New Course at OpenAI: Broadcom Partnership to Overcome Chip Shortages

Sam Altman's prolonged complaint about Nvidia chip shortages, which have delayed the introduction of new ChatGPT features, may now come to an end. OpenAI announced a remarkable partnership with Broadcom to develop custom chips worth $10 billion. This alliance reveals the new reality: companies are eager to escape the pricing dictate and dependency known as the "Nvidia tax." Although Nvidia chips dominate AI training, specialized solutions in the inference field are increasingly gaining importance. Inference processes, which occur with every use of ChatGPT or similar services, could significantly reduce costs and energy usage, promising improvements in performance capacity. These developments illustrate the shift toward a broader chip market environment beyond a monolithic provider. A significant step towards independence is that companies like Positron promise chips that are two to three times better per dollar, while Groq claims to significantly enhance ChatGPT's speed. These promises are transforming the market into a dynamic field where big names like Amazon and Microsoft also want to reduce Nvidia dependence by developing their own inference-focused chips. While Nvidia remains strong and advances new technologies such as the Blackwell system, the market offers opportunities for innovation. Big players and newcomers are strategically moving to secure market shares— a scenario that outlines multi-dimensional growth in the chip sector over the coming years and marks a shift in technological and economic balance. This competition could bring far-reaching benefits for the AI industry and beyond.