Nvidia forecasts 70% sales growth as AI infrastructure boom accelerates

Nvidia has issued one of its most aggressive growth forecasts to date, projecting a 70% year-on-year sales increase for its next fiscal year as global demand for artificial intelligence infrastructure shows no signs of cooling. The projection, announced alongside the company’s latest earnings, follows a record-breaking quarter in which revenue reached $96.2 billion, nearly double the $48.7 billion reported in the same period a year earlier. The company said it expects $108 billion in revenue for the current quarter, a figure that would have been unthinkable just two years ago and one that underscores the scale of the AI build-out now underway across the global economy.

The forecast matters because Nvidia has become the clearest barometer of the AI economy. Its chips are the foundation upon which most advanced AI models are trained and deployed, and its financial results are scrutinised by investors, policymakers and corporate strategists worldwide. When Nvidia signals that demand is still accelerating, it suggests that the broader AI build-out—spanning cloud providers, startups, sovereign AI programmes and enterprise adoption—remains in an expansionary phase. But the same results also expose growing tensions in the supply chain and raise questions about how long the current capital expenditure cycle can be sustained. The company’s market capitalisation stands at $5 trillion, reinforcing its status as a bellwether for the entire AI economy and amplifying the significance of every data point in its earnings report.

Data centre dominance and the limits of supply

Nvidia’s data centre segment, which encompasses the high-end GPUs and networking equipment used for AI workloads, generated $89 billion in revenue during the quarter, a 117% increase from the prior year. That segment now accounts for the overwhelming majority of the company’s total sales, reflecting the extent to which Nvidia has transformed from a graphics chip maker into the primary supplier of computational power for the AI era. Chief Financial Officer Colette Kress was explicit about what is holding the company back: not demand, but supply. She said demand is expected to double, and that supply constraints remain the primary limit on growth. The statement is a rare admission from a company of Nvidia’s scale that its biggest challenge is not finding customers but physically producing enough chips to meet their orders.

The bottleneck is not only in advanced logic chips. Nvidia’s gross margin is expected to dip from 75% to 71% in the near term, a decline the company attributes to memory shortages. High-bandwidth memory, or HBM, is essential for AI accelerators, and the supply of HBM and related DRAM components has tightened dramatically as memory makers struggle to keep pace with demand from Nvidia and its competitors. The squeeze is also visible in Nvidia’s free cash flow, which fell to $21.3 billion from $48.6 billion a year earlier, partly because the company has extended customer payment terms to secure long-term orders. That shift suggests Nvidia is using its balance sheet as a strategic tool to lock in demand, but it also means the company is carrying more financial exposure to its customers at a time when the durability of AI spending is still being tested.

A broadening ecosystem beyond early leaders

Chief Executive Officer Jensen Huang sought to frame the current moment as a structural shift rather than a cyclical peak. He highlighted a broadening ecosystem of AI labs and startups that are now building on Nvidia’s platform, moving beyond the early dominance of companies like OpenAI. The implication is that demand is becoming more diversified, with new entrants in healthcare, robotics, autonomous systems and scientific computing all competing for compute capacity. Huang’s framing is significant because it addresses one of the most persistent concerns about Nvidia’s growth: that it depends on a small number of hyperscale customers who could abruptly reduce their spending. By pointing to a wider range of buyers, Huang is arguing that the demand base is now deep enough to withstand shifts in any single customer’s strategy.

Nvidia is not merely a supplier to this ecosystem; it is also a major investor in it. The company has made significant investments in OpenAI and Anthropic, and has participated in financing arrangements that have drawn scrutiny from analysts and commentators. One such arrangement involves a $500 billion chip financing consortium, while another provides a $100 billion backstop for an OpenAI data centre. Critics have labelled these structures “circular financing,” arguing that Nvidia is effectively helping to fund the customers that buy its chips, thereby inflating demand signals and masking the true economics of AI infrastructure spending. Kress pushed back against that characterisation, stating that the demand is real and that the risks associated with these arrangements are limited. The debate is unlikely to be resolved quickly, but it has become a central point of contention for investors trying to assess whether Nvidia’s growth is built on genuine end-user demand or on a self-reinforcing loop of vendor-financed purchases.

Whether or not the financing structures are problematic, they underscore a fundamental shift in how AI infrastructure is being funded. The capital intensity of frontier AI development has grown so large that even well-capitalised cloud providers and startups are seeking new financing mechanisms. Nvidia’s willingness to participate directly in those mechanisms gives it unusual influence over the pace and direction of AI deployment, and it means the company’s financial health is increasingly intertwined with the commercial success of its largest customers. For international executives and investors, this entanglement is both a source of confidence—Nvidia has every incentive to ensure its customers succeed—and a source of risk, since a slowdown in AI adoption would hit Nvidia not only through lower chip sales but also through the value of its investments and financing commitments.

The memory boom: Samsung, SK Hynix and Micron

The AI-driven chip boom extends well beyond Nvidia. Samsung Electronics is projected to post an 18-fold year-on-year jump in operating profit to 86 trillion won, approximately $72.6 billion, for the second quarter of 2026, according to analyst estimates cited in regional business media. The surge is being driven by soaring demand for memory chips, including HBM, DRAM and NAND flash. Prices for these components rose between 44% and 53% quarter-on-quarter, and analysts expect supply to remain tight through 2027. The scale of the profit increase is extraordinary even by the standards of the semiconductor industry, which has historically been characterised by sharp boom-and-bust cycles.

The market capitalisation of the three largest memory makers—Samsung, SK Hynix and Micron—has now surpassed $1 trillion each. That milestone reflects a profound change in the semiconductor industry’s centre of gravity. Memory, long considered a commoditised and cyclical business, has become a strategic bottleneck for AI systems. The tight supply of HBM in particular is now a limiting factor for Nvidia’s own production, creating a feedback loop in which memory makers and logic chip designers are mutually dependent on each other’s expansion plans. This interdependence means that capacity decisions made by Samsung, SK Hynix and Micron will directly shape how many AI accelerators Nvidia can ship over the next two years, and by extension how quickly AI services can scale across the global economy.

For procurement executives and technology leaders, the memory shortage is not an abstract concern. It is already affecting pricing, lead times and product roadmaps across the industry. Companies that depend on AI accelerators for their own services may need to plan for longer hardware delivery cycles and higher component costs well into 2027. The tight supply also gives memory makers unusual pricing power, which is reflected in the sharp quarter-on-quarter price increases. For businesses building AI infrastructure, the implication is clear: securing memory supply is now as important as securing GPU supply, and procurement strategies that treat memory as a commodity are likely to face significant disruption.

Risks beneath the surface

Despite the impressive headline numbers, several risk factors merit attention. JPMorgan has warned that AI’s growing share of cloud capital expenditure—projected to exceed 70% by 2027—may not be sustainable without clearer revenue returns from AI services. Cloud providers are spending enormous sums on AI infrastructure, but the revenue generated by AI applications has not yet caught up with the scale of investment in many cases. If that gap persists, some providers may slow their purchasing, which would ripple through the entire supply chain and hit Nvidia, memory makers and the broader ecosystem of AI startups simultaneously. The warning from JPMorgan is notable because it comes from one of the largest lenders to the technology sector, and it reflects a growing unease among financial institutions about the concentration of capital spending in AI.

Nvidia’s own financials also show signs of strain beneath the surface. The decline in free cash flow, the dip in gross margin and the extension of customer payment terms all point to a company that is managing a complex set of trade-offs. The company’s market capitalisation stands at $5 trillion, which means that even small disappointments in growth or margin trajectory could have outsized effects on investor sentiment. A company valued at that level is priced for sustained, exceptional performance, and any signal that the AI build-out is slowing—whether from supply constraints, customer concentration or financing concerns—could trigger a sharp repricing. Nvidia’s leadership is clearly aware of these risks, but the company’s ability to mitigate them is limited by factors outside its control, including the pace of memory capacity expansion and the spending decisions of a small number of hyperscale customers.

For international professionals, the message is nuanced. The demand for AI infrastructure is deep and structural, not a short-term speculative bubble. The breadth of the customer base, the scale of capital investment and the integration of AI into core business processes all point to a durable expansion. But the constraints are also structural: memory shortages, supply chain bottlenecks, capital intensity and concentration risk among a small number of suppliers. Executives and investors should monitor not only Nvidia’s revenue growth but also its gross margin, free cash flow and the health of its financing arrangements. Founders in AI and adjacent sectors can expect continued investment flowing into the ecosystem, but they will need to navigate rising component costs and a supply base that is increasingly concentrated among a handful of dominant players.

Looking ahead, the central question is not whether AI demand will continue to grow in the near term—it almost certainly will—but whether the industry can build the supply capacity and revenue models needed to sustain that growth over a longer horizon. Nvidia’s forecast of 70% sales growth is a statement of confidence, but it is also a test of the entire AI economy’s ability to absorb and justify investment at an unprecedented scale. The next two years will reveal whether the current expansion is the foundation of a new technological era or the prelude to a painful correction.