The TensorFlow machine learning framework that Google Inc. released under an Apache open-source license in late 2015 just gained an important endorsement: IBM Corp. said today that it will support TensorFlow 0.12 on its PowerAI machine learning platform.

IBM introduced PowerAI late last year as a general-purpose complement to its Watson natural language question answering computer, which IBM calls a “cognitive” processor. IBM is tuning and packaging Watson for different applications such as security analytics and financial analysis. In contrast, PowerAI is meant to be an open platform that anyone can use.

The move should be a boost for the rapidly growing machine learning market, since PowerAI is considered to be one of IBM’s most powerful systems. Machine learning, which is a form of predictive analytics, enables computers to become “smarter” by repeatedly churning through large sets of data and iterating on patterns that emerge. Popular applications include fraud detection, text analytics and speech recognition.

Based on the Power Systems S822LC and Power8 processor, the PowerAI platform also incorporates Nvidia Corp.’s NVLink processor using the high-speed NVLink interface to Nvidia’s Tesla Pascal P100 GPU accelerators. NVLink is a faster alternative to the Peripheral Component Interconnect bus that is standard in most personal computers and servers.

Google developed TensorFlow to run on a wide variety of platforms, including servers running multiple central processing units and GPUs. GPU acceleration enables tasks to be split away from main memory and processed in parallel on specialized processors. In theory, this permits almost unlimited scalability in high-performance computing clusters. TensorFlow is used in more than 50 commercial Google products, including speech recognition, Gmail, Google Photos, and search.

IBM said its Technology Support Services arm will introduce an enterprise software support offering specifically for the PowerAI stack. IBM Global Business Solutions has also established a deep learning design and development team to help customers build upon the PowerAI platform with TensorFlow and other frameworks.

IBM also announced support for the Chainer deep learning framework on the latest release of PowerAI. The platform already supports Caffe, Torch, Theano, Torch and several other machine learning libraries.