Broadcom (AVGO) has no shortage of artificial intelligence demand. The more difficult issue is whether its clients can afford to pay for everything.
Tigress Financial initiated Broadcom coverage on Oct. 22 with a Buy rating and a $600 price target, citing a multiyear opportunity in custom AI accelerators, high-performance networking, and infrastructure software.
The analyst believes that Broadcom’s AI infrastructure leadership and growing custom silicon business will drive the next phase of development, according to Investing.com, citing Tigress.
The target stands out after Broadcom closed Oct. 22 at $323.24, roughly 32% below its $495 52-week high. A move to $600 would represent nearly 78% upside from that closing price.
However, Hock Tan, CEO of Broadcom, has discovered an issue that may decide the success of that optimistic prediction.
Tan said this month that finance might become the largest barrier to continued AI progress, after years of struggles by chipmakers with shortages of wafers, memory, and other components.
Finding consumers who desire Broadcom’s chips may thus be less of a barrier than securing sufficient funding to construct the massive data centers needed for those chips.
Broadcom’s AI business is still exploding
There are several statistics that back up Tigress’s optimism.
At $29.6 billion, Broadcom’s fiscal third-quarter sales increased 86% year over year.
The AI industry expanded far more quickly.
AI semiconductor revenue reached $16.7 billion, up 221% from a year earlier and 54% from the previous quarter. Broadcom expects AI semiconductor revenue to reach $21.7 billion in the fourth quarter, up 236% year-over-year, due to strong demand for custom AI accelerators and networking.
That amounts to roughly $5 billion of additional quarterly AI semiconductor revenue in only three months.
Also read: Broadcom fought AT&T over VMware: Now a telecom giant is all in
On Broadcom’s third-quarter earnings call, management laid out a path toward about $115 billion of AI semiconductor revenue in fiscal 2027 and $230 billion in fiscal 2028.
These projections reflect massive expected deployments from clients like Google, Anthropic, OpenAI, and Meta. This scale contributes to Tigress’s belief that Broadcom can surpass its prior peak in the stock market.
It also presents an issue.
Customers still need data centers, energy, and tens of billions of dollars to operate the more advanced bespoke accelerators and networking equipment that Broadcom can provide.
Broadcom CEO identifies AI’s next bottleneck
Tan said that Broadcom is overcoming memory and wafer shortages at a conference in Singapore in October.
Capital is a different story.
Tan told Reuters that capital would be the biggest AI development constraint. He compared some AI investment to “casino gambling,” warning that investors can set unrealistic expectations without understanding what they are funding.
The warning is important since developing AI infrastructure is now very expensive.
To deal with that finance issue, Broadcom has already gone beyond just selling semiconductors. The business established an AI infrastructure finance platform in June with Apollo and Blackstone with the goal of enabling more than 20 gigawatts of processing capacity until 2028.
Over one gigawatt of Anthropic computing capacity was financed by the platform’s first deployment, which came with an initial $35 billion funding commitment.
Broadcom says the infrastructure will support frontier AI labs like Anthropic and OpenAI using its custom AI processors and networking products. The arrangement demonstrates the growing interdependence between funding and chip demand.
When AI businesses develop more advanced processing systems, Broadcom benefits. However, several of those businesses don’t now make enough money to fully fund massive infrastructure initiatives.
That relationship became even clearer with Anthropic.
Reuters reports that Broadcom consented to lend Anthropic up to $42 billion to fund infrastructure projects. Broadcom is anticipated to become one of Anthropic’s biggest infrastructure partners as it commits to $125.2 billion of TPU processing capacity over a five-year period.
For Broadcom, assisting clients in obtaining funding might lead to massive future chip sales.
However, it also modifies the danger.
In addition to the question of whether consumers desire its technology, Broadcom is also exposed to the question of whether the economics of large AI data centers eventually justify their expense.
Broadcom’s $600 target depends on more than chip demand
Wall Street is still immensely optimistic. There are now 47 buy ratings, three hold ratings, and 0 sell ratings for Broadcom. The average analyst price target is about $527.59, according to data from Investing.com.
Tigress’s $600 target sits above that consensus, but it isn’t completely isolated. KeyBanc has a $575 target, Bernstein is at $575 and Mizuho is at $530.
The bull instance is simple to comprehend.
Broadcom is not holding out for future AI income. Last quarter, it made $16.7 billion from AI semiconductors. This quarter, it anticipates making $21.7 billion.
Compared to Nvidia’s general-purpose graphics processors, the company’s custom-chip business offers an alternative path into the AI boom.
Large clients collaborate with Broadcom to create accelerators tailored to their unique workloads. Tan argues that this approach could enhance the economics for clients engaged in large-scale AI computations.
More Tech:
- Meta and Microsoft just sent employees a memo about Anthropic
- Microsoft cutting off popular software with no extension Oct. 13
- Apple’s biggest AI advantage may face new privacy challenge
In addition to Broadcom’s AI revenue, investors should keep an eye on the speed at which its clients can convert projected processing into operational data centers, even if only some of them do so.
During Broadcom’s earnings call, Tan emphasized that distinction.
Customer deployments of 30 gigawatts may be needed in 2027 and 2028, but not all may be operational on time. Broadcom cannot realize all the revenue from that demand until it builds data-center shells, electricity, and equipment.
Financing becomes more than a side concern as a result. It establishes the rate at which theoretical chip demand translates into real chip sales.
Therefore, Tigress’s $600 target depends on more than just Broadcom’s ability to produce AI chips that are competitive.
Customers must build the infrastructure. Financing must remain accessible. AI models must provide enough revenue to justify ongoing expenditures.
Recent findings from Broadcom indicate that demand is not the main issue.
Between the third and fourth quarters alone, the firm anticipates adding almost $5 billion in quarterly AI semiconductor sales.
The more difficult issue is whether the financial sector can continue to finance infrastructure at the rate assumed by Broadcom’s remarkable projections.
Tigress’s $600 target is ultimately a bet that enough capital will remain available.
Related: Jim Cramer sees something Wall Street may be missing in Broadcom