Reading time: 8 minutes
The balance sheets tell the story of yesterday, but the markets speculate on that of tomorrow. While accounting continues to freeze assets at the price paid sometimes several decades ago, investors ignite around artificial intelligence, valuing start-ups to tens of billions before they even prove their profitability. Between accounting prudence and speculative fever, where is the truth?
When accounting freezes the past
Accounting is based on conventions designed to ensure the reliability and comparability of accounts. Among them, the principle of historical coste is probably one of the most emblematic: an asset is shown on the balance sheet for its acquisition price and severs over the years, regardless of its market valuee.
This simple rule, however, raises a disturbing question. How can we justify that a building purchased in the 1970s, located in the centre of a capital city, is now almost zero in the accounts, whereas it could be sold for millions?
The entrepreneur knows this well: his heritage is largely underestimated. But the accountant defends the prudence and security of this convention.
A security... but at the cost of an economic myopia
If the historical cost reassures, it is because it is based on a tangible proof: an invoice, a notarial act, a price paid. It also avoids a tax puzzle: No one would have to pay taxes on a purely virtual surplus value, as long as no cash was received.
Yet this caution has a setback. It artificially impoverishes the heritage image of companies and blurs the reading of their true wealth.
The investor who relies solely on the accounts sees only one narrower version of economic reality.
International standards have tried to remedy this by introducing the fair value for certain financial or real estate assets, but always sparingly. The fear is great that the balance sheets will turn into barometers of speculation.
When markets pack
For beyond accounting principles, there is another world: financial markets, dominated by anticipations, enthusiasm and sometimes collective blindness. History is marked by bubbles: that of tulips in the 17th century, that of internet values in the late 1990s, or the real estate bubble that preceded the 2008 crisis.
Each time, the same scenario: prices that fly away, disconnected from real economic flows, then a brutal collapse.
Accounting at historical cost, at those times, acts as a guard: it does not immediately recognize the bubble. But it later suffers the consequences, when it is necessary to depreciate massively excess values that have become untenable.
The fragility of excess values
These surplus values, called goodwill in the language of standards, are at the heart of the problem. They arise from acquisitions made at very high prices, justified by expected synergies or the supposed scarcity of the target.
As long as growth continues, it seems legitimate. But when the context changes – slowing down, rising rates, increasing competition – they turn out to be fragile.
The balance sheet, artificially inflated, must then be corrected by dramatic depreciations that erode equity and tarnish the credibility of management. Major groups that have embarked on acquisition waves know this: surplus value is a promise that can become a burden.
Artificial intelligence, new bubble in gestation?
Today, artificial intelligence concentrates all expectations. Start-ups manage to raise hundreds of millions, sometimes billions, without having yet demonstrated the viability of an economic model. Some companies valued at tens of billions have abyssal losses and colossal operational costs, linked in particular to the necessary computing power.
However, investors are pouring in, convinced to enter a new industrial revolution. The record fundraising of companies such as OpenAI, Anthropic or Mistral AI in Europe is a testament to this enthusiasm. But in many cases, medium-term profitability remains hypothetical, and reliance on cloud giants for access to infrastructure still limits model sustainability.
In 2025, OpenAI a lifted $8.3 billion in an over-subscribed lifting, which gave it a post-money recovery of $300 billion. Some media even speak of a planned total lifting of 40 billion to achieve the group's ambitions.
Yet, despite this surge, OpenAI's business model is prudence: costs are low, the need for colossal infrastructure, technical partnerships with cloud providers are heavy. In other words: the promise of a bright future today justifies a value never seen, but the operational results must follow to legitimize it. If the market turns around, the balance sheet will have to absorb the corrections.
Mistral AI, a start-up founded in 2023 by former DeepMind and Meta researchers, embodies the European dream of catching up with American giants. In September 2025, it lifted EUR 1.7 billion in a series C, bringing its valuation to EUR 11.7 billion ($14 billion).
Yet, despite this enthusiasm, Mistral has not yet proved its worth in terms of sustainable margin, of solid recurring income or acquisition of large-scale market.
The association of a promising future with high valuation poses a challenge: the start-up will have to turn hopes into concrete results, or face a downward evaluation.
The risk of collective illusion
We see a well-known mechanics: the collective narrative, that of a technology that will upset the world economy, takes precedence over the concrete figures.
The valuations are blazing, fed by fear of missing a historic opportunity.
In this sense, the traditional accounting, cautious and frozen in the past, seems dephased. But it retains a virtue: reminding that the value of a company is also judged on its ability to generate cash flows, not on the mere promise of a bright future.
When enthusiasm falls back, it is the balance sheets inflated by high-priced acquisitions or by too optimistic projections that are exposed. The excess values must then be depreciated, revealing in the open the excesses of the euphoric period.
Between accounting prudence and speculative fever
The confrontation between the principle of historical cost and the formation of financial bubbles, whether they concern real estate, technology or today artificial intelligence, highlights a simple truth: The accounts never tell the whole story. They offer a solid but incomplete basis.
For an entrepreneur as for a student, the lesson is clear: It is necessary to learn to read beyond the figures, to distinguish the book value from the economic value, and to be wary of market value when dictated by collective enthusiasm..
Artificial intelligence is likely to open up real prospects for productivity and growth. But until economic models are stabilized, caution remains. Bulls never warn when they burst.
History repeats itself, but lessons quickly forget
From Dutch tulips in the 17th century to Internet values in the late 1990s to American real estate before 2008, the big financial bubbles all followed the same scenario: a collective package, a dizzying price rise, and then a brutal deflation that leaves exsangue balance sheets and ruined investors.
The euphoria of yesterday becomes today's disillusion, and the excess values recorded at the top of the cycle turn into massive losses.
In this context,he work of the auditors is particularly sensitive.
How can we attest to financial statements that sometimes include billions of euro of goodwill, while their justification is based on fragile assumptions of synergies and future growth?
Since The systematic depreciation of these excess values has been eliminated in favour of simple depreciation tests, sound balance sheets depend on highly uncertain financial projections.
These « oddly tested » are by nature sensitive to discount rate changes, margin forecasts, macroeconomic scenarios. The slightest inflection in these parameters can turn a promise into a dry loss.
This is the fragility of the present system: On the one hand, prudent accounting principles such as historical cost that underestimate the real value of strategic assets; on the other, speculative bubbles that push to put dizzying amounts on the balance sheet without guaranteeing their sustainability.
Between these two extremes, the role of the statutory auditor has become more complex than ever. Its mission is not only to certify figures, but also to shed light on the risks behind them. And at a time when artificial intelligence is fuelling disproportionate valuations, this vigilance demands more than ever, because history has proved: no bubble lasts forever; trees never climb to the sky.
