The massive wave of global investment in artificial intelligence infrastructure may not translate into the biggest gains for the companies building the technology itself, according to investment expert James Disney-May.
While spending on AI continues to surge, Disney-May says investors should look beyond the obvious beneficiaries of the boom and focus on businesses already embedded in enterprise workflows where the real commercial opportunities may lie.
AI expert James Disney-May said: “The AI spending wave is real and it is enormous. Goldman Sachs estimates hyperscale capital expenditure will reach $527 billion in 2026, with some forecasts putting the total closer to $700 billion.
“Jensen Huang called it the largest infrastructure buildout in human history at Davos. He is probably right.
“And for two years, the obvious trade was to back the companies supplying the hardware and compute that made it possible – Nvidia, the hyperscalers, the data centre operators. Goldman Sachs’ infrastructure basket returned around 44 per cent at its peak. The market was correctly pricing the first and most visible layer of demand.”
Despite the scale of investment, Disney-May believes the market is starting to question whether the companies building AI models will ultimately capture the greatest financial rewards.
James said: “The scale of spending is extraordinary. But extraordinary spending does not guarantee extraordinary returns. The market is beginning to realise that the biggest gains may not sit with the model makers.”
He points to a growing mismatch between infrastructure spending and the revenues currently generated by leading AI developers.
OpenAI, widely seen as the most prominent consumer of AI infrastructure, recorded around $20 billion in annualised revenue last year. While impressive, Disney-May says that figure looks modest when set against the enormous capital investment flowing into the broader AI ecosystem.
He also highlights the arrival of DeepSeek, which demonstrated that a frontier-capable model could potentially be built at a fraction of the compute cost many investors had assumed.
Meanwhile, enterprise adoption of AI has accelerated rapidly. Spending by businesses tripled in just one year, rising from $11.5 billion in 2024 to $37 billion in 2025.
However, that spending is not primarily benefiting AI model developers.
James said: “The more interesting businesses are not necessarily the ones building the best models. They are the ones already inside the workflow, with the data, the distribution, and a practical route to monetising AI.
“That is a less dramatic investment case, but often a more durable one.”
He says sectors such as legal technology, financial compliance, healthcare software and procurement automation are already seeing strong demand for AI-driven tools because they sit directly within everyday enterprise processes.
One example is ServiceNow, which initially saw its share price fall by around 33 per cent amid concerns that AI agents could disrupt its business model.
Instead the company reported 21 per cent subscription revenue growth, secured 35 deals worth more than $1 million for its AI suite in a single quarter, and maintained strong full-year guidance.
James said: “The threat became a tailwind.”
Looking ahead, Disney-May believes the next evolution of AI technology will further reinforce this trend.
Agentic AI systems, which operate autonomously inside workflows rather than simply responding to user prompts, are expected to deepen the advantage for companies already integrated into enterprise operations.
James said: “Investors forget this in every technology cycle. The infrastructure layer can be essential without capturing all the value.
“Railways changed economies, but plenty of railway investors never saw the returns they expected. AI may follow the same pattern.”
While he believes the companies building AI models are creating extraordinary technology, Disney-May says investors searching for long-term returns may find better opportunities elsewhere.
James said: “The model makers are building something extraordinary. But the returns, when they come, are unlikely to be found there.”
