Artificial intelligence was still a small part of the cloud market in October 2017, but industry leaders expected it to become an important source of growth. Bernie Trudel, then chairman of the Asia Cloud Computing Association, argued that the two technologies were beginning to reinforce each other.
Cloud platforms gave developers access to computing power, storage, software frameworks and large datasets without requiring every organisation to build its own infrastructure. AI, in turn, created new demand for those resources.
The forecast was that AI could generate 10 percent of cloud revenue by 2025. Whether or not a single percentage captured the market precisely, the underlying relationship was already becoming clear.
Cloud Lowered the Barrier to Machine Learning
Training and operating machine-learning systems can require specialised hardware, large volumes of data and engineering expertise. Purchasing that capacity upfront is difficult for companies that are still experimenting.
Cloud services offered a more flexible route. Teams could rent infrastructure for a project, increase capacity when workloads grew and shut resources down when they were no longer needed.
That model did not make AI simple, but it reduced the cost of getting started. A developer could test a framework or model without waiting for a lengthy hardware procurement process.
Framework Choice Was Becoming a Service
The 2017 cloud market included competing open-source machine-learning frameworks such as TensorFlow and MXNet. Each had different strengths, and organisations needed to decide which software and infrastructure combination suited their data.
Cloud providers saw an opportunity to package that complexity. Instead of supplying raw virtual machines alone, they could help customers select frameworks, configure environments and manage training jobs.
This marked a change in what customers expected from cloud computing. Infrastructure remained essential, but value increasingly came from higher-level services that removed repetitive engineering work.
Data Was Part of the Platform
Trudel highlighted Common Crawl, a public archive containing large quantities of web data, as an example of the resources available to cloud users. Datasets of that scale could support research, language processing and other forms of analysis.
Access alone did not guarantee a useful model. Teams still needed to assess quality, provenance, licensing, bias and whether a dataset represented the problem they wanted to solve.
Cloud platforms could make storage and processing easier, while governance remained the customer’s responsibility. That distinction became more important as organisations moved from demonstrations to systems affecting real people.
Google Connected Open Source and Cloud Adoption
Google had already used open-source software to build influence among AI developers. TensorFlow gave researchers and engineers a widely available toolkit, while the company’s cloud services offered managed infrastructure for running related workloads.
The company was also applying machine learning internally. DeepMind’s work on data-centre cooling demonstrated that AI could help optimise the infrastructure supporting cloud services themselves.
That created a feedback loop. AI increased demand for cloud capacity, while cloud operators used AI to improve efficiency, scheduling and energy consumption across their facilities.
Competition Extended Beyond Early Leaders
Amazon Web Services and Google were prominent examples, but the expected market included many more technology companies. Apple, Baidu, Alibaba, Intel, Tencent and Facebook were among those anticipated to develop their own AI services or supporting platforms.
The competitive boundaries were not straightforward. Chipmakers supplied accelerators, consumer platforms produced data and applications, and cloud providers offered the environments where models could be trained and deployed.
Some companies competed across several layers at once. Control over hardware, frameworks, models and distribution could create powerful advantages, while customers worried about being locked into one provider.
General AI Services Were the Larger Prize
Early cloud AI products often focused on specific tasks such as image recognition, speech processing or prediction. These application programming interfaces allowed software teams to add a capability without training a model from scratch.
The longer-term ambition was a broader collection of reusable AI services. Providers wanted to supply tools that could be adapted to many industries, datasets and business processes.
Managed services could widen adoption, but convenience introduced dependencies. Customers needed to understand costs, data location, service reliability and what would happen if a provider changed its model or terms.
Growth Brought Infrastructure Questions
AI workloads placed new pressure on data centres. Training could consume substantial computing power, and production systems needed reliable performance at unpredictable levels of demand.
Cloud scale was well suited to these requirements, but capacity was not unlimited. Providers had to invest in chips, networks, cooling and electricity while improving how resources were allocated.
Efficiency therefore became both a commercial and environmental concern. A model that delivered business value with less computation could reduce operating costs as well as energy use.
The 2017 Forecast Identified a Durable Partnership
The prediction that AI would fuel cloud growth captured a structural shift already underway. Cloud computing was evolving from rented infrastructure into a delivery system for data, models and increasingly sophisticated managed services.
AI depended on the cloud’s scale and accessibility, while cloud providers gained a new class of demanding, high-value workloads. Each technology expanded the market for the other.
The most important change was not one revenue estimate. It was the movement of machine learning from specialised research environments toward an on-demand service available to ordinary development teams. That transition would shape both cloud competition and enterprise AI adoption for years to come.
