Kathy Gibson reports from AWS Summit in Sandton – AWS has already invested R15,6-billion in building infrastructure in Africa – and has committed to a further R30,4-billion.
“Whether it is opening new office locations, or enabling the distribution connect backbone, or setting up local zones, we have seen an acceleration in the drumbeat of investment on the continent,” says Chris Erasmus, country GM: South Africa at AWS.
“We see Africa as an incredibly strategic area of growth for AWS.”
Some milestones that the organisation has already achieved include a number of societal gains.
For instance, in 2019, AWS set itself a goal of powering or matching 100% of its electricity usage by 2030 – and this was achieved seven years ahead of schedule.
It also aims to be water positive by 2030 and is already 40% of the way towards that goal.
Erasmus points out that a lack of skills is a major stumbling block to growth in the technology industry. “So, in 2020, we announced we will enable 29-million people by 2025 – and have surpassed that goal a year early. In sub-Saharan Africa, we have already enabled 300 000 people.”
In 2022, AWS opened its Cape Town Skills Centre – just the second such centre in the world. “We have had more than 20 000 people walk in, get access to skills and certification, and now entering the job market.”
The is also investing in the development of SMEs and has committed $30-million to this.
AWS is positioning itself as the leader in the burgeoning cloud and artificial intelligence (AI) era.
AWS continues to expand its operations throughout the African and South African regions, says David Brown, vice-president of AWS Compute and Networking Services.
“We are always looking to regions with data centres to provide the highest levels of redundancy,” he says. “And the regions we have in-country support our customers efficiently.”
In South Africa, issues of latency often come into play, which AWS addresses through local zones, which can quickly deploy infrastructure geographically closer to the customer.
“We are looking at a number of local zones In South Africa, and also several in the rest of Africa.
“This gives us the capability to bring our technology to customers. We are committed to providing capacity and operating stability, and security is always critically important.”
Brown explains that, throughout history, the world has been fundamentally changed by a series of quantum changes: the industrial, digital and information revolutions.
In the latest era, the Internet is driving massive changes in the way we live and work.
It also provided the catalyst for Amazon to set up first its e-commerce business and later the technology organisation.
In the early days of Amazon, the company found itself facing technology challenges without available solutions.
“So we set about finding solutions for our problems – and we could then allow anyone else to use them.”
This is how AWS was born, and it continues to innovate, now offering more than 200 services that remove barriers to innovation for any users.
Not only has AWS created programmable APIs to all its service, it has also developed its own silicon and large-scale network topologies.
“The technologies we have built have allowed our customers to transform the world around us.”
Some examples include Netflix, which uses AWS to stream content to 190 countries. Moderna healthcare which completed its Covid vaccine sequencing in just two days using AWS and was able to ship a vaccine just 25 days later.
Closer to home, mPharma brings access to safe pharmaceuticals to patients in Africa.
“The cloud is setting us up to explore the next era,” Brown says. “And now we are on the precipice of another tectonic change. Generative AI (GenAI) will enable us to interact with and interpret data in profoundly new ways.
“AI is incredibly exciting,” he adds. “But it’s not just something that allows you to write poetry or complete homework. You can think about how it can bring deep reasoning to your applications, transforming the way we build products and businesses.
“As builders, GenAI gives us a new software component: the ability to reason in new ways, and to address the world’s largest challenges.”
Companies around the world are already developing new applications with intuitive and cost-effective features. Some notable examples of AWS customers include Lonely Planet, Aitable, Evolutionary Scale, Slack and Thomson Reuters.
In Africa, Hurone AI is transforming cancer care in Nigeria, Kenya, and Rwanda.
“We expect transformations from GenAI,” Brown says. “It is already automating complex jobs that previously required human thinking and reasoning.”
But developing AI and GenAI applications is complex, which is why AWS has done a lot of the upfront innovation on behalf of customers, Brown says.
“With AWS services, we can help you make what you deliver to your customer indistinguishable from magic,” he adds, referencing the Arthur C Clarke quote.
“In the fullness of time, GenAI will become vanilla and present in everything we do,” Brown says. “But today, and for the foreseeable future, we are going to have to lean in on GenAI – and we will have to learn together.”
AWS’s GenAI suite of applications form a three-tier stack: at the bottom is the infrastructure to training large language models (LLMs), in the middle tools to build and scale GenAI applications, and at the top are applications to leverage GenAI and LLM models with no specialist knowledge required.
When it comes to machine learning, there are two kinds of workloads training and inference – and both of these are energy-hungry.
“To make GenAI user cases economical and feasible, we need incredibly performant and cost-effective infrastructure purpose built with AI in mind,” Brown says.
Using the latest Nvidia GPUs, AWS has built an elastic fabric network that can provision ultra clusters of up to 20 000 GPUs. “AWS is the best place to run Nvidia GPUs.”
Indeed, Nvidia partnered with AWS to help it build its Project Sabre supercomputer, using 10 000 Nvidia Grace GPUs running at 414 exaflops.
“We have also been investing in our own custom silicon,” Brown says. “And this has already led to significant cost reductions.”
AWS training chips have helped to reduce the cost of training by up to 50% and inference ships have been sown to reduce inference costs by up to 40%.
They can also significantly reduce energy usage when training and inferencing.