TPU

These days everyone knows that NVIDIA is the leader in global production of graphics processing units (GPUs). The company is making the news this week after announcing remarkably higher earnings over the first quarter of 2017. They destroyed the estimates from Wall Street by posting a 48% gain over last year. The question, of course, is whether the chip seller can maintain their market dominance with other players beginning to join the marketplace with the emergence of new technologies such as TPU.

NVIDIA is a thriving business these days. Most of that business is still coming through the gaming sector, with large growth from 2016. However, with the proliferation of AI technology, the need for GPUs in other sectors such as data centers is becoming greater. The substantial rise in this market place year over year (186% growth) shows that the demand for chips of this capacity and quality is burgeoning and the market supply is not keeping up.

When a huge demand is met with a small supply, other manufacturers start to take no- tice. Enter Google, IBM and Microsoft. All three companies are working at new and custom made chips that will be better at dealing with the demands of AI than the GPUs of today are. Not only are these three working on custom solutions, but other large tech companies are looking for ways around the GPU problem.

Google also published a blog article last month providing some of the details on their coming chip called the TPU, or tensor processing unit. The chip must go through a training phase, but after the training phase, Google is saying that the chip will be between 15 and 30 times faster than the current fastest GPU model on the market place. Further, they have promised that the chip will be a 30-80x improvement over the energy usage of current chips. The entire system is built on their open source code TensorFlow, which is designed to function as a machine learning framework. Undoubtedly, as the TPU comes online and the benefits of the chip are tested on the broader marketplace, NVIDIA will see a drop in sales for this sector. Microsoft is also working on its own custom made chip capable of machine learning, and IBM has already promised that they will be releasing something similar.

However, NVIDIA is seeking to diversify by looking for other sources of business revenue. They’ve recently signed a long term partnership deal with Bosch in order to write and build an entire computing management system for the autonomous cars which seem to be making all the headlines. They’ve promised that their coming hardware will be able to produce a 100% autonomous car before the close of 2018.

Ultimately, the chip game is going to get competitive in a new and unprecedented way. With big market players like Google and IBM on board, there’ll no doubt be a massive surge in available chips alongside the TPU which will certainly change the face of computing. These companies are designed for rapid innovation and will no doubt produce excellent products for companies and consumers alike.

LEAVE A REPLY

Please enter your comment!
Please enter your name here