MENU


Sony, well known as a consumer electronics giant, has had to spend several years rebuilding itself after it fell prey to the undercutting pressure from other Asian companies and countries. No more Sony Walkman, no more Sony Viao computers and relative few Sony televisions.

Sony’s move into content in the form of films and games along with its leadership in CMOS image sensors have been key to that rebuilding.

And now it has had the smart idea of letting its image sensors evolve to become single-component imaging subsystems and mount its own software on top of that (see Sony plans to turn image sensors into a subscription platform).

It is a smart idea in a number of ways.

1) First and most important: the idea of performing the maximum amount of AI processing in the sensor component gives Sony the chance of owning a share of a tremendously rich and extensive set of applications in every sector that CMOS image sensors have been deployed, which is very many indeed. This condenses the data and produces useful information while minimizing communications and energy consumption and preserving privacy, which may be key in some applications.

2) Leading the market: While Sony is the market leader in image sensors it has the likes of Samsung and Omnivision hard on its heels (see Samsung, ST grow market share in strong image sensor market). In addition, there are Chinese image sensor vendors that are likely to rise up quickly undercutting its margin, so this is exactly the right time to be making a move and changing the game while it can. As market leader Sony is in a position to establish norms and standards for this sort of offering.

3) Customers’ customers: As a vendor of image sensor hardware Sony’s CMOS image sensor sales are dominated by five customers, the leading smartphone manufacturers. The loss of one or more of those customers to a rival selling something almost as good would have a severe impact on Sony. How long can Sony keep selling image sensors in Samsung smartphones? By adding a layer of software subscription, Sony gains potential access to its customers’ customers, reduces dependence on its big customers and, potentially, creates more stickability with those customers.

Next: Four?


4) PlayStation-style scalability: As well as opening up an additional customer base the same image sensors could have multiple applications and repeat business as the same image sensor is redeployed, whether that it is in the smartphone, in the doorcam, in security survellience, the automobile and on the PC and in the gaming console. In the gaming console the application of AI processing within the image sensor could literally be a game changer.

Sony’s vision is similar to its own PlayStation business model. There Sony can drop the price on PlayStations and make its money on the multiple gaming software titles it sells to owners. In image sensors Sony could sell the hardware at minimal margin knowing that the big bucks are to be made by licensing software that, hopefully, it will be able to provide to customers.

In addition, with most equipment now being in contact with the Internet, in many cases it could be possible to provide updates, upgrades and new sales over the air.

There is a fifth angle to this. The industry has been thinking about machine vision and the application of AI to image processing for many years. One of the main thrusts was that conventional image sensors were designed to make pretty pictures for human eyes and were not optimized for machine vision applications. Such applications might value dynamic range, energy consumption, hyperspectrality or other parameters above resolution.

Next: End run round startups


As a result, a number of startups – for example Prophesee, Synsense and Insightness – have looked at event-driven image sensors. Indeed, Sony has reportedly acquired a majority stake in Insightness AG (Sony acquires Swiss vision sensor firm).

While Sony’s approach does not exclude the use of more specialized image sensors packaged with an AI processor, it has to be recognized that such sensors will be expensive. In contrast, Sony has packaged a conventional high-volume CMOS image sensor together with a logic chip containing an image processor, DSP for machine learning and embedded memory and this has the advantage of lower cost.

The one disadvantage that Sony’s proposed business model has is that there is little to tie the customer to Sony, or prevent them from migrating to third parties that might be able to exploit the same hardware. Sony will need to build up an appropriately-shaped ecosystem to support is software subscription aspirations. It may be that ultimately the smartphone vendors or service operators will expect to own this business. But at the very least Sony’s first mover advantage will give reason for those primary customers to stick with Sony rather than defect to a lower-cost sensor maker.

Mobile phone service operators have always understood the commercial advantage of turning hardware provision into a long-running service subscription. It is to be expected as system-on-chips evolve into fully characterised hardware-software platforms more and more hardware vendors will be thinking about how they can make similar moves to those now being planned by Sony.

Related links and articles:

www.sony.net

News articles:

Sony plans to turn image sensors into a subscription platform

Samsung, ST grow market share in strong image sensor market

Swiss neuromorphic startup changes name, raises money

Swiss startup launches mega-neuron vision processor

Prophesee plans sensor to address automotive, consumer markets

Sony acquires Swiss vision sensor firm


Share:

Linked Articles
eeNews Analog
10s