Nvidia stock declined last month after forecasting a drop in its revenue growth rate, according to an August Forbes post.
The AI chip designer dominates the fast-growing market for such chips with a more than 80% share. So, Nvidia’s slower growth — from 122% in the second quarter to 80% forecast in the third — raises questions for investors:
- Is the AI chip market slowing down?
- Are rivals — drawn in by the market’s size and growth and the company’s high profitability (gross margin of 71.8% in the second quarter) — cutting into Nvidia’s market share?
- Or will Nvidia’s growth accelerate as it works out the glitches in its new chip and builds on capabilities that developers value and rivals can’t copy?
The short answers are yes, no, and maybe.
- AI chip market growth is slowing down. Gartner forecast the AI chip industry would increase 33% in 2024. Between 2024 and 2029, AI chip revenue is forecast to grow more slowly — at a 20.4% compound annual rate to nearly $312 billion, according to MarketsandMarkets.
- Rivals are not taking market share from Nvidia. Nvidia is gaining share in the GPU market. Between the fourth quarter of 2023 and the first quarter of 2024, the company’s market share rose from 80% to 88%, noted a Jon Peddie Research report featured in a June Forbes post. During the same period, AMD’s share of the GPU market fell from 19% to 12% while Intel’s share declined from 1% to “negligible,” according to the report.
- Nvidia growth might accelerate. Nvidia is facing manufacturing difficulties with its newest chip — dubbed Blackwell. Meanwhile, startup Cerebras Systems’ competing design is faster and cheaper than Nvidia’s product, the startup says. If Nvidia can solve Blackwell’s technical problems and continue an annual pace of launching powerful new chips, the company’s enormous software advantage could help accelerate its revenue growth.
If the company resumes the triple-digit growth rates it achieved in recent quarters, Nvidia stock could rise. If investors think the AI chip designer’s growth is likely to keep decelerating, now could be a good time to sell the stock.
MORE FOR YOU
Reasons Nvidia Growth Could Accelerate
Two drivers — especially in combination — could accelerate Nvidia’s growth. These include an increase in the rate of demand growth for AI chips and Nvidia’s successful launch of new chips that deliver a much higher return on investment for customers than previous versions did.
Growth In Demand For AI Chips Depends On The Emergence Of A Killer App For Generative AI
The growth in demand for AI chips depends on whether a killer app emerges for generative AI. While companies were initially afraid of missing out on the promise of AI chatbots, they are growing terrified of being sued if the technology hallucinates.
That is making them hesitant to deploy their AI investments. “Boards are going bipolar,” Yoav Shoham, co-founder of AI21, an AI software maker competing with OpenAI, told me in an August interview.
Of 200 to 300 generative AI experiments the typical large company is undertaking, a mere 10 to 15 have been rolled out internally, and perhaps one or two have been released to customers, according to my June interview with Liran Hason, CEO of Aporia, a startup offering guardrails to protect companies from AI hallucinations.
To unlock further demand, a generative AI killer app — that gives many people an overwhelming need to use a new technology because it saves them so much time and money — must emerge. For example, the killer app for the personal computer was the electronic spreadsheet while the iPod's was the iTunes store.
Nvidia Says Its Chips Deliver The Industry’s Highest Return On Investment
Customers are enjoying a fast return on investment on the company’s chips. “The people who are investing in Nvidia infrastructure are getting returns on it right away,” Nvidia CEO Jensen Huang said on an August 29 call with analysts. “It’s the best ROI infrastructure, computing infrastructure investment you can make today.”
What is the source of Nvidia’s return on investment? Despite charging a higher price than competitors do, the company’s chips perform better and cost less to run — more than offsetting their higher price due to their lowest total cost of ownership, according to my new book, Brain Rush.
Demand for Nvidia’s Blackwell chips is “incredible,” Huang said in an August 29 press release. “Global data centers are in full throttle to modernize the entire computing stack with accelerated computing and generative AI,” he added.
Nvidia said it shipped samples of Blackwell chips during the quarter and changed the product to make it more efficient to manufacture. “In the fourth quarter, we expect to ship several billion dollars in Blackwell revenue,” Nvidia CFO Colette Kress said in an August 29 statement.
Why Nvidia Earns Higher Margins Than Rivals Do
Compared to AMD and Intel, Nvidia is growing faster and earning more profit. Nvidia’s most recent growth rate — 122% in the second quarter — outpaced AMD (8.9%) and Intel (-1%). Nvidia’s 71.8% gross margin was also higher than rivals’ — AMD (49%) and Intel (35%), according to Yahoo! Finance.
Nvidia’s relatively high margins are due to higher prices and higher demand. “I was in a meeting at one of Nvd’s competitors once and the CFO said to a room filled with C level executives, ‘If your products are so damn good, why is the GM so damn low?’,” Peddie wrote me in an August 30 email.
“And the answer was, although no one dared say it, ‘We don’t really think we’re as good as they are so our differentiation is price’, ” he added.
Two essential capabilities enable Nvidia to outgrow and outearn rivals:
- A stream of improved AI chip designs. Nvidia keeps developing new chip designs that surpass rivals’. “The company has developed a full stack and only uses outside memory vendors to build its systems,” according to Peddie’s email. “It’s got a huge team of the best engineers in the industry. It is producing developments and papers faster than it can patent them all. And it has invested billions of dollars in R&D. Its R&D budget is bigger than the sales forecast of the startup — that’s a headwind for the startups that will be very difficult if not impossible to overcome,” he added.
- Better software for Nvidia developers. No other chip company has software to rival Nvidia’s CUDA, noted Brain Rush. “Nvd works with their customer, of which it has more than all others combined,” Peddie’s email noted. “These customers develop programs on Nvd chips, and most of them share their development with Nvd as a way of influencing the next generation design. So not only does Nvd have tens of thousands of SW & HW engineers (probably 5:1 SW), which is more than all its competitors except Intel. It also has a ghost force of thousands more in universities and customer sites,” he added.
Why Nvidia Revenue Could Slow Down
Despite these advantages, Nvidia’s revenue growth is clearly slowing down. Technical problems with Blackwell may have contributed to that slowdown. If rivals can deliver chips that work faster and cost less, perhaps coders will write software for competing chips — costing Nvidia market share.
The Product Issues Holding Up Nvidia’s Blackwell Chips
Nvidia said product issues were largely responsible for a decline in gross profit margins between the first and second quarters of 2024, noted the Wall Street Journal. More specifically, the company’s second quarter gross margin of 75.1% declined 3.3 percentage points from the previous period.
While Nvidia did not provide details of the product issues, analysts and industry executives said the large size of the Blackwell chips led to manufacturing challenges. That’s because Nvidia’s design requires the joining of “two advanced new Nvidia processors and numerous memory components in a single, delicate mesh of silicon, metal and plastic,” noted the Journal.
The complexity of the chips boosts the odds they will not operate properly. Defects in and heat from any component can cause the whole chip to fail — lowering the percentage of usable $40,000 Blackwell chips coming out of the manufacturing process, the Journal wrote.
Nvidia has significant experience dealing with such challenges. The technical challenges include “race problems – speed is everything and moving data from one part to another creates timing challenges,” noted Peddie’s email.
“Zillions of hours in a simulator are spent trying to find them all and correct, and still once the silicon is made, new ones show up. Nvd (and other semi companies) know this and design in compensations and redundancies to work around it if an issue pops up,” he added.
It will take time before it becomes clear whether Nvidia will overcome these technical challenges.
Startup Cerberas Systems Says Its Chip Is Faster And Cheaper Than Nvidia’s
Cerebras Systems is a $4.2 billion (2021 private market valuation, according to PitchBook), 430-employee Sunnyvale, Calif-based chip-making startup that competes with Nvidia. The company has taken a different approach to designing Blackwell-like chips — which enables its chip to process information faster and less expensively than Nvidia’s.
“To do meaningful work in AI, you need a huge amount of compute, and that converts to many transistors, many more than can fit on a single chip,” Andrew Feldman, a Cerebras founder, told the Journal. “The technology to get to two [chips] is difficult to develop, the technology to get to four is harder, and to get to eight, even harder,” he added.
Cerebras — whose customers include AstraZeneca and the Mayo Clinic — “figured out how to connect and operate as one huge chip,” noted the Journal. Moreover, last week the startup launched a cloud-computing service for AI inferencing — namely responding to natural language user queries — after the AI chatbot has been trained.
Cerebras — which filed confidentially for an IPO in April and could go public in October 2024 — claims the inferencing service operates “with 20 times the speed at a fraction of the cost” of competing chips, according to AI News.
Using the company’s Wafer-Scale Engine, Cerebras claims its service is 20 times faster — achieving speeds of 1,800 output tokens per second — than the usual hyperscale cloud products using Nvidia’s GPUs, AI News wrote. What’s more, the Cerberas-powered service is “also more cost-efficient,” sources told AI News.
Cerebras’ Faster Speed And Lower Cost May Not Be Enough To Challenge Nvidia
While large companies might opt for Cerebras to save time and money, smaller ones are likely to stick with Nvidia. “The key question is whether enterprises are willing to adapt their engineering processes to work with Cerebras’ system,” David Nicholson, an analyst at Futurum Group, told AI News.
Since Nvidia already offers established solutions, smaller companies lacking the larger firms’ capital may consider it worth changing their engineering processes, noted Nicholson.
While it could be years before such systems work flawlessly, the future of generative AI could be serving as agents — for example, enabling consumers to pick an itinerary and delegating to the AI agent the process of searching for flights and buying the best ones at the lowest price, Brain Rush explained.
Cerebras’ technology is well-suited for powering such AI agents which “must operate rapidly, repeatedly, and in real-time,” AI News noted, due to “Cerebras’ 16-bit accuracy and faster inference capabilities.”
If Nvidia concludes Cerebras has built a better mousetrap, the AI chip leader could build something like the startup’s WSE. “If a giant chip like Cerebras was the best answer, don’t think Nvd could and would build one too? What can any startup do that Nvd hasn’t done or could easily replicate?” Peddie concluded.
Cerebras — which is operating its own data centers t0 provide the inference service — aims to win Microsoft and Amazon as customers, noted Bloomberg. Could Cerebras take market share from Nvidia? “Enough to make them angry,” Feldman told Bloomberg.
