By Tarun Dua, Managing Director at E2E Networks Ltd
Infrastructure-as-a-service market has been on a double-digit growth path in the last few years. In 2022, the worldwide IaaS market crossed $100 billion for the first time, growing from $92.8 billion in 2021 to $120.3 billion in 2022. By 2032, it is expected that this number will touch $350 billion.
These numbers aren’t surprising. The key reason behind this growth is the value that IaaS platforms offer to businesses in an increasingly digital landscape.
IaaS is essentially a vital piece of technology that provides organizations with on-demand access to virtualized computing resources such as cloud GPUs, compute resources, storage, database-as-a-service (DBaaS), and other fundamental building blocks that power cloud applications and IT infrastructure.
A potent aspect of IaaS is the pay-as-you-go pricing model, resulting in massive cost optimization. This enables businesses to scale resources up or down on demand, without having to worry about upfront investments in infrastructure.
As an increasing number of businesses have undergone digital transformation, or taken the step towards moving their legacy infrastructure to the cloud, IaaS platforms have been powerful enablers and grown along with these businesses.
However, I believe that we are on the verge of witnessing a major inflection point, which will be driven by the emerging potential of artificial intelligence, machine learning and generative AI technologies, that will lead to further uptick in the IaaS market in the coming decade.
The Era of AI
The years 2022-23 will always be remembered as the year when AI went mainstream. In the second half of 2023, we are witnessing an unprecedented surge in the adoption of public large language models (LLMs). Similarly, other Generative AI technologies – such as image generation, audio generation, automated speech recognition (ASR) and speech to text – captured the imagination of everyone, from CXOs to LinkedIn and Twitter influencers, with hundreds of new platforms flooding the market.
Yet, even as public generative AI platforms have showcased the potential of AI to marketeers and leaders, their pitfalls have also come to light.
Public LLMs and generative AI technologies are built upon general purpose deep learning AI models which are trained on petabyte-scale public datasets, have billions of parameters, and constantly improve through user feedback and usage. Therein lies their pitfall for enterprises and businesses.
When trying to leverage them for internal use-cases, businesses often end up leaking confidential corporate information, a risk that businesses simply cannot afford to take. Furthermore, due to the potential biases that may be present in the public datasets on which they have been trained, businesses worry about introducing similar biases in their own processes and systems when adopting them in real-world scenarios. Last but not the least, these public generative AI technologies have been shown to hallucinate, often introducing facts and references that are completely imaginary.
The question then arises, how does a business leverage the incredible potential of generative AI and machine learning, while avoiding some of its pitfalls?
This is where private AI models come into play.
Private Enterprise-Grade AI
While public LLMs have grown, a growing number of research papers, open source LLMs, generative AI models, and public datasets have simultaneously been launched in public domain with commercial usage licence.
These open-source generative AI models enable developers and organisations to build private generative AI applications which have similar capabilities to the public ones, but without the pitfalls that come along.
For instance, on 25 May 2023, Technology Innovation Institute made the 40 billion parameter model Falcon 40B available for free under Apache 2.0 licence, which can be used for research, development and commercial uses. On 18 July 2023, Meta released the latest version of its large language model Llama2, free of charge for research and commercial use.
A similar trend is currently taking place in other domains of AI and machine learning, giving developers and businesses the opportunity to create their own private generative AI stack purely built on open-source technologies. They can be fine-tuned, trained on private company and market-specific data, and refined to work for use-cases that apply for the company.
The major advantage of a private AI stack is around privacy and security of enterprise data. Additionally, if trained on data sets that are specific to a domain, the probability of the AI model hallucinating reduces drastically. Since they avoid several of the pitfalls of public generative AI platforms, this is the key method through which enterprise adoption of AI would take place in the future.
IaaS and AI
As enterprises increasingly start testing and training AI workloads, they need access to advanced GPUs with top-notch specs and performance.
Given the cost implications of owning GPU hardware, the most favorable option for businesses is to leverage cloud GPUs offered by providers who have access to the latest and greatest of GPU models being launched. Additionally, cloud providers who specialize in handling AI/ML workloads ensure that they allow developers to take advantage of the best that the underlying architecture has to offer.
Furthermore, Indian businesses would additionally need to ensure that any sensitive data they train their model on resides on an infrastructure that offers them the peace of mind of adhering to regulations of Indian IT laws.
Future of IaaS
As enterprise-AI adoption starts to grow, infrastructure provided by IaaS companies will play a key role in providing them with the cutting-edge hardware, platform and technologies that enable them to build fast and leverage the underlying capabilities.
In addition to the capabilities IaaS platforms have already been offering, AI adoption will be a key driver of growth in the coming decade for IaaS companies. Already, by 2023, the demand for advanced GPUs like H100 and A100 has reached a point where supply is barely able to keep up.
In the coming future, as an increasing number of enterprises and startups look to leverage AI to enhance their customer experience, we will witness a surge in demand for AI-focussed cloud computing technologies that only AI-focussed IaaS companies can seek to offer. This would unlock new opportunities cloud platforms that we haven’t seen before.
