Training advanced AI models, like the ones that power chatbots and other smart tools, takes a huge amount of computing power. However, tech giant IBM (IBM) is working to make this process faster and more efficient by building special computer systems that can handle many tasks at the same time, such as training, testing, and running the models. Indeed, in early 2024, IBM introduced a powerful new supercomputer called Blue Vela, built with help from NVIDIA (NVDA) and Dell (DELL). It was designed to train IBM’s Granite models and was the first system to use IBM’s newest storage technology.
Elevate Your Investing Strategy:
Take advantage of TipRanks Premium at 50% off! Unlock powerful investing tools, advanced data, and expert analyst insights to help you invest with confidence.
It is worth noting that saving progress when it comes to training AI models, known as checkpointing, is a big challenge for researchers. If saving fails at any point, all work since the last checkpoint is lost, which wastes time and money. Interestingly, though, when IBM tested Meta’s (META) largest Llama model, which has up to 1 trillion parameters, it showed impressive speed. In fact, it could load its saved progress in just 23 seconds and save it in 37 seconds.
What’s more is that IBM achieved these results by using only 4% of its Blue Vela system, while the rest was still running other AI projects. This shows how powerful and efficient its technology is. Normally, training a massive model like this would take over 500 GPUs, but new AI methods are helping reduce that number. In addition, IBM’s newest AI system, which was built with CoreWeave (CRWV) using NVIDIA’s latest platform, was able to train a large Llama model in under 28 minutes, even before the full system was up and running.
Is IBM a Buy, Sell, or Hold?
Turning to Wall Street, analysts have a Moderate Buy consensus rating on IBM stock based on seven Buys, four Holds, and one Sell assigned in the past three months, as indicated by the graphic below. Furthermore, the average IBM price target of $297.33 per share implies 19.3% upside potential.
See more IBM analyst ratings
Disclaimer & DisclosureReport an Issue