Solid results, strong reactions, expectations that tolerate no deviation. Two key AI stocks are beginning to show subtle fractures that the market no longer seems willing to overlook.
What happens when companies long considered untouchable deliver solid results, fully consistent with the dominant narrative, and are nevertheless ruthlessly punished by the market? That is the moment when the market starts speaking a different language — quieter, less euphoric, more analytical. And it is precisely in these phases that the first cracks in major bubbles tend to emerge. Not when earnings collapse, but when they fall short of perfection. Today, two emblematic AI infrastructure stocks are sending a signal worth paying attention to. A signal that doesn’t shout, but whispers something potentially far more dangerous.
This is no longer a normal market
In recent quarters, equity markets have experienced a phase of almost linear narrative expansion. Artificial intelligence has become the dominant driver of expected growth, capital expenditure, and asset allocation. In this environment, valuations detach from current fundamentals and begin to discount a frictionless future. This is where the concept of multiple compression becomes critical.
When valuation multiples expand not because earnings have surged, but because expectations have been extrapolated forward, even marginal signs of deceleration take on outsized importance. The market stops focusing on absolute results and instead concentrates on the delta between what was implicitly priced in and what is actually delivered.
Oracle: when the pivot begins to creak
Oracle is the first to blink. In pre-market trading, the stock drops 11%. The stated reason is almost trivial: revenues of €16.06 billion versus expectations of €16.21 billion. A marginal miss, statistically insignificant under normal macro conditions. But the current environment is anything but normal.
Oracle is no longer viewed merely as a legacy software company. It is perceived as a core piece of AI infrastructure — spanning databases, cloud services, data centers, and workload orchestration. A significant portion of the enterprise AI growth narrative is built on this foundation. When even a micro-fracture appears in such a linchpin, the credibility of the entire mechanism comes under scrutiny.
The reaction does not stop with Oracle. Contagion is immediate. Nvidia, AMD, CoreWeave, and even SoftBank — one of the most concentrated AI exposure vehicles in Asia — come under pressure. This is not indiscriminate panic. It is a rapid reassessment of risk. The market is delivering a clear message: it is no longer rewarding the narrative alone. It is beginning to stress-test its durability.
Broadcom: strong numbers, muted response
Broadcom reinforces the signal. Following Oracle’s results, attention shifts to another cornerstone of AI infrastructure. On a standalone basis, the figures are outstanding. Revenues of €18.02 billion versus €17.5 billion expected. Adjusted EPS of €1.95 versus €1.87. Revenue guidance for the current quarter of €19.1 billion compared with estimates of €18.4 billion. By any conventional metric, these are strong results.
And yet, the market reaction is subdued — almost skeptical. The issue is not performance, but expectations. Broadcom is currently valued for smooth, uninterrupted, and perfectly linear growth. In effect, the stock is priced as if AI infrastructure adoption will proceed without delays, bottlenecks, or cyclical pauses. That is an inherently fragile setup, because it leaves virtually no margin for execution risk.
The real issue: asymmetric expectations
This is the crux of the matter. The concern is not an earnings downturn, but the asymmetry between valuation levels and tolerance for error. When a stock trades at elevated multiples, even minor deviations can trigger disproportionate repricing. At present, the technology sector — and AI-related names in particular — is trading at forward P/E ratios well above the 80th percentile of their 10- and 30-year historical ranges. In practical terms, the market is discounting an almost idealized growth path.
The problem is that real-world growth is never linear. It consists of accelerations, slowdowns, capacity constraints, and capital investments that precede revenue realization. Oracle’s decision to revise capital expenditure guidance upward — to roughly €50 billion per year from the previously expected €35 billion — is emblematic. On one hand, it underscores commitment to AI. On the other, it raises the risk of a timing mismatch between investment intensity and monetization.
The repricing mechanism
Markets move through causal chains. If earnings momentum decelerates, valuation multiples come under pressure. If multiples prove unsustainable, prices must adjust to restore equilibrium. As prices decline, management guidance becomes more cautious. And if losses breach certain technical and psychological thresholds — typically around 20% — the regime shifts. At that point, it is no longer a correction. It becomes a sector-specific bear market.
The critical takeaway is that this process may now be triggered not by macro shocks, but by a series of modest disappointments. Oracle and Broadcom are not underperforming. They are simply demonstrating that perfection is unattainable — and that is precisely what the market appears increasingly unwilling to accept.
Be cautious, not fearful
This is not a call for panic, nor a prediction of an imminent collapse. It is a call for clarity. Large bubbles rarely burst overnight. They begin with subtle signals, often dismissed because they are inconvenient. Artificial intelligence remains a powerful structural transformation. But in markets, the price paid to participate matters just as much as the story being told.
Original article published on Money.it Italy 2025-12-16 10:42:22. Original title: Queste 2 azioni segnalano che la bolla AI sta per scoppiare
Argomenti
