Nvidia's latest outlook offers one of the clearest indications yet that the artificial intelligence infrastructure boom has not reached its spending peak. The chipmaker expects revenue to grow by about 70% in its fiscal year ending January 2028, a projection far above the roughly 44% growth analysts had been expecting. More significant than the number itself is what Nvidia says is driving it: demand for computing capacity is expanding beyond a small group of technology giants into AI laboratories, cloud providers, businesses, governments and industrial users.
The forecast matters because Nvidia is effectively providing a forward-looking test of the durability of the AI investment cycle. For several years, the biggest cloud and technology companies have spent enormous amounts on data centres and advanced processors. Investors have increasingly questioned whether such spending can continue if the companies making those investments cannot demonstrate sufficient returns from AI services. Nvidia's latest numbers do not prove that the spending will remain profitable indefinitely, but they indicate that customers are still planning for substantially greater computing requirements rather than preparing to slow investment.
That distinction is central to understanding the next phase of the AI boom. The first wave was dominated by the construction of infrastructure needed to train increasingly powerful models. The next wave is increasingly about running those models at scale, integrating them into businesses and building new forms of AI that require continuous computing capacity.
Demand Is Broadening Beyond the Original Buyers
Nvidia's strongest evidence for continued growth is the changing composition of its customer base. Its data centre revenue reached $89 billion in the latest quarter, up 117% from a year earlier, while total quarterly revenue rose 106% to $96.2 billion. The company expects revenue of approximately $108 billion in the following quarter, showing that its growth remains exceptionally strong even after several years of rapid expansion.
The important change is that demand is no longer coming only from the largest cloud companies. Nvidia says AI laboratories are expected to contribute roughly a quarter of its overall business next year, while regional cloud providers, enterprises, government-backed computing projects and industrial customers are becoming increasingly significant. This broadening reduces the industry's dependence on a handful of hyperscale technology companies and creates additional sources of demand for advanced computing systems.
The expansion of Nvidia's relationship with Amazon Web Services illustrates the scale involved. The two companies plan to deploy another two million Nvidia GPUs across Amazon's global infrastructure during 2027 and 2028. That agreement follows an earlier plan to add more than one million Nvidia GPUs, which Amazon and Nvidia now say was overtaken by stronger-than-expected demand.
Such commitments suggest that AI infrastructure spending is increasingly becoming a multiyear capital programme rather than a short-term response to the launch of individual models. Companies are planning data centre capacity years ahead because building the electricity, networking, cooling and computing infrastructure required for large-scale AI takes considerable time.
The AI Boom Is Moving From Training to Production
The next stage of spending is also being shaped by a change in how AI is used. Early investment focused heavily on training increasingly large models. Businesses are now attempting to deploy AI systems continuously, including automated software agents, enterprise applications, scientific computing and physical systems such as advanced robotics.
This shift could make computing demand more persistent. A model that is trained once represents a large but relatively concentrated computing requirement. An AI system that is continuously answering users, executing tasks, analysing information or operating machines requires computing capacity every time it is used. As AI becomes embedded in ordinary business processes, the amount of computing consumed can therefore increase even if individual models become more efficient.
Nvidia is positioning its next-generation Vera Rubin platform around this broader workload. The company says the platform is already entering production and is being adopted by major cloud providers and other customers. Its strategy is also expanding beyond graphics processors into central processors, networking, memory and complete data centre systems. This allows Nvidia to capture a larger share of the infrastructure required to operate AI rather than depending solely on sales of individual processors.
That broader strategy may be increasingly important as customers develop their own chips. Amazon, Google and Microsoft are all investing in alternatives to Nvidia's processors. The existence of these competing systems means Nvidia cannot assume that every dollar spent on AI infrastructure will automatically become Nvidia revenue. Its response has been to make its technology ecosystem broader and harder to replace, combining processors with networking, software and complete computing platforms.
Supply Constraints Reveal the Scale of the Investment Cycle
One of the more unusual features of Nvidia's outlook is that the company is not warning about weak demand. It is warning that it may not have enough components to satisfy that demand. Memory shortages and rising component costs are already affecting the company, while limited supply is constraining how quickly Nvidia can increase production.
That creates a significant complication for investors. Strong demand does not automatically translate into unlimited revenue growth if suppliers cannot provide enough memory and other components. Nvidia expects gross margins to weaken temporarily as component costs rise, demonstrating that the AI infrastructure boom is creating pressure throughout the semiconductor supply chain rather than benefiting chip designers alone.
The supply problem is nevertheless revealing. Companies generally do not commit enormous amounts of capital to infrastructure unless they expect it to be used. Amazon's decision to expand its Nvidia GPU deployment after previously announcing a smaller programme is one example of customers increasing their infrastructure plans as AI workloads expand. Nvidia's own management has said customer demand is accelerating even at its existing scale.
At the same time, shortages should not be interpreted as proof that every planned AI investment will eventually generate attractive returns. Companies can build capacity faster than applications become profitable. The current evidence establishes strong demand for computing infrastructure; it does not establish that every AI service using that infrastructure will become commercially successful.
The Spending Cycle Is Becoming More Diversified
A major reason Nvidia can project another year of exceptional growth is that AI infrastructure is no longer confined to the largest American technology companies. Sovereign computing initiatives are encouraging governments and regions to build domestic AI capacity, while companies in finance, manufacturing, healthcare and automotive industries are deploying increasingly sophisticated AI systems.
The emergence of regional cloud providers is particularly important. These companies allow governments and businesses without enormous internal data centres to obtain access to advanced computing. Their growth effectively spreads AI infrastructure investment across a larger number of customers and geographic markets. Nvidia says this non-hyperscale portion of its data centre business is becoming increasingly important, suggesting that future growth may depend less exclusively on the investment decisions of a handful of technology giants.
India is also beginning to emerge as part of this broader infrastructure expansion. An AI infrastructure company backed by the founders of Greenko has ordered 9,000 Nvidia Vera Rubin processors for a planned AI facility in Hyderabad, with a broader commitment to invest billions of dollars in global AI computing infrastructure. Such projects illustrate how demand is spreading into markets that were not at the centre of the first generation of hyperscale AI infrastructure.
The broader customer base could make the AI spending cycle more durable, but it also makes the market more complicated. Enterprises and governments have different investment priorities from technology companies, and their adoption rates will depend on whether AI produces measurable improvements in productivity, research, automation or public services.
Nvidia Is Betting That Compute Will Remain the Scarce Resource
The deeper message behind Nvidia's forecast is that artificial intelligence is moving toward a stage where computing capacity itself becomes an economic input. The company's argument is no longer simply that increasingly powerful processors are necessary to train better models. It is that AI systems are becoming productive tools whose usage will create continuing demand for computing.
That thesis is supported by the scale of current investment. Nvidia's latest results show that data centre demand is still accelerating, while its Amazon partnership and expanding customer base point toward large infrastructure commitments extending well into 2028.
There are still substantial risks. Customers are developing competing processors, memory shortages can restrict Nvidia's ability to fulfil orders, and the economic returns from AI investment remain uneven across industries. Nvidia's exclusion of Chinese data centre revenue from its latest outlook also demonstrates how export controls and geopolitical restrictions can affect the company's addressable market.
Yet the immediate evidence points in one direction: the AI infrastructure cycle is broadening rather than contracting. Nvidia's 70% forecast is therefore less important as a promise about one company's future revenue than as a signal about how customers themselves are planning. They are still building for significantly greater AI computing requirements several years ahead. Whether that spending ultimately produces equally large economic returns remains unresolved, but the infrastructure race itself shows little sign of ending soon.
(Source:www.investing.com)
The forecast matters because Nvidia is effectively providing a forward-looking test of the durability of the AI investment cycle. For several years, the biggest cloud and technology companies have spent enormous amounts on data centres and advanced processors. Investors have increasingly questioned whether such spending can continue if the companies making those investments cannot demonstrate sufficient returns from AI services. Nvidia's latest numbers do not prove that the spending will remain profitable indefinitely, but they indicate that customers are still planning for substantially greater computing requirements rather than preparing to slow investment.
That distinction is central to understanding the next phase of the AI boom. The first wave was dominated by the construction of infrastructure needed to train increasingly powerful models. The next wave is increasingly about running those models at scale, integrating them into businesses and building new forms of AI that require continuous computing capacity.
Demand Is Broadening Beyond the Original Buyers
Nvidia's strongest evidence for continued growth is the changing composition of its customer base. Its data centre revenue reached $89 billion in the latest quarter, up 117% from a year earlier, while total quarterly revenue rose 106% to $96.2 billion. The company expects revenue of approximately $108 billion in the following quarter, showing that its growth remains exceptionally strong even after several years of rapid expansion.
The important change is that demand is no longer coming only from the largest cloud companies. Nvidia says AI laboratories are expected to contribute roughly a quarter of its overall business next year, while regional cloud providers, enterprises, government-backed computing projects and industrial customers are becoming increasingly significant. This broadening reduces the industry's dependence on a handful of hyperscale technology companies and creates additional sources of demand for advanced computing systems.
The expansion of Nvidia's relationship with Amazon Web Services illustrates the scale involved. The two companies plan to deploy another two million Nvidia GPUs across Amazon's global infrastructure during 2027 and 2028. That agreement follows an earlier plan to add more than one million Nvidia GPUs, which Amazon and Nvidia now say was overtaken by stronger-than-expected demand.
Such commitments suggest that AI infrastructure spending is increasingly becoming a multiyear capital programme rather than a short-term response to the launch of individual models. Companies are planning data centre capacity years ahead because building the electricity, networking, cooling and computing infrastructure required for large-scale AI takes considerable time.
The AI Boom Is Moving From Training to Production
The next stage of spending is also being shaped by a change in how AI is used. Early investment focused heavily on training increasingly large models. Businesses are now attempting to deploy AI systems continuously, including automated software agents, enterprise applications, scientific computing and physical systems such as advanced robotics.
This shift could make computing demand more persistent. A model that is trained once represents a large but relatively concentrated computing requirement. An AI system that is continuously answering users, executing tasks, analysing information or operating machines requires computing capacity every time it is used. As AI becomes embedded in ordinary business processes, the amount of computing consumed can therefore increase even if individual models become more efficient.
Nvidia is positioning its next-generation Vera Rubin platform around this broader workload. The company says the platform is already entering production and is being adopted by major cloud providers and other customers. Its strategy is also expanding beyond graphics processors into central processors, networking, memory and complete data centre systems. This allows Nvidia to capture a larger share of the infrastructure required to operate AI rather than depending solely on sales of individual processors.
That broader strategy may be increasingly important as customers develop their own chips. Amazon, Google and Microsoft are all investing in alternatives to Nvidia's processors. The existence of these competing systems means Nvidia cannot assume that every dollar spent on AI infrastructure will automatically become Nvidia revenue. Its response has been to make its technology ecosystem broader and harder to replace, combining processors with networking, software and complete computing platforms.
Supply Constraints Reveal the Scale of the Investment Cycle
One of the more unusual features of Nvidia's outlook is that the company is not warning about weak demand. It is warning that it may not have enough components to satisfy that demand. Memory shortages and rising component costs are already affecting the company, while limited supply is constraining how quickly Nvidia can increase production.
That creates a significant complication for investors. Strong demand does not automatically translate into unlimited revenue growth if suppliers cannot provide enough memory and other components. Nvidia expects gross margins to weaken temporarily as component costs rise, demonstrating that the AI infrastructure boom is creating pressure throughout the semiconductor supply chain rather than benefiting chip designers alone.
The supply problem is nevertheless revealing. Companies generally do not commit enormous amounts of capital to infrastructure unless they expect it to be used. Amazon's decision to expand its Nvidia GPU deployment after previously announcing a smaller programme is one example of customers increasing their infrastructure plans as AI workloads expand. Nvidia's own management has said customer demand is accelerating even at its existing scale.
At the same time, shortages should not be interpreted as proof that every planned AI investment will eventually generate attractive returns. Companies can build capacity faster than applications become profitable. The current evidence establishes strong demand for computing infrastructure; it does not establish that every AI service using that infrastructure will become commercially successful.
The Spending Cycle Is Becoming More Diversified
A major reason Nvidia can project another year of exceptional growth is that AI infrastructure is no longer confined to the largest American technology companies. Sovereign computing initiatives are encouraging governments and regions to build domestic AI capacity, while companies in finance, manufacturing, healthcare and automotive industries are deploying increasingly sophisticated AI systems.
The emergence of regional cloud providers is particularly important. These companies allow governments and businesses without enormous internal data centres to obtain access to advanced computing. Their growth effectively spreads AI infrastructure investment across a larger number of customers and geographic markets. Nvidia says this non-hyperscale portion of its data centre business is becoming increasingly important, suggesting that future growth may depend less exclusively on the investment decisions of a handful of technology giants.
India is also beginning to emerge as part of this broader infrastructure expansion. An AI infrastructure company backed by the founders of Greenko has ordered 9,000 Nvidia Vera Rubin processors for a planned AI facility in Hyderabad, with a broader commitment to invest billions of dollars in global AI computing infrastructure. Such projects illustrate how demand is spreading into markets that were not at the centre of the first generation of hyperscale AI infrastructure.
The broader customer base could make the AI spending cycle more durable, but it also makes the market more complicated. Enterprises and governments have different investment priorities from technology companies, and their adoption rates will depend on whether AI produces measurable improvements in productivity, research, automation or public services.
Nvidia Is Betting That Compute Will Remain the Scarce Resource
The deeper message behind Nvidia's forecast is that artificial intelligence is moving toward a stage where computing capacity itself becomes an economic input. The company's argument is no longer simply that increasingly powerful processors are necessary to train better models. It is that AI systems are becoming productive tools whose usage will create continuing demand for computing.
That thesis is supported by the scale of current investment. Nvidia's latest results show that data centre demand is still accelerating, while its Amazon partnership and expanding customer base point toward large infrastructure commitments extending well into 2028.
There are still substantial risks. Customers are developing competing processors, memory shortages can restrict Nvidia's ability to fulfil orders, and the economic returns from AI investment remain uneven across industries. Nvidia's exclusion of Chinese data centre revenue from its latest outlook also demonstrates how export controls and geopolitical restrictions can affect the company's addressable market.
Yet the immediate evidence points in one direction: the AI infrastructure cycle is broadening rather than contracting. Nvidia's 70% forecast is therefore less important as a promise about one company's future revenue than as a signal about how customers themselves are planning. They are still building for significantly greater AI computing requirements several years ahead. Whether that spending ultimately produces equally large economic returns remains unresolved, but the infrastructure race itself shows little sign of ending soon.
(Source:www.investing.com)
