Microsoft is developing a network card to improve the performance of its server artificial intelligence chip, Maia, assures Reuters.
To that end, Microsoft CEO Satya Nadella recently recruited Pradeep Sindhu, co-founder of networking equipment maker Juniper Networks. Microsoft also acquired Funzible, Sindhu’s startup specializing in server chips.
Network cards are hardware used to send and receive computer signals at the network layer. Nvidia offers cards of this type, the ConnectX-7, with its graphics processors (GPU). The development of this equipment by Microsoft could take more than a year. If successful, Microsoft could help ChatGPT developer OpenAI reduce the time needed to train models, and therefore reduce costs.
Reuters requested confirmation of this information from Microsoft, but did not receive a response.
It is now on the infrastructure field that the battle for artificial intelligence is moving
Microsoft is well positioned in the explosive AI market following its billion-dollar investment in OpenAI. This recently enabled it to integrate this technology, via its Copilot tool range, into its own products.
But it is now on the field of infrastructure that the battle for artificial intelligence is moving. Microsoft entered the race against Nvidia with the launch of the Maia 100, a GPU that runs large language models (LLM).
Nvidia’s stock has been hitting record highs in recent months, thanks to an unprecedented surge in demand for its graphics processing units (GPUs) driven by strong demand for artificial intelligence products. Last week, the company overtook Alphabet to become the third-largest U.S. company by market capitalization.
“Microsoft, Amazon, Meta and Google plan to significantly increase their investments in AI infrastructure this year”
For the fourth quarter of 2023, Nvidia just announced record revenue of $22.1 billion, up 22% from the third quarter and 265% from the previous year.
Record annual revenue for 2023 is $60.9 billion, up 126%.
“All four (Microsoft, Amazon, Meta and Google) plan to significantly increase their investments in AI infrastructure this year, which bodes well for Nvidia’s Q4 and Q1 earnings guidance 2024. But it is not certain that the four companies will reduce their investments to meet demand,” said Gil Luria, analyst at DA Davidson.
“Other analysts question whether Nvidia has enough supply to meet near-term demand,” notes CNBC. A new high-end NVIDIA chip, the B100, is due to start shipping this year.
Source: “ZDNet Korea”