Broadcom CEO Hock E. Tan (left), Oracle CEO Larry Ellison
Analysis

Oracle and Broadcom rattle markets as AI trade hits a reality check

Strong earnings failed to calm investors as rising capital costs and stretched valuations prompted a pullback in AI-linked stocks.

Fear of heights may be the simplest way to describe what has gripped Wall Street in recent days around AI-related stocks. Oracle and Broadcom, two of the most prominent names in what is now widely referred to as the “AI trade,” a broad bet on the artificial intelligence revolution, sent tremors through the market when Oracle fell 15% over the course of the week and Broadcom plunged 11.5% on Friday alone. The sell-off came on a day when risk aversion had already taken hold, with the Nasdaq down 1.7% and the S&P 500 falling 1.1%.
Yet viewed from a longer perspective, 2025, now just two weeks from its end, still looks like another strong year on Wall Street, one likely to leave investors smiling and traders and analysts collecting generous bonuses.
1 View gallery
מימין לארי אליסון ו הוק טאן
מימין לארי אליסון ו הוק טאן
Broadcom CEO Hock E. Tan (left), Oracle CEO Larry Ellison
(Aaron Schwartz/Sipa / Bloomberg)
What, then, should be made of the renewed anxiety that has resurfaced after several weeks in which talk of an AI bubble seemed to fade? It is not clear that the fear reflects any deep or structural change. Markets that occasionally pause or pull back after sharp gains often signal a measure of sanity rather than excess. When every piece of news, good or bad, is interpreted as bullish and pessimism disappears entirely, that is typically when bubbles burst.
For now, the market continues to stumble intermittently. Friday’s decline of more than 1% in the S&P 500 marked the index’s 28th such drop this year. Historically, that places 2025 squarely within the norm: since 1928, markets have averaged 29 declines of more than 1% per year. Even after the recent sell-off, the bears have yet to celebrate. Short interest in leading AI stocks has not risen meaningfully, and the longer-term performance of the sector remains striking. Oracle’s shares are still roughly in line with the S&P 500 for the year, while Broadcom remains up about 55% year to date, after doubling in both 2023 and 2024. Even following Friday’s drop, Broadcom continues to trade at a rich earnings multiple, particularly compared with Nvidia or Alphabet, the two companies currently dominating the AI narrative.
Oracle erases September’s surge
The quarterly results and forecasts released by Oracle and Broadcom late in the week were underwhelming, but far from disastrous. Just a quarter or two ago, numbers like these might have propelled both stocks higher, possibly by double digits. That is especially true for Broadcom, which comfortably beat expectations. Revenue reached $18 billion, above analysts’ forecasts of $17.5 billion, while operating profit came in at $11.9 billion, compared with expectations of $11.4 billion.
Broadcom, which designs custom AI chips for companies including Google, also raised its revenue forecast for the coming quarter from $18.3 billion to $19.1 billion and said AI-focused sales are expected to double to $8.2 billion. The company disclosed that its much-discussed $10 billion contract with an unnamed customer is with Anthropic, while also confirming ongoing work with Google and Meta on chips aimed at loosening Nvidia’s grip on the market. What unsettled investors was not demand, but margin pressure, as profitability declined alongside the rapid growth in AI chip sales.
Oracle’s picture is less dramatic, though that is partly because expectations had been pushed to extremes. In September, the company’s previous quarterly results sent its stock soaring 30% in a single day. Alongside that rally, Oracle, one of the pillars of the AI ecosystem through its database and cloud infrastructure business, laid out an ambitious four-year growth forecast. Since then, the stock has given back all of those gains, as investors have begun to focus on the cost of delivering on those promises.
To meet its targets, Oracle must invest heavily and quickly. It has already raised its capital expenditure forecast by $15 billion, to $50 billion in 2026. Combined with a slight revenue miss in the most recent quarter, where sales reached $16 billion, and a swing to negative operating cash flow of $10 billion, the investment burden has weighed heavily on the stock. Adding to the pressure is the $18 billion in debt Oracle has taken on to finance the rapid construction of data centers for AI customers, led by OpenAI, with which it has signed a massive $300 billion agreement. By the end of the week, Oracle’s bonds were trading at yields approaching 5.9%, a level often associated with distress, even as the company denied reports of delays in some data-center projects.
A reality check, and a reminder about diversification
The surge in AI stocks over the past two years, which also lifted the broader indices, was fueled largely by vision and narrative. Companies spoke of a technological revolution that would reshape industries and society, and investors priced in an extraordinary future. Three years after the launch of ChatGPT ignited that excitement, 2026 is shaping up as the year when the promise begins to translate into reality.
Reality, however, is rarely as generous as the dream. In the case of AI, the central asterisk is the immense investment required to build the underlying infrastructure. Today’s systems can support limited, consumption-based uses of AI, but not yet the large-scale organizational deployments that are expected to underpin sustainable business models. This challenge has drawn less attention, particularly from retail investors who chased Nvidia’s soaring stock without fully considering who would ultimately pay for the vast number of GPUs it sells.
That burden falls on companies such as Oracle, which has become a convenient symbol of the excesses of the AI trade: a company buying massive quantities of Nvidia chips to build data centers for customers like OpenAI, without certainty that those customers can ultimately generate enough revenue to justify the cost.
The AI market remains highly concentrated, with a handful of stocks absorbing most investor capital, both on the way up and on the way down. As large private players such as Anthropic, which is reportedly planning an IPO at a valuation of $300 billion, enter the public markets, investors may gain more opportunities to diversify their exposure. Clearer business models could also bring greater stability to a sector that, for now, continues to oscillate between euphoria and anxiety.