Specialist
Former executive at Nvidia Corp
Agenda
- Evolution of Nvidia’s (NASDAQ: NVDA) Cuda library
- Trajectory and general roadmap, noting optimisation of Cuda cores
- Demand outlook, including application-specific and end market growth opportunities
- Potential market share erosion if another player develops compatible binary-GPU (graphics processing unit) solution that can run Cuda code
Questions
1.
How do you assess the trajectory of computing in the industry? What’s important to keep an eye on?
2.
Could you take us through the evolution of Nvidia’s Cuda platform? Could you highlight its significance?
3.
You mentioned HPC [high-performance computing] and the growing need for simulations. What other industry technological trends are you seeing that would impact Cuda’s growth directly?
4.
How has Nvidia more or less consistently captured and locked in customer demand and dollars with Cuda?
5.
We’re starting see other companies – as you mentioned, competitors such as AMD – realise the importance of building their own ecosystems and the sustainability they could find within this. Could you elaborate on this trend? How far ahead of competitors does Nvidia remain with Cuda? Is it 1-2 years?
6.
What are your adoption expectations in the general parallel computing market over the next 3-5 years if competition slowly trickles in?
7.
Do you see any fundamental outstanding issues with Cuda, whether in the SDK [software development kit] or elsewhere?
8.
Cuda is essentially free – anyone can download it. The revenue derives from GPU [graphic processing unit] hardware sales as well as integrating the software. How do you perceive Nvidia’s revenue split between hardware and software?
9.
For what reasons could Nvidia start to charge for its software and AI models?
10.
Could you break down the advantages of the Cuda cores specifically? Which applications do you think will require their use as we look across the rest of the cloud data centre or enterprise ecosystem, thinking about a Google TPU [Tensor Processing Unit], an Amazon Graviton or an Intel x86 CPU [central processing unit]?
11.
You discussed application-specific drawbacks of the different options. What are the benefits and potential shortcomings of the Hopper architecture that we’ve already discussed?
12.
We’ve seen US-China geopolitical tensions build up throughout the past year and more. How do these tensions aggravate the uncertainty of locating and receiving computing power supply for data centres and computing platforms generally?
13.
What will it take for Intel or AMD to step up and challenge Nvidia’s dominance in the accelerated data centre space?
14.
What kind of challenge or threat is Intel’s entry into the GPU market and what we’re starting to see with its Meteor Lake developments?
15.
What’s your outlook on Nvidia’s developments within Cuda and parallel computing?