Broadcom AI outlook tests chip rally’s perfection trade

Broadcom shares fell more than 13 per cent in late trading Wednesday after a quarter that put the artificial-intelligence rally’s demands in sharper relief. According to Reuters, the chipmaker posted $22.19bn in second-quarter revenue, while its latest AI-chip outlook failed to clear the higher bar investors had set for one of the market’s most crowded semiconductor winners.

Awkwardly for Broadcom, the company did not tell investors the AI story had broken. Management said the story was still running, only not faster than the market had already priced. After Bloomberg reported the shares had added more than $280bn in market value over four sessions before the print, that distinction mattered. Good quarters can be defended. Merely good quarters are harder to defend after that kind of move.

Viewed that way, the report matters less as a conventional beat-or-miss exercise than as a live test of the AI trade’s valuation discipline. For public-market investors, Broadcom remains one of the clearer ways to own custom chips, networking silicon and hyperscale infrastructure demand outside Nvidia. Wednesday’s selloff still narrowed the question for the next leg of the trade: how much upside is left once expectations already assume a near-flawless AI buildout?

The bar moved first

One number carried more weight than the headline revenue figure: Broadcom’s current-quarter AI-chip forecast of $16bn, a level that would still be large in almost any other context. Wall Street wanted management to lift the slope of the curve. Instead, the company confirmed a curve the market had already drawn for it.

Ryan Lee, senior vice president of product and strategy at Direxion, put the pricing problem plainly in comments carried by Reuters:

“Today’s miss on revenue and subsequent post-market pull back (in shares) shows the market demands perfection for this chip rally to keep running.”

His point explains the violence of the move. Semiconductor companies have not been rewarded simply for participating in AI demand. Shares have been rewarded for proving, again and again, that demand was underestimated. Nvidia trained investors to look for acceleration, backlog expansion and fresh evidence that hyperscalers were not just buying chips, but scrambling for them. Broadcom is now being measured against a similar template, even though its business mix and revenue recognition are different.

Fast-rising stocks create their own expectations trap. The quicker the rally into earnings, the less patience investors have for management restraint. Broadcom’s unchanged long-range AI target of more than $100bn may still point to a durable opportunity. For a share price already anticipating a step-up, leaving the target alone became its own signal.

Demand is not the problem

Management can point to evidence that the underlying AI business remains strong. MarketWatch reported that April-quarter AI-chip revenue reached $10.8bn, up 143 per cent from a year earlier, even as the stock sold off after hours. Reuters also reported that Broadcom kept its long-range AI chip sales target above $100bn and said executives were “very comfortable” securing supply for 2026 and 2027.

“Nothing slows down what was estimated prior - they just didn’t raise it,”

Ben Bajarin, Creative Strategies analyst, told Reuters.

Context matters here because Broadcom’s AI revenue is not a one-quarter consumer-electronics cycle. The company is selling into multi-year cloud capex plans, custom accelerator projects and the networking layer that allows large clusters to function. Its buyer base is concentrated and powerful, which creates risk. Orders also tend to be tied to deep engineering commitments rather than impulse demand.

Revenue recognition is the short-term problem. Investors can believe Broadcom’s backlog and still question how much of it lands in each quarter. They can accept that supply is available and still punish the company if the timing does not beat the Street’s path. In a calmer tape, the $16bn forecast would probably have read as confirmation. In the post-Nvidia AI market, confirmation can trade like disappointment.

Networking changes the lens

A stronger case for Broadcom may sit less in the headline AI accelerator number than in infrastructure plumbing. MarketWatch argued this week that Alphabet’s heavy AI spending gives Broadcom a fresh shine because the company sits inside the custom silicon and networking stack that hyperscalers need as model training and inference expand. Owning that layer is a different wager from simply buying the next GPU cycle.

In Broadcom’s pitch, custom chips, Ethernet switching and data-center connectivity become more valuable as AI clusters spread beyond a handful of training campuses. Bloomberg has tied the company’s AI opportunity to work with customers including Google, Anthropic and Meta, while MarketWatch’s analysis stressed the high-barrier nature of its networking silicon. If that view is right, the selloff is not a vote against AI demand. It is a vote against the speed at which that demand becomes reported revenue.

Competition still sits inside the story. Reuters noted that Broadcom races with Nvidia in parts of the AI infrastructure market and faces rivals including Marvell in custom chips. Scarcity is the point because the market is paying semiconductor leaders for capacity, customer locks and difficult-to-replace designs. If customers can spread work across suppliers, it becomes harder for any one vendor, Broadcom included, to sustain a premium multiple on backlog alone.

So the investor question is not whether Broadcom can benefit from AI infrastructure. It plainly can. The harder question is whether it can benefit at a pace that keeps justifying a stock move already discounting the answer.

The software drag matters

Another detail kept Broadcom from reading as a pure AI-chip proxy. CNBC framed the drop around weak software sales and an unchanged annual AI forecast, even as the semiconductor business carried the narrative. That mix matters. Broadcom’s acquisition-led software business is part of the cash-flow story investors use to underwrite the stock, not a footnote below the chip segment.

When software growth lags while AI guidance merely confirms previous expectations, the blended company looks less like a clean accelerator on hyperscale spending and more like a diversified technology supplier with one very hot division. That is not a bad business. It is just a different one from the market’s favorite shorthand.

Valuation follows from that distinction. A diversified supplier deserves credit for recurring cash flows, customer stickiness and lower cyclicality. A scarce-asset AI story gets paid for urgency. Broadcom has been trying to occupy both categories. Wednesday’s reaction suggests investors will grant that status only when both sides of the business cooperate.

The wider AI trade is on trial

Timing made the selloff more uncomfortable for the broader AI complex. MarketWatch argued this week that the huge run-up in AI stocks may rest partly on a “token mirage”, a warning that investors may be extrapolating usage and spending patterns faster than the economics can settle. Broadcom’s print gave that debate a concrete stock-market test: when the numbers are strong but not stronger than hoped, does the multiple hold?

For now, the answer is no. A one-night drop does not mean the AI capex cycle is rolling over. It means the equity market has become more selective about where the next positive surprise comes from. That distinction matters for Broadcom, Nvidia, Marvell and the other semiconductor names that have become shorthand for the infrastructure boom.

Credit markets are starting to ask a related question. Semafor wrote this week that Anthropic needed a co-signer as lenders weighed debt tied to AI hardware purchases, with Broadcom standing behind part of the financing. The structure shows how AI demand is moving beyond chip orders into credit, leasing and balance-sheet engineering. It also explains why investors are trying to separate durable infrastructure demand from aggressive extrapolation.

Broadcom sits in the middle of that separation. The company has real AI exposure, real customer relationships and a long-dated revenue target that management did not walk back. Its share price had also started to treat those strengths as settled facts. The earnings reaction was the market’s reminder that, in the AI trade, strong demand is now the baseline. Upside is the product.