Accelerator market shifts
Investor reports say Nvidia remains the dominant datacentre accelerator vendor while AMD is gaining ground through aggressive product launches and pricing. Smaller signals—like rising Vertiv interest and increased institutional positions in Micron—are also being reported alongside the chip conversation. (ibtimes.com.au) (ibtimes.com.au) (dailypolitical.com) (dailypolitical.com)
Nvidia still sets the pace in datacentre accelerators, but Advanced Micro Devices is widening the fight as 2026 spending shifts from one chipmaker to a broader stack of suppliers. (investor.nvidia.com) Nvidia reported $62.3 billion in fourth-quarter datacentre revenue on February 25, 2026, up 75% from a year earlier, and $215.9 billion in full-year revenue for fiscal 2026. The company said Blackwell systems drove the latest surge as cloud groups kept building artificial-intelligence clusters. (investor.nvidia.com) Advanced Micro Devices reported record 2025 revenue of $34.6 billion on February 3, 2026, with datacentre revenue of $16.6 billion, up 32% year over year. Chief Executive Lisa Su said the company entered 2026 with “rapid scaling” in its datacentre artificial-intelligence business, and the company’s investor site lists expanded partnerships with Meta on February 24, 2026, and Hewlett Packard Enterprise on December 2, 2025. (ir.amd.com 1) (ir.amd.com 2) These chips are the engines inside artificial-intelligence datacentres: Nvidia and Advanced Micro Devices sell accelerators that train models and answer user prompts faster than general-purpose server processors. When buyers add more of them, they also need more memory, more electricity conversion, and more cooling equipment in the same buildings. (investor.nvidia.com) (investors.vertiv.com) (micron.com) That is why the market shift is showing up beyond the chip names. Vertiv, which sells power and cooling gear for datacentres, said on February 11, 2026 that fourth-quarter organic orders jumped about 252% from a year earlier, backlog reached $15.0 billion, and 2026 sales are expected at $13.25 billion to $13.75 billion. (investors.vertiv.com) Vertiv has tied that demand directly to denser artificial-intelligence facilities. In January it said adaptive liquid cooling and higher-power designs were reshaping datacentre operations, and in June 2025 it said it was aligning new 800-volt direct-current products with Nvidia’s coming platform roadmap. (investors.vertiv.com 1) (investors.vertiv.com 2) Micron is the memory readout on the same story. Its newsroom said on March 18, 2026 that it had reported second-quarter fiscal 2026 results, and on March 16 it said it was in high-volume production of HBM4, the high-bandwidth memory stacked next to accelerator chips for artificial-intelligence workloads, designed for Nvidia Vera Rubin systems. (micron.com) High-bandwidth memory matters because accelerator performance is limited not only by compute power but by how quickly data moves in and out of the chip package. As Nvidia keeps the largest installed base and Advanced Micro Devices pushes more systems into the field, suppliers that feed power, cooling, and memory demand are getting pulled into the same capital-spending cycle. (micron.com) (investor.nvidia.com) (ir.amd.com) The next test is whether Advanced Micro Devices can turn product launches and customer wins into a bigger slice of 2026 deployments without slowing Nvidia’s revenue machine. For now, the clearest signal is that the accelerator race is no longer moving only Nvidia and Advanced Micro Devices shares; it is moving the equipment and memory companies around them too. (investor.nvidia.com) (ir.amd.com) (investors.vertiv.com) (micron.com)
