Here’s how the incredible success of Nvidia AI could lead to its loss

Nvidia (Nvda 5.27%)) was able to set up the wave of artificial intelligence (AI) launched by Chatgpt at the end of 2022 and dominate the ACC accelerators market, in part because the company had spent years building a solid base in accelerated computer. NVIDIA GPUs are useful not only for AI workloads, but also in the academic world and industry to manage simulations and perform other tasks with high computer intensity.

CUDA, the NVIDIA owner software toolbox which allows developers to exploit the power of the company Gpushas existed for almost two decades. The vast ecosystem around Cuda was a huge advantage for Nvidia when the AI ​​boom started. The technology giants and start-ups moving quickly to advance in the AI ​​race, the NVIDIA GPUs were the road to the slightest resistance.

The tide begins to turn

When the GPU market in the data center was much smaller, it made no sense that Nvidia customers spend the engineering resources necessary to free themselves from their dependence on Cuda. The return on investment was simply not there.

With the AI ​​boom, causing an explosion of GPU demand from the data center, mathematics are starting to change. The NVIDIA data centers segment has generated more than $ 35 billion in income in the last quarter only, and the company’s beneficiary margins are via the roof. Customers love Microsoft,, Meta-platformsAnd Openai spends billions, or even tens of billions of dollars each year on the NVIDIA GPUs while they run to meet the IA computer capacity.

The stakes are now much higher. A company like Meta could save billions by removing Nvidia and designing its own AI chip and its software stack from zero. This push to bring internal accelerators seems to take steam in the technological industry.

Some companies, including Amazon And Alphabethave already created several generations of personalized AI chips. Other companies now follow suit. Meta would have tested a personalized AI chip for workloads in AI, which are much more intensive in calculation than the workloads on AI. Openai is not far behind, which would have worked on the finalization of its first design of Personalized AI fleas.

Companies that design their own AI chips and reduce the dependence on the expensive AI accelerators of NVIDIA could have a major cost advantage over competitors who do not follow the same path. This could force even more companies to provide internal AI fleas.

Erode the domination of Nvidia

While Nvidia has some competition from Dmla On the AI ​​accelerator market, its own customers represent a much greater threat. The figures are now large enough to encourage giants of technology and AI companies to rethink their hardware and AI software batteries. Beyond the reduction in the NVIDIA token demand, this trend could exert pressure on the prices and margins of Nvidia.

It is only because the GPU market in the data center has become so important that the CUDA grip on accelerated computers is attacked. Nvidia always sells a lot of ai chips, and she will continue to do so as long as the AI ​​boom continues. Once the IA computer capacity demand cools – the trees do not develop in the sky, after all – growth will be difficult for Nvidia because market share Eroded due to competition from local AI chips.

Nvidia’s actions have already dropped from its summit of all time, because the concerns concerning the resilience of AI demand collided with concerns about prices and the state of the global economy. While personalized AI chips constitute an increasing part of the ACC Accelerator market, the stock could undergo additional pressure.