Broadcom briefly crossed $2 trillion before an OpenAI report dragged shares sharply lower
Broadcom had its moment in the sun. Then an OpenAI report arrived and took some of the shine off in a hurry.
Earlier this month, shares of Broadcom briefly crossed a $2 trillion market capitalization for the first time, touching a record closing high near $422 on April 22 after Google unveiled the design for its next-generation AI chip. By today, April 28, the stock had given back 4.23% at the open, dragged lower alongside the broader semiconductor sector by a Wall Street Journal report suggesting OpenAI had missed internal targets for both user growth and revenue.
The swing in less than a week captures the central tension facing every AI-adjacent stock right now: extraordinary long-term visibility sitting alongside very real short-term anxiety.
What pushed Broadcom to $2 trillion
The catalyst for the record high was a specific and meaningful piece of news from Broadcom’s most important customer. Google announced its eighth-generation tensor processing unit, and this time it came with a notable structural change: two chips instead of one. TPU 8t is designed as a training powerhouse while TPU 8i is built for inference, the computationally intensive work of AI models answering questions and completing tasks in real time.
The announcement landed with particular force for Broadcom because CEO Hock Tan had predicted almost exactly this development on the company’s most recent earnings call. He told investors that customers would begin developing two chips simultaneously each year, one for training and one for inference. Google’s announcement confirmed that prediction publicly, and the market rewarded Broadcom with a 5% single-day gain.
Why the 2-chip roadmap matters for Broadcom’s revenue
The strategic implications extend well beyond one customer. Google notes that TPU 8i delivers 80% better performance per dollar compared to its seventh-generation chip, which should meaningfully reduce the cost of inference and in turn drive higher demand for the chips themselves. More inference volume means more chips needed, and more chips needed means more revenue for the company designing them alongside Google.
There is also a deeper structural benefit. In a single-chip architecture, companies can optimize performance through software, but the gains available at the software level are far smaller than what can be achieved through hardware-level specialization. By splitting into two purpose-built chips, Google is effectively moving optimization from software to hardware, which is precisely where Broadcom’s expertise sits. That shift could allow the company to capture meaningfully more value per design cycle.
Developing two chips also requires considerably more engineering work than one, and Broadcom should expect compensation for that added complexity. Taken together across multiple hyperscaler customers, the financial impact compounds quickly.
The 5 customers and why confirmation matters
Broadcom has said it is seeing two-chip-per-year roadmaps across five of its hyperscaler customers. So far, two have publicly confirmed the framework. Meta announced its own roadmap in mid-March under the headline of four chips in two years, a cadence that aligns precisely with Tan’s prediction. With Google now a second public confirmation, the level of confidence around the remaining three customers increases substantially. If all five follow through, the benefits of Google’s two-chip approach would extend across Broadcom’s entire custom chip business.
Why the stock fell 4% anyway
Today’s decline was not a response to anything specific at Broadcom. The company reported first-quarter revenue of $19.3 billion in its most recent results, a 29.5% increase from the same period a year earlier, and provided optimistic second-quarter guidance. All 11 analysts covering the stock carry buy-equivalent ratings, and the median price target among 22 analysts sits at $475.
The pullback came from a macro-level concern about AI demand broadly. The OpenAI growth report rattled the entire sector, sending Nvidia down 3.6%, AMD down 5.4% and Micron down 4.9% alongside Broadcom’s own decline. Analysts described the move as a technical cooldown and profit-taking following an extended period of sharp appreciation, rather than any deterioration in Broadcom’s own fundamentals.
3 risks that deserve attention
Despite the broadly constructive outlook, there are meaningful risks investors are tracking.
First, Broadcom carries $66 billion in debt, a legacy of its VMware acquisition, which limits financial flexibility and could create pressure if market conditions deteriorate. Second, the company faces ongoing regulatory scrutiny in the European Union related to its VMware licensing practices, including price increases and restrictive terms that have drawn complaints from enterprise customers.
Third, while hyperscaler AI spending remains robust for now, any meaningful slowdown in that investment cycle would land directly on Broadcom’s revenue outlook given how concentrated its AI chip business has become.
Earnings from several major technology companies are due shortly and will offer the market its clearest read yet on whether large-scale AI infrastructure spending is holding or beginning to soften. For Broadcom, that data could either confirm the current trajectory or reignite the concern that briefly surfaced today.
Disclaimer: This article is for informational purposes only and does not constitute financial advice. The author and publication are not registered investment advisors and do not provide personalized investment recommendations.