September 16, 2016

How big is “cloud-scale”?

35,000 CPU, or ~10,000 laptops - all available within 10 min, enabling the most advanced exploration methods for new materials Can we do better?

35,000 CPU, or ~10,000 laptops - all available within 10 min, enabling the most advanced exploration methods for new materials

Can we do better?

What do you think about when you hear the word “cloud”? Do you imagine an atmospheric phenomenon, an icon on your smartphone, or endless racks of interconnected servers? Likely — the latter.

Our goal at is to put this tremendous amount of computing power and all the brilliant minds behind it to work for the benefit of humanity. And yes, of course, it already benefits us all while storing party pictures, documents, text messages, contact information, recent trips and more. But can we do better?

Can we use cloud to solve humanity’s most important problems? How about finding better ways to capture carbon dioxide to prevent global warming? New ways to travel to space to prevent overpopulation? Affordable solar panels and long-lasting batteries to switch to sustainable energy sources? New materials for semiconductor industry to smear out the end of Moore’s law?

Yes we can.

Everything around you is built out of atoms and molecules. If you were to become billion times smaller you would see how these atoms combine together to form a material. And it is through the way by which these atoms interact with each other that a material acquires its characteristics. It could be hard as steel or kevlar, conduct electricity like copper or silver, or even cure diseases like penicillin. What if we could use the cloud to predict this?

Predicting the properties of materials from fully atomistic perspective requires many factors, but without doubt it needs computational power. There are 1 million billion billion atoms in just 1 gram of all common substances, like silicon. Although one does not always need to keep track of all of them to predict the properties or a resulting material, you can still imagine that the complexity is very substantial.

That’s why one needs computational power. The more — the better. And that’s why we at are incorporating a “cloud-scale” compute platform into the foundation of our product.

So what is “cloud-scale”?

It turns out, however, that everyone understands “cloud-scale” differently. For some it is a multitude of small programs, like web-servers, running on a remote computer. For others it is a workstation with 8 CPU available through a remote client. For us at, as the video below demonstrates, “cloud-scale” now is 35,000 CPU. Imagine a football stadium with 8,750 people who each brought a laptop with them. This is the scale.

Cloud providers have “unlimited” resources, however, scalability is not only determined by the sheer numbers, but also by how the control and data flow is designed. Implementations that do not consider this can either lead to low performance or high costs. By analogy: the football stadium crowd could be intensely watching the field, or wander around, like during breaks.

Now the beauty is that you can have compute power whenever you have the need: when you get the right idea to test. At that time you can pay a few thousand dollars to have a football stadium of people with laptops to work for you for a few hours. Without needing to care about how all these people got to their seats (buying hardware) or will get back home (maintaining it).

To learn more about our platform in action, take a look at the video below: