OpenAI is hungry for as much compute power as it can get its hands on, and the company has signed another deal with a chipmaker to help make that happen. This time around, it’s teaming up with Broadcom to make custom chips and systems for use in both OpenAI’s infrastructure and its partners’ data centers.
OpenAI is designing the “AI accelerators” and systems. Broadcom will start deploying those racks in the second half of next year, the companies said. The aim is to complete the rollout by the end of 2029. The two companies are said to have started working together 18 months ago.
The deal is for 10 gigawatts of chips and it’s worth “multiple billions of dollars,” according to The Wall Street Journal. It was reported last month that OpenAI and Broadcom were making custom chips together. For what it’s worth, the latter’s CEO said recently that a new, unnamed client had put in an order worth $10 billion.
The Broadcom deal follows agreements that OpenAI recently struck with both NVIDIA and AMD. NVIDIA is investing $100 billion into OpenAI and will provide it with 10 gigawatts of AI infrastructure. The deal with AMD is for six gigawatts of compute power. OpenAI is said to be paying AMD tens of billions of dollars under that agreement and it could ultimately take up to a 10 percent stake in the company. As with the Broadcom rollout, both the NVIDIA and AMD deployments are expected to start in the second half of 2026. OpenAI also inked a deal with Oracle in July for 4.5 gigawatts in data center capacity as part of its Stargate Project.
According to recent reports, OpenAI CEO Sam Altman told employees that he wanted the company to build out 250 gigawatts of compute power over the next eight years, significantly up from the 2GW it’s expected to have by the end of this year. (For context, 250GW is about a fifth of the energy generation capacity of the entire US, which sits at around 1,200GW.)
As things stand, it would likely cost around $10 trillion to buy that much capacity. Altman said OpenAI would have to develop new financing tools to make that happen, but he hasn’t elaborated much on what those might look like. Even its current deals have OpenAI on the hook for hundreds of billions of dollars.
While the likes of NVIDIA and Microsoft have invested heavily into OpenAI, there isn’t a backer on the planet that can plow $10 trillion into the company. As things stand, OpenAI is very, very far away from making up the difference in revenue too. It’s reportedly expecting to make $13 billion in revenue this year.
(Reuters) -Broadcom raised its annual forecast for revenue from chips that help in artificial intelligence work by 10% on Wednesday, and announced a stock split to take advantage of a rally in its shares this year.
Shares of the Palo Alto, California-based chipmaker surged 12% in extended trading.
The company expects $11 billion in revenue from AI-linked chips in 2024, up from its previous forecast of $10 billion.
Broadcom manufactures advanced networking chips that help move around vast amounts of data used by AI applications such as OpenAI’s ChatGPT, making it one of the beneficiaries of businesses heavily investing in the boom.
Broadcom recorded revenue of $3.1 billion from AI products during the second quarter.
The company, which has seen its stock rally more than 30% so far this year, after almost doubling in 2023, will carry out a 10-for-1 forward stock split, in a bid to make its shares more affordable for retail investors.
The split-adjusted trading is expected to begin on July 15.
Its custom chips unit has also attracted orders from large cloud providers looking to reduce their dependence on Nvidia’s pricey processors. Broadcom is widely considered to be making custom chips for Google and Meta.
Revenue from Broadcom’s semiconductor solutions segment, which houses its networking and custom chips, rose about 6% to $7.20 billion in the quarter.
“As the data center market moves to AI servers, Broadcom’s upside is extremely high. In many ways (Broadcom) will be the second-biggest beneficiary of this shift, next to Nvidia,” said Ben Bajarin, analyst at Creative Strategies.
Revenue from the company’s infrastructure software segment more than doubled, thanks in part to its purchase of VMware.
Broadcom raised full-year revenue forecast by $1 billion to $51 billion. It also raised its annual core profit projections and beat LSEG estimates for second-quarter adjusted earnings per share and revenue.
(Reporting by Arsheeya Bajwa in Bengaluru; Editing by Shilpi Majumdar)
Semiconductor stocks have been on fire over the past year or so as the sector received a nice shot in the arm from rising demand for artificial intelligence (AI) chips, which explains why shares of chip giant Nvidia have jumped a whopping 216% in the past year.
Nvidia boasts a 92% share of the data center GPU (graphics processing unit) market, and that dominant position has led to outstanding growth in the company’s revenue and earnings in recent quarters. It could also lead to another solid year of growth for Nvidia given the massive opportunity in AI chips.
According to a forecast by investment banking firm Raymond James, the market for chips powering generative AI applications could double in 2024. Given Nvidia’s position in that market, that growth should translate into a huge revenue bump: Third-party estimates foresee the company generating $76 billion in data center revenue this year. That would be a substantial jump over the $44 billion data center revenue Nvidia may have clocked in its fiscal 2024 (which ended Jan. 31).
However, Nvidia is not the only way to capitalize on the AI chip market. JPMorgan analyst Harlan Sur believes that chipmaker Broadcom(NASDAQ: AVGO) could become the second-largest supplier of AI semiconductor chips this year.
Broadcom could generate significant AI revenue this year
The JPMorgan analyst has an overweight rating on Broadcom stock with a price target of $1,550 — more than 20% above current levels. According to Sur, the chipmaker could generate AI revenue of $8 billion to $9 billion in 2024 thanks to its dominant position in the market for custom chips, especially high-end application-specific integrated circuits (ASICs) deployed for AI workloads.
According to JPMorgan, Broadcom reportedly commands a 35% share of the high-end ASIC market. This puts it well ahead of second-place Marvell Technology, which has a 12% market share. Broadcom’s solid position means that it is well-placed to make the most of a fast-growing opportunity.
The market for high-end custom ASICs was reportedly worth $13 billion to $18 billion in 2023. This market is expected to clock annualized growth of 20% in the long run as more companies look to manufacture custom AI chips. The good part is that Broadcom has reportedly built a solid pipeline of customers for its custom chips, reportedly landing two to three major deals last year.
Broadcom has been involved in designing every generation of Google’s tensor processing units (TPUs), and also counts the likes of Microsoft and Meta Platforms as customers. It is worth noting that all these companies have been working on custom chips to train AI models efficiently and cost-effectively.
Not surprisingly, Broadcom’s AI revenue has been growing at a nice pace of late. The company sold $1.5 billion worth of generative AI chips in the fourth quarter of its fiscal 2023 (which ended on Oct. 29), which was 16% of its total revenue. That translates into an annual revenue run rate of $6 billion. So JPMorgan’s estimate that Broadcom could generate $8 billion to $9 billion in AI chip revenue this year points toward a potential jump of 33% to 50% based on its annual revenue run rate last quarter.
The valuation makes the stock an attractive bet
Analysts anticipate Broadcom’s revenue will increase by almost 40% in its fiscal 2024 to $50 billion, including a $12 billion contribution from VMware, which it recently acquired. The stock currently trades at 14.6 times sales following a 105% spike in the past year.
Still, it is significantly cheaper than Nvidia, which trades at 38 times sales. Of course, Nvidia can justify its steep valuation thanks to its sizable AI business, which is much bigger than Broadcom’s. However, conservative investors may want to look for cheaper AI chip plays, which is where Broadcom comes in.
The company enjoys a healthy market share in custom AI chips, a market that’s expected to grow at a solid pace in the long run. Also, if Broadcom does hit $50 billion in revenue this year and maintains its current sales multiple, its market cap could increase to $730 billion. That would be a 25% jump from current levels. That indicates that it isn’t too late for investors to buy this AI stock, despite the terrific gains it has clocked in the past year.
Should you invest $1,000 in Broadcom right now?
Before you buy stock in Broadcom, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Broadcom wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Stock Advisor provides investors with an easy-to-follow blueprint for success, including guidance on building a portfolio, regular updates from analysts, and two new stock picks each month. The Stock Advisor service has more than tripled the return of S&P 500 since 2002*.
JPMorgan Chase is an advertising partner of The Ascent, a Motley Fool company. Randi Zuckerberg, a former director of market development and spokeswoman for Facebook and sister to Meta Platforms CEO Mark Zuckerberg, is a member of The Motley Fool’s board of directors. Harsh Chauhan has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends JPMorgan Chase, Meta Platforms, Microsoft, and Nvidia. The Motley Fool recommends Broadcom and Marvell Technology. The Motley Fool has a disclosure policy.
