Quick Read
-
Broadcom’s AI chip revenue jumped 221% to $16.7 billion, with management forecasting $230 billion in AI sales by fiscal 2028.
-
NVIDIA leads general-purpose AI compute with a 45‑times P/E, whereas Broadcom’s custom XPUs cater to hyperscalers at a forward P/E close to 20.
-
Because XPUs are co‑designed with each hyperscaler, switching to another supplier would require restarting a multi‑year development cycle, creating high switching costs.
Broadcom (NASDAQ:AVGO) delivered a quarter that ranks among the most significant AI results of the year. AI semiconductor revenue hit $16.7 billion—up 221% year‑over‑year and 54% sequentially—and management lifted its fiscal‑2027 AI revenue forecast to about $115 billion, with fiscal‑2028 seen reaching $230 billion.
Despite the strong numbers, the stock reacted negatively because the Q4 total‑revenue guidance of roughly $34.8 billion fell just short of, or barely met, consensus expectations. The regular session ended down 0.66% at $367.24, with a more pronounced move in after‑hours trading.
Broadcom’s Actual Results
Revenue totaled $29.591 billion, up 85.5% year‑over‑year, while non‑GAAP EPS came in at $3.32 versus the $3.2382 estimate. Operating margin widened by 240 basis points to 67.9%, and free cash flow reached $13.665 billion, representing 46% of revenue.
The market focused on the disappointing Q4 top‑line guide, an understandable overreaction given that the more important disclosure was the forward AI outlook.
A custom AI accelerator (XPU) is silicon co‑designed with a particular hyperscaler for a specific model family. XPUs accounted for 73% of AI revenue, and XPU shipments rose more than 3.5‑fold year‑over‑year. Hock Tan summed it up: ‘When you co‑develop a chip optimized for your specific LLM workloads, you will outperform any GPU.’
Free Report, Just Released
Did Any of Your Stocks Make the Top 10 List?
It is an uncomfortable question, and there is now an answer to it.
24/7 Wall St has helped investors make money for over two decades, and our top analysts just finished ranking the definitive Top 10 Stocks To Buy Now. Not the ten biggest companies. Not the ten everyone is arguing about. The ten best stocks to buy right now.
Open your account and look at what you own. Some of it you bought for a reason you could still defend today. Some of it you bought years ago for a reason you can no longer remember.
Story Continues
The report is free. Put the ten next to what you own and find out which is which.
Enter Your Email and See the Ten →
Free from 24/7 Wall St. It lands in your inbox.
NVIDIA (NASDAQ:NVDA) sells a general-purpose platform with the CUDA ecosystem attached. Broadcom plays a different role, supplying design expertise and networking for hyperscalers who already know what they want to run.
Gigawatt‑Scale Visibility and Customer Concentration
Management outlined the deployments. Anthropic plans to field one gigawatt of Ironwood in 2026, another five gigawatts of TPU v8i in 2027, and an additional 10 gigawatts in 2028. OpenAI has Jalapeno on track for 1.3 gigawatts in 2027 and more than five gigawatts across generations in 2028. Gigawatts represent ordered capacity that still hinges on power, land, and packaging before turning into revenue. The concentration is notable: six XPU customers, four of which are expected to be especially large, meaning any shift in their plans could have a material impact.
The underappreciated piece is networking. Tomahawk 6, a first‑to‑market 100‑terabit switch, is deployed by essentially every AI hyperscaler building XPUs with Broadcom, while Tomahawk 7 is taped out as the industry’s inaugural 200‑terabit‑per‑second Ethernet switch. The shift toward custom silicon dilutes blended gross margin because memory content is heavier, which is why Q4 gross margin is guided to roughly 73%. This is a feature, not a bug: a hyperscaler that co‑designs a chip cannot move that program elsewhere without restarting a multi‑year design cycle.
Is Broadcom Stock a Buy?
NVIDIA continues to dominate general‑purpose accelerated computing and its developer ecosystem, trading at a 45‑times P/E. Marvell (NASDAQ:MRVL) offers the only other merchant custom‑silicon alternative, albeit on a smaller scale. Broadcom’s forward P/E near 20×, paired with a $525.97 analyst price target and Hock Tan’s goal of surpassing $30 in EPS by fiscal 2028, makes the risk‑reward profile attractive even with the customer‑concentration overhang.
The idea of Broadcom as an NVIDIA alternative is the central theme. We have identified the early characteristics shared by the market’s biggest tech winners in a free playbook titled *The Next Nvidia Playbook*. While the setup looks attractive on a research basis, the customer‑concentration discount warrants a measured position size.
Contact editorial@247wallst.com for any questions or corrections.
