What Nvidia actually reported

Nvidia (NASDAQ: NVDA) posted Q1 fiscal 2027 revenue of $44.1 billion, according to the company's earnings release dated 20 May 2026. That figure exceeded the analyst consensus of roughly $43.3 billion, as compiled by LSEG, representing year-on-year growth of approximately 69% from the $26.0 billion reported in Q1 FY2026.

The datacenter segment, which now accounts for the vast majority of Nvidia's turnover, was once again the principal driver. While the company did not break out a precise datacenter figure in its initial statement, analysts had projected the segment would surpass $39 billion for the quarter, according to estimates tracked by Bloomberg.

Jensen Huang, Nvidia's chief executive, described the current wave of infrastructure investment in characteristically expansive terms.

"The buildout of AI factories, the largest infrastructure expansion in human history, is accelerating at extraordinary speed. Agentic AI has arrived, doing productive work, generating real value, and scaling rapidly across companies and industries."

The results extended a run of consecutive earnings beats stretching back to early 2023, as reported by the Guardian. Nvidia's share price, which had already risen sharply over the preceding 12 months, moved higher in after-hours trading.

What the numbers say about AI infrastructure spending

Nvidia's revenue is widely treated as a proxy for the pace of global AI capital expenditure. The company's largest customers, the hyperscale cloud providers, have collectively committed to record spending programmes.

In their most recent quarterly earnings, the four largest buyers of Nvidia hardware each signalled sustained or increased investment:

  • Microsoft (NASDAQ: MSFT) guided to approximately $80 billion in capital expenditure for its fiscal year ending June 2025, with the bulk directed at AI-capable datacentres, according to its January 2025 earnings call.
  • Alphabet (NASDAQ: GOOGL), Google's parent company, disclosed plans for roughly $75 billion in 2025 capex, as stated in its Q1 2025 results.
  • Amazon (NASDAQ: AMZN) indicated AWS-related capital spending would reach approximately $100 billion in 2025, according to its Q1 earnings disclosure.
  • Meta Platforms (NASDAQ: META) raised its 2025 capex guidance to a range of $64 billion to $72 billion, per its most recent quarterly filing.

Taken together, these four firms alone have earmarked north of $300 billion for infrastructure this year, a substantial share of which flows directly or indirectly to Nvidia. The sustained pace of orders suggests that demand for GPU clusters has not yet peaked, even as each provider races to build out capacity for training and inference workloads.

Implications for UK operators buying AI compute

For UK scale-ups and mid-market firms, the relevant question is not Nvidia's share price but whether this capital flood translates into more accessible, more affordable compute.

There are reasons for cautious optimism. Intense competition among hyperscalers for enterprise AI workloads has already begun to compress pricing on inference-as-a-service products. AWS, Azure and Google Cloud have each introduced GPU instance pricing tiers that are 30% to 50% lower than equivalent offerings 18 months ago, based on published list prices across their respective platforms.

Domestic supply is also expanding. The UK government's AI Growth Zone initiative, announced in early 2025, designated sites in Culham, Oxfordshire and elsewhere for accelerated planning approval of large-scale datacentres. Private-sector commitments have followed: Microsoft confirmed a £2.5 billion investment in UK datacentre capacity, according to a company announcement in late 2024, while Amazon Web Services pledged £8 billion in UK infrastructure spending over five years, as disclosed in its own statement the same year.

These investments do not guarantee that a 50-person SaaS firm in Manchester will see GPU costs halve overnight. But they do increase the total pool of available compute within UK borders, reducing latency for domestic workloads and, over time, creating pricing pressure that benefits smaller buyers.

When cheaper inference might arrive

The economics of AI inference are shifting faster than many operators realise. Nvidia's own product roadmap plays a role: the company's Blackwell architecture, which began shipping in volume during the second half of fiscal 2026, delivers substantially higher inference throughput per watt than its predecessor. As hyperscalers deploy Blackwell-based clusters at scale through 2026 and into 2027, the per-query cost of running large language models is expected to decline further.

Several independent analyses support this trajectory. Research published by Epoch AI in early 2025 found that the cost of running inference on frontier models had been falling at a rate of roughly 10x every 18 months, driven by a combination of hardware improvements, software optimisation and model efficiency gains.

For UK finance directors weighing AI integration budgets, the practical takeaway is that the infrastructure buildout reflected in Nvidia's results is not an abstract Silicon Valley phenomenon. It is actively expanding the supply of compute available through major cloud platforms, and the competitive dynamics among those platforms are working in buyers' favour.

The pace of cost reduction is unlikely to be linear. Supply-chain bottlenecks, energy constraints and export controls on advanced chips all present risks. But the direction of travel, more compute, at lower unit cost, delivered through increasingly competitive cloud marketplaces, appears well established for the next 12 to 18 months.