Intel In Talks To Buy AI Chip Startup Habana Labs: Reports
'The most important thing for them with their new products and any potential acquisition is that they develop a real significant three- to five-year road map for R&D, and they stick to it,' one Intel partner says of the report that Intel wants to expand its AI portfolio with another acquisition.
Intel is reportedly in advanced talks to acquire Habana Labs, an Israeli artificial intelligence chip startup that has previously raised funding from the semiconductor giant's venture capital arm.
The news emerged Tuesday morning in Israeli news publications Globes and Calcalist, with the former pegging the deal's value at around $1 billion and the latter saying it could be up to $2 billion. Both publications cited sources familiar with deal deliberations.
[Related: Intel AI Exec Gadi Singer: Partners 'Crucial' To Nervana NNP-I's Success]
Intel declined to comment while Habana Labs did not respond to a request for comment.
The Santa Clara, Calif.-based company's reported interest in acquiring Habana Labs, which has offices in Tel Aviv and San Jose, Calif,. comes as AI computing becomes a central focus for the chipmaker. The company recently said its AI products are expected to generate $3.5 billion this year, which would mark a 250 percent increase from the $1 billion figure the company reported last year.
Habana Labs is a fabless semiconductor company that makes AI processors for inference and training applications. Most recently, the startup launched its Gaudi training processors, which it has said can outperform GPU training workloads nearly four times over. Habana Labs, which launched its Goya inference processor last year, has raised a total of $120 million from investors, including Intel Capital.
Just a few weeks ago, Intel launched its Nervana neural network processors for deep learning training and inference. At a launch event, the chipmaker said its NNP-I1000 inference processors provide nearly four times the compute density of Nvidia's T4 GPUs while its NNP-T1000 training processors offer 95 percent scaling capabilities on important training models.
Marc Fertik, vice president of technology solutions at Elk Grove Village, Ill.-based Ace Computers, an Intel partner that is No. 261 on CRN's 2019 Solution Provider 500 list, said with Intel introducing new hardware like the Nervana neural network processors, the company needs to show that these products have staying power, unlike Intel products that have been discontinued or had their road maps cleared, such as Xeon Phi and Omni-Path Interconnect.
"If you go back when Intel had Xeon Phi, they had a lead for about a year and then as usual they got out of that market. Now they're going back in buying," he said.
Fertik said Intel's three core strengths have been in CPUs, solid state drives and networking hardware, so it would benefit the company to find another sustainable product line.
"The most important thing for them with their new products and any potential acquisition is that they develop a real significant three- to five-year road map for R&D, and they stick to it," he said.