Qualcomm targets Nvidia in lower-cost AI inference

Lupa described a move that deserves attention far beyond the semiconductor sector. Qualcomm, long associated mainly with smartphones and mobile networks, wants to compete with Nvidia in the operation of artificial intelligence. Not in model training, where the field is already largely divided, but in inference, the stage when a model actually serves users or enterprise applications. That is where the battle over everyday AI deployment in companies, cloud platforms and enterprise systems will be decided. This matters because the inference market is commercially different from the training market. While only the biggest labs and hyperscale operators can afford to train large models, inference concerns a much wider set of customers. If a company runs a customer-service chatbot, a translation tool, an internal assistant or a document-analysis system, it cares not only about performance, but also about the cost per query, energy consumption and chip availability. Qualcomm is therefore not trying to beat Nvidia in its strongest area, but to outflank it where spending will be broad and long-term. Lupa also notes that Qualcomm had previously bet on distributed edge AI, for example around mobile towers, yet computing power ended up concentrating in huge data centers instead. The company is effectively admitting that its earlier vision did not materialize and is changing course. For investors and customers alike, this signals that AI infrastructure is entering another phase: less ideology, more pragmatism. The winner will not be the one that talks the loudest about the future, but the one that lowers operating costs. From a regulatory and geopolitical perspective, this is good news for customers as well. The more strong suppliers emerge in AI chips, the less dependent the market becomes on one dominant player. That can help restrain prices, improve the bargaining position of cloud companies and indirectly benefit European businesses that today mostly watch the American race from the sidelines. If Qualcomm succeeds, it will not just be a technical novelty. It will put pressure on the entire economics of AI operations. Source: Lupa.cz

Qualcomm targets Nvidia in lower-cost AI inference

The full article is available to users registered at Hard Skills