state-of-the-art AI models. To put things into perspective, OpenAI’s o1 reportedly cost hundreds of millions of dollars to train, while DeepSeek R1 was developed for just $5 million—a mere 2% of the cost.

But the real game-changer lies in inference costs. Running advanced AI models like OpenAI’s o1 for time-consuming streaming avatar agents would have been cost prohibitive before, limiting their practical applications in cost-sensitive industries. With DeepSeek R1, these costs can now be easily affordable while not sacrificing any performance. For Agentic Companies, this means streaming AI applications, which rely on continuous inference, are now economically viable at scale. This technological breakthrough is an excellent case of Jevons Paradox: by making resource usage significantly more efficient, it’s likely to drive increased demand and total resource consumption. This shift is likely to significantly benefit the Agent layer, which stands on the shoulders of LLMs, positioning it as one of the clear winners in this efficiency-driven surge in demand."

Hygia: Affordable AI for Healthcare

Consider Hygia, an AI HealthTech company that we are working with to develop 24/7 streaming avatars to support aging and chronic care populations. These avatars act as conversational companions, health monitors, and administrative assistants, helping to improve the quality of service for the elderly and all patients. However, the company had been grappling with the reality of embedding LLM costs of $20/hour for older LLMs with a subset of the skills needed.

DeepSeek R1 changes the equation entirely. With inference costs potentially dropping to $0.20/hour, LLM costs can become a negligible factor, enabling Hygia to scale their solutions affordably and deliver services to a broader audience. This could be a great boost for the economy, with US healthcare costs on track to reach 26% of GDP in 5 years. Hygia is also a great example of how the new breed of agent-layer companies can become LLM-agnostic, and quickly shift to running whichever fundamental layer model delivers the best quality-cost mix.

Furthermore, open-source models like DeepSeek R1 can be customized or fine-tuned to work even better with agent-layer companies. This flexibility opens the door for Hygia to optimize the model for specific healthcare use cases and even explore vertical integration, such as deploying tailored edge-computing instances to enhance scalability and reduce latency for real-time applications.

CEEVUE: Edge Computing for Video Conferencing

Another exciting company that we are working with is CEEVUE, which is revolutionizing video conferencing with its agent-driven 180-degree video conferencing system. CEEVUE’s innovative platform combines streaming avatars with live guests, creating immersive and engaging meeting experiences. With DeepSeek R1, CEEVUE now has the ability to embed a customized open-source version of the model directly onto Nvidia Nano SuperComputer that can accompany each screen.

This approach offers several advantages:

• Eliminates Bandwidth Constraints: By running the model locally on edge computing devices, CEEVUE reduces reliance on cloud infrastructure and avoids network bottlenecks.

• Avoids Inference Time Limits: Local deployment ensures uninterrupted performance without worrying about API rate limits or latency.

• Ensures Cutting-Edge Performance: Using a customized version of DeepSeek R1 ensures that the platform is equipped with a state-of-the-art LLM tailored to its specific needs.

CEEVUE’s integration of edge computing with DeepSeek R1 is a prime example of how agentic companies can harness the power of open-source AI to push the boundaries of innovation.

The Power of Open Source for Agentic Companies

DeepSeek R1’s open-source release is not just a win for the AI community—it’s a catalyst for innovation across industries. By making the model and its methodology freely available, DeepSeek has empowered agentic companies to experiment, replicate, and innovate without prohibitive costs. Prominent voices in the AI community, like Meta’s Yann LeCun, have celebrated this as a victory for open-source collaboration.

Open-source AI democratizes access to cutting-edge technology, allowing smaller companies to compete with tech giants. It also accelerates the development of specialized applications. For agentic companies, this means faster deployment of tools that address critical challenges, such as reducing costs, improving performance, and expanding use cases.

Customized open-source models also pave the way for vertical integration. For example, agentic companies like Hygia and CEEVUE could explore deploying tailored edge-computing instances that reduce latency, enhance scalability, and align AI capabilities with specific business objectives.

Broader Implications for the Agentic Ecosystem

The release of DeepSeek R1 is more than just a technological breakthrough—it’s a signal of what’s possible when efficiency meets innovation. For agentic companies, the implications are profound:

• Lower Costs: Reduced inference costs make compute-intensive applications economically viable.

• Greater Flexibility: Open-source models allow for customization, enabling tighter alignment with business needs.

• Improved Scalability: Edge computing solutions can localize AI capabilities, reducing dependency on cloud infrastructure and enabling seamless performance.

Conclusion: A New Era for Agentic Companies

DeepSeek R1 represents a monumental opportunity for agentic companies. From Hygia’s healthcare avatars to CEEVUE’s cutting-edge video conferencing systems, the ability to deploy state-of-the-art AI at dramatically reduced costs is transformative. By leveraging open-source innovation, these companies can not only scale their solutions but also explore new frontiers, such as vertical integration and edge computing.

For agentic companies, the timing couldn’t be better. DeepSeek R1 democratizes access to advanced AI, enabling a new wave of creativity and efficiency that will shape the future of industries worldwide