Clarification:
GeForce is a gaming line, and some previous high end game cards had some extensions for doing compute work as they had the same die as their big-brother Quadros but simply with a different driver and lacking ECC and other redundancies (Specifically the 8800/9800/2?0, 470, 570, and 580 were popular for this use), but the focus is towards game performance.
Quadro is a professional graphics production line, this is specifically for video and photo use, but many have found other more general compute loads for it as well, even though it is not what it is designed for. While their specs are great, the core clocks are relatively low compared to GeForce, and due to the driver being written for a different load they tend not to game as well as a comparable (and much cheaper) GeForce card.
Tesla is a compute card line, very similar to Quadro but much bigger. This is for computing mass amounts of complex numbers, but more generalized than Quadro, and not specifically designed for video related tasks. These are support cards; They do not game, and they do not even have display outputs. While the article says they will not have enough ram, they will still likely have 4-16GB of ram and possibly more. When those amounts of graphics ram are not enough then you can begin to understand the types of people these cards are aimed at, and why the entry level price begins ~$2000, and goes up from there.