It is tough at the top, and being at the top of the most hyped industry globally \u2014 let alone being the most valuable company in the world \u2014 is tougher still. Often described as the maker of the shovels of the AI revolution, with its powerful graphics processing units (GPUs), Nvidia occupies that poisoned chalice of prime position in the tech world. <\/p>\n
While it digs in its heels for the foreseeable future, rivals are circling and wondering: could 2026 be the year that Nvidia\u2019s powerful technology is knocked off its perch? <\/p>\n
\u2022 Nvidia says AI boom is accelerating as revenues rise 62%<\/a><\/p>\n Certainly, the reigning champion is facing some formidable opponents. <\/p>\n In January came China\u2019s DeepSeek<\/a>, an advanced AI model allegedly trained on far fewer GPUs than its peers, knocking $600 billion off Nvidia\u2019s share price. Then in November Google pumped out its latest AI model to rave reviews. Painfully for Nvidia, it was trained on Google\u2019s bespoke chips \u2014 tensor processing units (TPUs) \u2014 which are not only effective but about half the price of Nvidia\u2019s. This sent Google\u2019s share price up 31 per cent in a month and Nvidia\u2019s into reverse. <\/p>\n Challengers include Amazon with its custom Trainium models, and even OpenAI, a longstanding Nvidia partner, is exploring making its own semiconductors, in partnership with Broadcom.<\/p>\n Yet experts like David Harold, from Jon Peddie Research, the leading computer graphics analyst, are unequivocal: with an 80 to 90 per cent market share, Nvidia\u2019s tech is not going anywhere soon.<\/p>\n In a pun fitting for a semiconductor analyst, Harold described the current wave of bespoke models made by Big Tech as a \u201cchipping away\u201d of Nvidia\u2019s dominance, rather than a fatal blow.<\/p>\n A risk now for Nvidia is that these hyper-scalers, making their very own chips, are its biggest customers. But Big Tech will not be able to pivot easily away from the empire of Nvidia boss Jensen Huang<\/a>. <\/p>\n The custom-designed semiconductors, made in-house for specific purposes, have a narrow focus. They could \u201cchip away\u201d at Nvidia\u2019s revenue by performing specific tasks, but most big companies will still rely on GPUs for what Harold calls \u201cgeneral capacity filler\u201d. <\/p>\n Away from the custom silicon experiments of the Magnificent Seven tech stocks \u2014 Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla \u2014 traditional semiconductor rivals such as AMD are making strides in AI. The latter signed a big deal recently with OpenAI<\/a>, but it still has a long way to go to catch up and market share is in the single digits. There are also start-ups such as Groq with its language processing units (LPUs) \u2014 a fast, cheap way of powering the use of language models. It is still a minnow.<\/p>\n When it comes to Nvidia\u2019s tech dominance, merely talking about hardware is too simplistic; most of its engineers are in software, working in areas such as industrial automation, robotics and digital twins.<\/p>\n In addition, it has created a toolkit that has captured the market: compute unified device architecture (Cuda). This interface between the hardware and the software allows developers easy access to the full power of the graphics card. Cuda has become the AI standard in research and commercial applications. It is the trusted go-to for most companies. <\/p>\n And Nvidia has stitched itself deep into the AI ecosystem, investing in some of the world\u2019s hottest start-ups, such as Wayve and ElevenLabs. It has also signed contracts that include vast purchases of its chips, most notably with OpenAI, raising eyebrows about circular deals and fears it is too big to fail.<\/p>\n The real tech challengers to Nvidia will be less obvious than the crop of rivals around it. More of a threat is the rise of new technologies that will enable artificial intelligence in a different way. <\/p>\n