September 19, 2025

news  
DeepSeek R1 Model, Nvidia GPUs, OpenAI Cost Comparison

Chinese AI startup DeepSeek has significantly challenged industry cost estimates by training its R1 reasoning model for just $294,000, utilizing 2,000 NVIDIA H800 GPUs over 45 days. This figure stands in stark contrast to the millions, or even over $100 million, reportedly spent by U.S. competitors like OpenAI for models such as GPT-4. DeepSeek's R1 model demonstrates strong performance in reasoning and coding, positioning China as a growing force in AI development. Meanwhile, Nvidia is investing £2 billion into UK AI startups, signaling a broader global push in AI innovation. Beyond model development, AI is being integrated across various sectors. Oracle is investing $5 billion in the UK for cloud infrastructure and generative AI services, including a sovereign cloud for government and defense. In the foodservice industry, AI is optimizing supply chains and menu development, while S-Oil has launched an AI sales analysis system for gas stations. Security operations are also seeing AI advancements with Exaforce's new platform and a partnership between Dataminr and Genetec for AI-powered physical security. S&P Global is using AI to enhance its data management and products, with its Spark Assist Platform already adopted by two-thirds of employees. However, a Harvard Business Review article cautions that AI might inadvertently reinforce company silos, hindering overall strategic goals if not managed carefully. In education, a new bill proposes standards for teaching AI in K-12 schools to prepare students for the future workforce. Onix has been recognized as a Leader in AI Services by QKS Group, leveraging its Google Cloud partnership to help organizations operationalize AI.

Key Takeaways

  • Chinese AI startup DeepSeek trained its R1 reasoning model for $294,000, using 2,000 NVIDIA H800 GPUs over 45 days, a cost significantly lower than U.S. competitors.
  • Nvidia is investing £2 billion in UK AI startups to foster innovation in the region.
  • Oracle is investing $5 billion in the UK to expand cloud infrastructure, including a sovereign cloud offering with generative AI services for government and defense.
  • AI is being implemented in the foodservice industry for supply chain optimization, route management, and menu development.
  • S-Oil has launched the first AI and big data-based sales analysis system, MAAS, for gas stations in South Korea.
  • Exaforce has introduced a new AI SOC platform utilizing multi-model AI and agentic automation for security operations.
  • Dataminr and Genetec are partnering to integrate real-time AI event and risk intelligence into physical security systems.
  • A Harvard Business Review article warns that AI may reinforce organizational silos if not managed proactively.
  • A new bill, the RAISE Act, proposes academic standards for teaching AI and emerging technologies in K-12 schools.
  • Onix has been named a Technology Leader in AI Services by QKS Group, focusing on helping organizations operationalize AI, particularly as a Google Cloud partner.

DeepSeek AI model training cost under $300,000

Chinese AI startup DeepSeek announced its R1 reasoning model was trained for $294,000. This cost is significantly lower than what U.S. competitors like OpenAI reportedly spend. DeepSeek used 2,000 NVIDIA H800 GPUs for 45 days to train the model. The R1 model shows strong performance in reasoning and coding, competing with top AI models. This development raises questions about AI development costs and China's growing role in AI.

China's DeepSeek trains AI model for $294,000, challenging US tech

Chinese AI firm DeepSeek revealed it trained its R1 model for $294,000, a fraction of the cost for U.S. tech giants. The company used 2,000 NVIDIA H800 GPUs over 45 days. This low cost challenges industry estimates for training advanced AI models, which can reach millions of dollars. DeepSeek's R1 model shows competitive performance in reasoning and coding. The development highlights China's advancements in AI and could impact the global AI landscape.

DeepSeek AI model training cost reported at $294,000

Chinese AI developer DeepSeek stated its R1 model training cost $294,000, using 512 Nvidia H800 chips. This figure is much lower than estimates for U.S. AI models like OpenAI's GPT-4, which could exceed $100 million. DeepSeek also acknowledged using A100 GPUs in earlier development stages. The company's lower-cost AI systems previously caused concern among investors about competition with AI leaders like Nvidia. This disclosure is likely to intensify discussions about China's AI capabilities.

DeepSeek AI model training cost significantly lower than US rivals

Chinese AI developer DeepSeek claims its R1 model cost only $294,000 to train, a stark contrast to U.S. competitors. The company published this estimate in the academic journal Nature, noting the use of 512 Nvidia H800 chips. This revelation follows DeepSeek's earlier release of lower-cost AI systems that impacted tech stocks. U.S. officials have questioned some of DeepSeek's claims, particularly regarding chip usage. DeepSeek also admitted to using A100 GPUs for initial development phases.

Exaforce launches AI SOC platform for security operations

Exaforce has introduced a new AI SOC platform designed to improve security operations. The platform combines multi-model AI, agentic automation, and data exploration to detect threats, reduce false positives, and automate investigations. It features specialized AI agents called Exabots for detection, triage, investigation, and response. Exaforce aims to provide comprehensive security coverage for companies with or without a dedicated SOC. The platform offers advanced detection for cloud and SaaS environments and automates alert triage to save analyst time.

Dataminr and Genetec partner for AI-powered physical security

Dataminr and Genetec are collaborating to integrate real-time AI event and risk intelligence into physical security systems. This partnership aims to break down data silos and provide organizations with early warnings of risks and safety events. Dataminr's AI platform will feed intelligence into Genetec's Security Center, allowing security teams to combine external risk signals with internal controls. The integration will enhance situational awareness and enable automated workflows for faster responses to protect people and facilities.

New Bill Proposes AI Standards in Education

Senator Jon Husted introduced Senate Bill 2740, the Recommending Artificial Intelligence Standards in Education Act of 2025 (RAISE Act). This bill encourages states to develop academic standards for teaching artificial intelligence and emerging technologies in K-12 schools. The goal is to better prepare students for a future workforce where AI is prevalent. States would need to integrate AI concepts into curricula and train teachers. Potential challenges include resource allocation and keeping standards updated with rapid technological changes.

S-Oil launches AI sales analysis system for gas stations

South Korean company S-Oil has launched the Market Analysis Automation System (MAAS), the first AI and big data-based sales analysis system for gas stations. MAAS replaces intuition with data-driven strategies for pricing, marketing, and expansion. It analyzes competitor prices, customer traffic, and market conditions to suggest optimal pricing. The system also tailors marketing by tracking customer behavior and aids in selecting new station locations. S-Oil plans to refine MAAS based on operator feedback and enhance its algorithms.

AI risks reinforcing company silos, warns Harvard Business Review

A Harvard Business Review article warns that while AI promises efficiency, it may be reinforcing organizational silos. Companies might see individual departmental improvements but struggle to achieve overall strategic goals. This happens as departments focus on their own AI-driven operations, becoming isolated. The article suggests leaders should actively work to prevent AI from deepening these divisions. It highlights the need for cross-functional collaboration to ensure AI benefits the entire organization.

AI drives efficiency and innovation in foodservice industry

Artificial intelligence is already delivering value in the foodservice industry, with many CEOs implementing or considering generative AI. AI helps optimize supply chains, manage transportation through route optimization, and improve menu development by analyzing trends in real time. Experts advise companies to focus on data quality and identify specific problems to solve with AI. Key strategies for adoption include establishing an organizational model, creating an AI strategy framework with governance, and educating teams on AI tools and policies.

Onix named Leader in AI Services by QKS Group

QKS Group has recognized Onix as a Technology Leader in its 2025 SPARK Matrix for Artificial Intelligence Services. Onix received strong ratings for its technology and customer experience. The company offers a comprehensive AI services portfolio, including proprietary tools like Eagle for planning and Phoenix for AI development. Onix leverages its expertise as a Google Cloud partner to help organizations operationalize AI across various industries such as manufacturing and healthcare. Their approach focuses on reducing implementation risks and accelerating value realization.

S&P Global uses AI to improve data management and products

S&P Global is integrating artificial intelligence into its operations to enhance its products and internal processes. CEO Martina Cheung stated that AI integration and data management are key strategic priorities. The company's S&P Spark Assist Platform is already used by two-thirds of its employees, with expectations of reduced headcount. S&P Global is also exploring partnerships to make its financial data more accessible to AI tool users, recognizing the growing client interest in AI for productivity and cost savings.

Oracle expands UK investment with sovereign cloud and AI plans

Oracle is investing $5 billion in the UK to expand its cloud infrastructure, including a sovereign cloud offering for government and defense users. This platform will provide generative AI services and ensure data compliance with national regulations. The initiative aims to offer a secure, shared platform for the UK, US, and NATO. Oracle CEO Safra Catz highlighted the investment's role in supporting allied nations' defense efforts. NATO's Communications and Information Agency is already migrating workloads to Oracle's sovereign cloud for enhanced data management and control.

Nvidia invests £2 billion in UK AI startups

Nvidia announced a £2 billion investment in the United Kingdom's artificial intelligence startup ecosystem. This investment aims to support the growth of AI innovation within the UK. Several asset managers will join Nvidia in this significant funding initiative. The announcement follows Nvidia's earlier $5 billion investment in Intel. This move underscores Nvidia's commitment to advancing AI technology globally and fostering emerging AI companies.

Sources

AI development cost DeepSeek R1 model NVIDIA H800 GPUs OpenAI GPT-4 China AI AI in security Exaforce AI SOC platform Exabots Dataminr Genetec AI in education RAISE Act AI standards K-12 education AI in foodservice supply chain optimization route optimization menu development generative AI Onix AI services QKS Group Google Cloud partner AI implementation S&P Global data management AI products S&P Spark Assist Platform financial data Oracle UK investment sovereign cloud generative AI services NATO Nvidia UK AI startups AI innovation Intel