The artificial intelligence (AI) boom has generated some of the most extraordinary financial claims in stock market history.

For accountants – trained to look past the narrative to focus on the numbers – several features of the current landscape warrant close and sceptical attention.

1. Circular accounting: profits built on recycled capital

Let’s start with what is, charitably, a novel accounting arrangement. Microsoft injected $13 billion into OpenAI – not as cash, but as cloud credits redeemable against Microsoft’s own servers.

OpenAI used those credits to train its models, and Microsoft recorded this usage as cloud revenue from a paying customer. The investor and the customer are, in substance, the same entity.

Corporate filings reveal that OpenAI and Anthropic together account for more than half of the $2 trillion cloud backlog held by Microsoft, Oracle, Google and Amazon.

Beyond this, three of the four so-called cloud computing ‘hyperscalers’ recorded large non-cash gains this year resulting from their AI stakes: Alphabet booked $36.8 billion in equity mark-ups on Anthropic; Amazon $16.8 billion; and Microsoft $5.9 billion over nine months.

Each new funding round, at a higher valuation, generates a further mark-up through the income statement.

Earnings look robust. The underlying economics are circular.

Accountants should be demanding far more detailed disclosure of what these backlogs represent in terms of durable, third-party commercial demand.

2. Forward purchasing: a one-off in the revenue line

The extraordinary capital expenditure surge has a less publicised driver: the front-loading of GPU [Graphics Processing Units] and infrastructure purchases ahead of US tariff increases and in anticipation of IT sector inflation.

This pre-buying has pulled forward demand, boosting both reported revenues at semiconductor suppliers and capex figures that the markets have treated as confidence signals, rather than inventory management.

As the effect reverses, the revenue comparatives risk looking considerably less impressive.

3. Productivity: where’s the beef?

The intellectual case for current AI valuations rests on productivity transformation but the macro data does not yet support it.

US non-farm business sector labour productivity rose just 0.3 percent in the first quarter of 2026, the weakest reading in a year.

A survey of close to 6,000 senior executives across four advanced economies found that while 70 percent of firms report using AI, 90 percent say it has had no measurable impact on productivity or employment so far.

Penn Wharton’s Budget Model estimates AI will increase US GDP by 1.5 percent by 2035 and about three percent by 2055.

This is a reasonable long-run case. It does not justify valuations priced for societal transformation.

4. Profitability: corporate v tech sector

The AI investment cycle has been lavish. Its impact on profitability outside the technology sector remains invisible in aggregate earnings data.

The productivity gains that are being promised to enterprise customers – and implicitly priced into the market – are not showing up in the numbers of the enterprises buying AI.

5. Extraordinary valuations: the SpaceX case study

SpaceX completed the biggest Initial Public Offering in history last month with the aim of raising $75 billion at a $1.77 trillion valuation.

The company’s stock price jumped 19 percent on its first day of trading, briefly surpassing Microsoft and Amazon in market capitalisation, before falling about 30 percent from its intraday peak within one week.

At $1.77 trillion, SpaceX trades at roughly 94 times trailing revenue – a multiple exceeding the most optimistic valuations seen during the dot-com bubble.

Research firm Morningstar’s discounted cash flow valuation came in at $780 billion, roughly 48 percent below the IPO price, with analysts describing the stock as “significantly overvalued”.

6. Hyperscaler stocks: the market is

starting to notice The Magnificent Seven shed roughly $2.3 trillion in market value in June alone, with Microsoft heading for its worst month since 2000.

Hyperscaler stocks as a group are down 9.3 percent in 2026. Group free cash flow fell 23.7 percent in 2025 despite strong earnings growth, and Amazon’s free cash flow is projected to turn negative this year.

The four largest hyperscalers are collectively guiding to approximately $725 billion in capital expenditure in 2026, up 77 percent from $410 billion last year.

The market has begun asking a simple question: when does this investment pay back, and with what rate of return?

The bottom line The history of transformative technology waves – railways, electrification, the internet – shows us that the technology eventually delivers on its promise while most of the early investors lose money.

The valuation premium is extracted before the productivity arrives.

For accountants reviewing tech-sector financials, the AI boom presents a distinct set of challenges: – Circular revenue arrangements that comply with standards while obscuring economic substance;

– Non-cash gains inflating reported earnings;

– Forward purchasing distorting comparatives; and

– Price multiples that price in a future the macro data has not yet begun to confirm.

“Buyer beware” is not a sophisticated analytical framework but sometimes it is the correct one.