May 16 (Reuters) - Ampere Computing on Thursday said it was pairing its chips with those from Qualcomm in a new offering aimed at lowering the power bills from operating artificial intelligence chips.

Founded by Renee James, a former president of Intel , Ampere uses technology from Arm Holdings to make central processing chips that are used by Oracle, Alphabet's Google and others. The startup has focused on making chips that are more energy efficient than industry leaders Intel and Advanced Micro Devices.

Qualcomm, which dominates the market for mobile phone chips, has been working to break into the market for AI chips in data centers since 2019 with a power-efficient offering of its own. Ampere and Qualcomm said on Thursday they have integrated their chips into a single data center server.

"Treat this as ... the first of the things that we're working on," said Jeff Wittich, Ampere's chief product officer.

"Obviously, can do things that are much bigger than the building server level solutions as we get out in the future given that we're both tackling a similar type of problem."

The joint Ampere-Qualcomm offering will not compete directly against AI chip leader Nvidia, whose chips are used to train AI systems with huge amounts of data.

The Ampere-Qualcomm servers are instead meant to run those models efficiently after they are trained.

Both Ampere and Qualcomm also compete indirectly with Nvidia because AI chips are often sold in systems that pair more than one kind of chip. The two working together can block potential rivals from gaining inroads with customers, said Jim McGregor, founder of Tirias Research.

"For both of these companies, it's about keeping their competition out of the data center," McGregor said.

Ampere on Thursday also announced the next generation of its central processing unit, which will have 256 processing cores, up from the current chip's 192. The new chip, due out next year, is being manufactured on Taiwan Semiconductor Manufacturing's 3-nanometer manufacturing process, Ampere said. (Reporting by Stephen Nellis and Max Cherney in San Francisco; Editing by Tom Hogue)