Pat Gelsinger and the DeepSeek revolution: How openness and efficiency are transforming the world of AI

Last update: 07/02/2025
Author Isaac
  • DeepSeek launches its R1 model: A model of IA open source that significantly reduces development and operation costs compared to its competitors.
  • Pat Gelsinger bets on openness: The former president of Intel highlights the historical lessons of computing and the transformative potential of an open ecosystem.
  • Impact on the market: R1 has sparked debates in the tech industry, a drop in stocks of Nvidia and a rethinking of how AI innovation happens.
  • Gloo adopts DeepSeekGelsinger's startup deploys R1 to develop new AI tools efficiently and openly.
pat gelsinger uses deepseek-0
LISBON, PORTUGAL – NOVEMBER 08: Pat Gelsinger, CEO, VMware, and Ross Mason (not pictured), founder, MuleSoft debate on “The future of technology: Will everyone benefit?” during the third day of Web Summit in Altice Arena on November 08, 2017 in Lisbon, Portugal. Web Summit (originally Dublin Web Summit) is a technology conference held annually since 2009. The company was founded by Paddy Cosgrave, David Kelly and Daire Hickey. The topic of the conference is focused on internet technology and attendees range from Fortune 500 companies to smaller tech companies. This contains a mix of CEOs and founders of tech start ups together with a range of people from across the global technology industry, as well as related industries. This year's edition, starting on November 06, is the second to be held in Lisbon and will congregate almost 60,000 participants. (Photo by Horacio Villalobos – Corbis/Getty Images)

The world of Artificial Intelligence (AI) is undergoing a significant change thanks to the launch of the R1 model from DeepSeek, a Chinese startup that has managed to attract both admirers and critics. This model has shaken the foundations of the industry by demonstrating that it is possible to build high-performance systems at a fraction of the cost of traditional approaches. Among the most prominent supporters is Pat Gelsinger, former CEO of Intel and current president of Gloo, a messaging and engagement platform geared toward faith-based communities.

DeepSeek has changed the game with the launch of R1, an AI-based reasoning model that is not only highly efficient, but is also open source. This means that any developer can access the model, modify it, and adapt it to their needs without the limitations imposed by closed models, such as those of OpenAI or Anthropic. According to Gelsinger, this open approach is key to driving innovation and democratising AI.

Innovation under constraints: A model created with limited resources

DeepSeek has attracted attention not only for the performance of its model, but also for how it achieved it. Using 2.000 Nvidia H800 GPUs and a budget of around $5,5 million, DeepSeek managed to train R1 in just two months. This efficiency contrasts with the multi-million dollar resources that other AI giants regularly invest. Gelsinger highlighted that this feat is an example of how “Engineering flourishes under constraints", noting that DeepSeek's Chinese engineers had to find creative solutions in the face of significant constraints, including export restrictions on advanced chips from the U.S. to China.

  Generating clean and efficient code with DeepSeek: a complete guide

Pat Gelsinger talks about DeepSeek

Key lessons from the history of computing

Gelsinger used this occasion to recall three historical lessons that R1 highlights. The first is that Lowering computing costs expands markets. According to the former Intel CEO, making AI more affordable will not reduce its adoption, but rather amplify it, integrating it into applications and devices that previously could not afford it. This includes everything from auriculares to electric cars with voice recognition.

The second lesson is that engineering thrives when faced with constraints, as demonstrated by the development of R1. Finally, and perhaps most importantly, he noted that “the opening wins«. Gelsinger strongly believes in the power of open ecosystems to foster collaboration, drive transparency and promote ethical adoption of technology.

Impact on Gloo and the market

In his role as president of Gloo, Gelsinger decided to implement R1 on his platform, mentioning that R1 will replace models like OpenAI in the development of its AI assistant, Kallm. This move not only reflects Gelsinger’s confidence in DeepSeek’s potential, but also underlines the impact this technology is having on the market.

However, it hasn't all been enthusiasm. The launch of R1 caused a significant drop in the shares of Nvidia and other chipmakers, as investors feared that cost cutting would dampen demand for hardware advanced. Gelsinger disagreed with this interpretation, noting that the accessibility of AI will only increase its use, which in the long run will spur demand for more advanced hardware.

DeepSeek and Pat Gelsinger

Controversies and future challenges

Despite the praise, DeepSeek’s path has not been without its challenges. Some critics have questioned the accuracy of its training data and its performance compared to closed-source models like OpenAI. Others have raised concerns about transparency, as much of DeepSeek’s work was done in China, a country facing international criticism over issues like censorship and privacy.

  Complete Guide to Using the DeepSeek API Step by Step

However, Gelsinger has played down these concerns, saying the important thing is that DeepSeek demonstrates that innovation does not depend solely on intensive use of resources, but on creativity and efficient engineering.

A blueprint for the future of AI

With the success of R1, DeepSeek has positioned itself as a catalyst for a radical shift in the way artificial intelligence is developed and used. Its open and affordable approach could revolutionize not only AI-related hardware and software, but also the way the world perceives its access and potential. Gelsinger noted that “The story of technology has always been about how lower prices expand industries.", and DeepSeek seems to perfectly embody this principle.