It is tough at the top, and being at the top of the most hyped industry globally — let alone being the most valuable company in the world — 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.
While it digs in its heels for the foreseeable future, rivals are circling and wondering: could 2026 be the year that Nvidia’s powerful technology is knocked off its perch?
• Nvidia says AI boom is accelerating as revenues rise 62%
Certainly, the reigning champion is facing some formidable opponents.
In January came China’s DeepSeek, an advanced AI model allegedly trained on far fewer GPUs than its peers, knocking $600 billion off Nvidia’s share price. Then in November Google pumped out its latest AI model to rave reviews. Painfully for Nvidia, it was trained on Google’s bespoke chips — tensor processing units (TPUs) — which are not only effective but about half the price of Nvidia’s. This sent Google’s share price up 31 per cent in a month and Nvidia’s into reverse.
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.
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’s tech is not going anywhere soon.
In a pun fitting for a semiconductor analyst, Harold described the current wave of bespoke models made by Big Tech as a “chipping away” of Nvidia’s dominance, rather than a fatal blow.
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.
The custom-designed semiconductors, made in-house for specific purposes, have a narrow focus. They could “chip away” at Nvidia’s revenue by performing specific tasks, but most big companies will still rely on GPUs for what Harold calls “general capacity filler”.
Away from the custom silicon experiments of the Magnificent Seven tech stocks — Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla — traditional semiconductor rivals such as AMD are making strides in AI. The latter signed a big deal recently with OpenAI, 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) — a fast, cheap way of powering the use of language models. It is still a minnow.
When it comes to Nvidia’s 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.
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.
And Nvidia has stitched itself deep into the AI ecosystem, investing in some of the world’s 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.
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.
The acceleration of quantum computing, for example, could move some of the hardest work off Nvidia GPUs and onto quantum chips. Or computers could start being operated with photonics, using light instead of electricity, which might replace silicon with something entirely different.
But for now, while Nvidia’s feathers may be being gently ruffled, it is pretty secure in the tech pecking order.