Is DeepSeek's Breakthrough Really a Disaster For Nvidia Stock?
We are constantly reminded not to get too comfortable in the world of investing. When things seem to chug along on autopilot, an innovation shakes things up. A recent example is the late 2022 release of the original ChatGPT, which thrust artificial intelligence (AI) into the forefront and set off a race in an industry worth trillions of dollars. A breakthrough from a Chinese company called DeepSeek may be shaking things up again (or there may be more to the story).There are several layers to this onion. Here's what to know.DeepSeek is a Chinese tech company that created DeepSeek-R1 to compete with ChatGPT-4 and other large language models (LLMs), like Alphabet's (NASDAQ: GOOG) (NASDAQ: GOOGL) Google Gemini and Llama 3 created by Meta Platforms (NASDAQ: META). But that isn't the headline-inspiring story. DeepSeek "trained" its model with $6 million and just 2,000 somewhat outdated Nvidia (NASDAQ: NVDA) graphics processing units (GPUs). This is a startling claim when competing programs reportedly cost hundreds of millions of dollars and many thousands of top-shelf GPUs. For example, xAI's Colossus uses 200,000 GPUs, with plans to expand to 1 million.Continue reading
We are constantly reminded not to get too comfortable in the world of investing. When things seem to chug along on autopilot, an innovation shakes things up. A recent example is the late 2022 release of the original ChatGPT, which thrust artificial intelligence (AI) into the forefront and set off a race in an industry worth trillions of dollars. A breakthrough from a Chinese company called DeepSeek may be shaking things up again (or there may be more to the story).
There are several layers to this onion. Here's what to know.
DeepSeek is a Chinese tech company that created DeepSeek-R1 to compete with ChatGPT-4 and other large language models (LLMs), like Alphabet's (NASDAQ: GOOG) (NASDAQ: GOOGL) Google Gemini and Llama 3 created by Meta Platforms (NASDAQ: META). But that isn't the headline-inspiring story. DeepSeek "trained" its model with $6 million and just 2,000 somewhat outdated Nvidia (NASDAQ: NVDA) graphics processing units (GPUs). This is a startling claim when competing programs reportedly cost hundreds of millions of dollars and many thousands of top-shelf GPUs. For example, xAI's Colossus uses 200,000 GPUs, with plans to expand to 1 million.
