Broadcom Q3 earnings preview: AI demand faces a tougher test 

Broadcom Q3 earnings preview: AI demand faces a tougher test

Broadcom reports its fiscal third-quarter results on Wednesday, 2 September, with investors looking for evidence that the extraordinary growth in custom AI chips and networking remains on track despite rising concerns over customer concentration and intensifying competition.

The semiconductor and infrastructure software group enters the results with expectations already elevated. Broadcom generated record revenue of $22.2 billion in fiscal Q2, up 48% year-on-year, while AI semiconductor revenue surged 143% to $10.8 billion. Management subsequently guided for Q3 revenue of around $29.4 billion, implying growth of 84% year-on-year, with AI semiconductor revenue expected to reach $16 billion, more than double the prior-year figure.

That guidance will provide the immediate benchmark for the September results. But, as with many of the biggest AI stocks, investors are likely to focus less on whether Broadcom simply meets its forecast and more on whether management can raise the bar again.

Those looking to trade around the results can do so through spread betting or CFD trading, while longer-term investors can access Broadcom shares through IG Invest or our share dealing service.

AI is driving the earnings story

​Broadcom's biggest opportunity is its position in the rapidly expanding market for custom AI accelerators and networking chips.

​Unlike Nvidia, which predominantly sells general-purpose GPUs, Broadcom works with hyperscalers to develop customised silicon designed around their specific AI workloads. Its customers include Google, Meta, Anthropic and OpenAI, with management identifying six core custom-chip customers.

​The strategy is becoming increasingly important as the world's biggest technology companies seek to build their own AI infrastructure and reduce their dependence on off-the-shelf accelerators.

​Broadcom expects full-year 2026 AI semiconductor revenue of approximately $56 billion, around 180% above fiscal 2025, and has reiterated its expectation that AI semiconductor revenue will exceed $100 billion in fiscal 2027.

​That makes the September earnings call an important test of whether the company's extraordinary growth trajectory remains intact.

The $16 billion question

​The immediate number to watch is the $16 billion Q3 AI semiconductor revenue forecast.

​Broadcom's previous quarter demonstrated just how demanding the market has become. Q2 revenue came in slightly below the Wall Street estimate despite the huge increase in AI sales, while the company's $16 billion Q3 AI forecast was also fractionally below analysts' expectations. The shares subsequently fell sharply as investors questioned whether Broadcom could continue to exceed an increasingly demanding set of expectations.

​That makes another strong beat-and-raise particularly important.

​Investors will want to hear that AI demand is broadening rather than relying on a handful of hyperscalers, that customer deployments are progressing on schedule and that supply is sufficient to meet demand.

​Broadcom has previously said it expects its AI semiconductor momentum to continue through 2027, with its six core customers providing substantial visibility into future demand. Management has also indicated that its AI backlog through the second half of 2026 and fiscal 2027 is substantial.

​For those wanting to buy shares in Broadcom or trade around the results, our platform gives you access to US-listed stocks alongside thousands of other global markets.

Google becomes the elephant in the room

​The biggest new issue going into the results is Google.

​Broadcom has been Google's main custom AI chip partner, particularly around its Tensor Processing Units. But Google's recent agreement with Marvell to help develop custom AI chips has raised questions about whether Broadcom's share of Google's future spending could decline.

​The deal gives Google the right to acquire up to $12.2 billion of Marvell shares, with the partnership covering AI accelerators, storage and networking components. Broadcom shares fell more than 5% when the agreement was announced.

​The key issue is whether this represents genuine diversification by Google or the beginning of a more meaningful shift in Broadcom's share of the custom-chip market.

​Management will therefore face pressure to explain the strategic importance of its Google relationship and whether its other customers can compensate for any eventual reduction in Google's contribution.

​The risk should not be overstated, however. Broadcom's AI opportunity extends well beyond Google, with Meta, Anthropic and OpenAI among its other major customers. OpenAI and Broadcom have also announced a multi-year collaboration involving 10 gigawatts of OpenAI-designed AI accelerators, with deployments scheduled to begin in the second half of 2026.

Inference could extend the AI cycle

​Another important theme is the shift from AI training towards inference — the stage where trained models actually respond to users and applications.

​Broadcom has argued that inference workloads will require significant additional compute capacity as AI becomes more widely embedded in products and services. OpenAI's partnership with Broadcom is explicitly focused on developing an inference-optimised accelerator, highlighting the potential for custom silicon to play a growing role as AI usage expands.

​This could be particularly important for Broadcom because inference workloads can be optimised for specific applications, potentially making customised accelerators more attractive to hyperscalers.

​If management provides evidence that inference demand is accelerating alongside training demand, investors may become more confident that the current AI infrastructure boom is not simply a temporary spending cycle.

Debt and AI infrastructure spending

​Broadcom's expansion is also taking place against a backdrop of enormous investment requirements across the AI industry.

​Reuters reported last week that Broadcom is exploring raising more than $60 billion of debt, potentially reaching as much as $100 billion through a financing structure intended to support AI-focused customers including Anthropic and OpenAI. That follows a separate $35 billion financing arrangement with Blackstone and Apollo linked to Anthropic's computing capacity.

​Such developments illustrate both sides of the AI investment story.

​On one hand, the willingness of investors to finance huge amounts of AI infrastructure suggests that demand remains exceptionally strong. On the other, the sheer scale of the capital being deployed raises questions about how quickly those investments will generate returns.

​Broadcom therefore needs to demonstrate that its customers' spending plans translate into durable chip demand rather than a short-lived infrastructure build-out.

VMware provides another earnings test

​AI may dominate the headlines, but Broadcom's infrastructure software business remains an important part of the investment case following its acquisition of VMware.

​Infrastructure software revenue was $7.18 billion in Q2, up 9% year-on-year, considerably slower than the semiconductor business. As AI chips become a much larger proportion of Broadcom's revenue mix, investors will be watching whether the software division can continue providing stable, recurring cash flow and support the group's margins.

​The changing revenue mix is also important because semiconductor growth can carry different margin characteristics from software. Investors will therefore be looking closely at gross margins, operating margins and free cash flow alongside the headline revenue number.

​For those monitoring the broader stock market impact of AI earnings seasons, our trading platform gives you the tools to track and act on fast-moving developments across technology stocks.

What investors will listen for

​Broadcom's September earnings call is likely to revolve around five questions:

​First, can the company beat the $29.4 billion Q3 revenue guide? A strong result will probably need to come with an equally strong outlook to satisfy investors.

​Second, can AI semiconductor revenue exceed the $16 billion target? This is arguably the most important number in the release.

​Third, is the $100 billion-plus fiscal 2027 AI revenue target still firmly on track? Any change in tone could have a significant impact on sentiment.

​Fourth, how much risk does Google's new relationship with Marvell create? Investors will want reassurance that Broadcom's six-customer pipeline remains intact.

​Finally, is inference becoming a meaningful new driver of demand? Evidence that AI workloads are shifting from training towards large-scale inference could strengthen the argument that AI infrastructure spending has several years left to run.

Broadcom technical analysis

​The Broadcom share price – year-to-date up 6.5% - is hovering above its current August low at $357.61, itself made between the January highs and July lows at $357.80-to-$356.43 which acted as support.

​Were it to give way on a daily chart closing basis and the next lower February-to-March highs at $353.14-to-$352.34 were to be slipped through as well, the September-to-December 2025 lows at $329.06-to-$324.05 may be reached.

​Broadcom weekly candlestick chart

​Only a rise and daily chart close above the 200-day simple moving average (SMA) at $369.25 and 21 August high at $375.13 may lead to another bounce off the July-to-August support zone taking place. If so, the late June-to-July highs at $407.53-to-$414.64 may be revisited, ahead of the current August high at $432.73. This level and the 14 May high at $442.36 need to be exceeded for the June peak at $495.00 to be back in the picture.

​Broadcom daily candlestick chart

How to invest in Broadcom shares

1. ​Do your research on Broadcom, the Q3 results and the key AI metrics to watch on 2 September
2. ​Download IG Invest or open a share dealing account with us
3. ​Search for Broadcom (AVGO) in our platform or app
4. ​Choose the number of shares or value of money you'd like to invest
5. ​Place your trade