Broadcom on Wednesday evening beat quarterly earnings expectations and provided a strong multiyear revenue forecast for its artificial intelligence custom chip business. Revenue in the fiscal third quarter of 2026, which ended Aug. 2, rose 85% year-over-year to $29.59 billion, a slight beat versus the $29.36 billion consensus forecast, according to estimates compiled by LSEG. Adjusted earnings per share (EPS) increased 96% to $3.32, beating expectations of $3.24, LSEG data showed. AVGO YTD mountain Broadcom YTD However, Broadcom shares down modestly in up-and-down extended-hours trading. A silver lining is that the stock was down 7% at one point. Broadcom has become a big disappointment of late, with shares only up about 6% for the year. Bottom line CEO Hock Tan needed to come out swinging on the conference call, and he did just that when he went over what he expected Broadcom’s AI revenues will be over the next few years. How the market rewards his bullish framework remains to be seen. Based on current fiscal fourth-quarter AI revenue guidance of $21.7 billion, Broadcom expects to do $58 billion of AI revenue in fiscal year 2026. That’s up from its prior view of $56 billion. That was a good first step, but the market was much more interested in Tan’s estimates for fiscal year 2027 and beyond. Recall, Broadcom shares sold off last quarter when Tan simply reiterated expectations of at least $100 billion in AI revenue in fiscal year 2027. On Wednesday evening, Tan updated that guide to $115 billion on the earnings call, a figure that is constrained by supply. That number was actually a little disappointing at first blush. Many expected Broadcom to raise its outlook somewhere in the $20 billion to $25 billion range. So, what got the stock to climb after hours? It was when Tan said he expects AI revenue to double again in fiscal year 2028 to $230 billion. That was the blowout the market had patiently waited for. This longer-term outlook was above the FactSet consensus estimate of $177 billion. On the bottom line, Tan said he believes the company is on track to exceed EPS of $30 in fiscal year 2028. That’s firmly above the FactSet consensus estimate of $26.38. Importantly, this outlook not only incorporates the chip supply Broadcom expects to receive, but also physical infrastructure constraints such as land, power, and data center shells. The math behind these figures is based on the company’s relationship with Alphabet’s Google, Anthropic, OpenAI, and Meta Platforms . Broadcom’s relationship with Alphabet has been questioned for the last few months given a recent deal with Marvell and ongoing share loss concerns against MediaTek. Tan defended his relationship with Google on the call. “Our partnership with Google will continue to sustain because we have the strongest IP portfolio in semiconductor design, including industry leading service chip to chip interconnect, leading-edge HBM [high-bandwidth memory] and SRAM [static random-access memory] integration, and simply differentiated advanced packaging,” Tan said on the call. ‘Most of all, we have consistently delivered the fastest time to market for TPUs from product definition to production.” TPUs, or tensor processing units, are the chips that Broadcom co-designs. Beyond Google, Tan said he expects Anthropic will be its largest customer of XPUs, which is what Broadcom calls its specialized chips, over the next two years due to the expected 5 gigawatts of TPUs they expect to deploy, with a line of sight into an incremental 10 GW. Computing power is measured in GW because energy is the limiting factor in AI infrastructure. With OpenAI, Broadcom is on track to deploy 1.3 GW next year with a line of sight to deploying 5 GW of its new Jalapeno processor and successor generation XPUs in the future, making them its second largest XPU customer. The 1.3 GW is consistent with prior commentary. To help support Anthropic and OpenAI fund their deployments, Broadcom established a special purpose vehicle (SPV) platform in partnership with Apollo and Blackstone to facilitate more than 20 GW of compute infrastructure by the end of 2028. The company explained on the earnings call that this SPV helps these frontier AI labs bridge a gap between their current cash flows and upfront investments. Nvidia did a similar financing arrangement with major Wall Street banks. With Meta, the two companies are on track to deliver three generations of MTIAs, or Meta Training and Inference Accelerator, with a line of sight to deploy 3 GW through 2028. That’s no change from prior guidance. We come away mixed from the quarter. On one hand, we are pleased to see Broadcom has enough visibility into both demand and supply to raise some of its outer-year targets. Using management’s new $30 adjusted EPS target, the stock trades at a very cheap multiple of roughly 12 times earnings. On the other hand, the market appears unwilling to ascribe a higher multiple to those earnings. That could be due to concerns around circular financing tied to the SPV, customer concentration among frontier labs, or, more recently, political pushback against data centers that could slow the buildout. The latter concern is why we cut this position in half last week. The smaller-sized position allows us to be more patient with this one. Even though Broadcom’s earnings estimates will go up for fiscal 2028 based on Tan’s guide, we’re taking our price target down from $480 to $430 due to the lower P/E multiples the market is giving the semiconductor group. We’re reiterating our hold-equivalent 2 rating . Segment commentary In Semiconductor Solutions, the much larger of the two operating segments and the one Wall Street is focused on because it houses its AI business, revenue increased 70% year over year to $20.84 billion. That beat analyst estimates of $20.29 billion, according to FactSet. AI semiconductor revenue surged 221% year over year and 51% sequentially to $16.7 billion. That was better than the $16 billion the company guided to one quarter ago. Broadcom’s AI business includes both custom chip revenue and networking products — things like Ethernet switches that move data between Broadcom’s XPUs and help chips work together. In Broadcom’s other operating segment, Infrastructure Software , revenue increased 29% year over year to $8.75 billion, missing the FactSet consensus estimate of $8.82 billion. The only real discussion about this business on the call is how its VMware private cloud AI allows enterprises to run a secure, cost-effective platform to build and run AI alongside existing applications. Guidance For its current fiscal fourth quarter, Broadcom forecasted total revenue of about $34.8 billion, below the $35.03 billion expected, according to estimates compiled by LSEG. However, the shortfall may be from software revenue because the company guided AI semiconductor revenue to accelerate to $21.7 billion. That’s a beat versus the FactSet consensus of $20.93 billion and represents 236% year-over-year growth. The company expects fiscal Q4 adjusted operating margin to be 66%, implying $22.97 billion of operating income. That’s slightly below the consensus estimate of 66.5% margins and $23.05 billion in operating income, according to FactSet. Higher memory content was called out as a reason why margins are not expected to be stronger. (Jim Cramer’s Charitable Trust is long AVGO, GOOGL, META, NVDA. See here for a full list of the stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust’s portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
Broadcom guides to big numbers but investors wanted more. Why we’re OK to hold for now
