- Intel further away from AI monetisation
- The market is becoming more selective about its AI winners
After ChatGPT emerged as a global phenomenon, investors rushed to acquire any semiconductor stocks they believed could benefit from the demand for more computing power. However, the latest US earnings season has shown that just because a semiconductor company says it will benefit from artificial intelligence (AI) demand doesn’t mean it will. At least, not imminently.
Underwhelming results at semiconductor designers Intel (US:INTC) and AMD (US:AMD) in the first quarter of this year have seen their share prices drop 18 per cent and 8 per cent, respectively, in the past month.
Both companies compete in the design of central processing units (CPUs), and their biggest markets are in PCs and non-AI data centres. Intel was the leader in these markets but poor decisions in the past few years have allowed AMD to catch up.
The issue for both is that in AI most of the demand is for graphics processing units (GPUs), which have more cores and can do parallel computing. As most people know by now, Nvidia (US:NVDA) is the leader in this market. Intel and AMD are both trying to convince investors they can catch up, but recent results suggest they have a long way to go.
Intel’s unfulfilled promises
Revenue at Intel increased 9 per cent to $12.7bn in the three months ending in March, which was just ahead of analysts' expectations. However, the shares fell on guidance that its gross margin would weaken in the second quarter – to 40.2 per cent, from 41 per cent in the first.
Chief executive Patrick Gelsinger tried to convince analysts on its earnings call that things would subsequently improve, with revenue to be boosted by “momentum” for AI PCs. “We expect sequential revenue growth to strengthen throughout the year and into 2025,” he said.
This should underpin an improvement in gross margin, which Gelsinger is forecasting will strengthen next year and ultimately hit a target of 60 per cent. This strategy relies on it successfully being able to produce 18A chips, which will be the equivalent of Taiwan Semiconductor Manufacturing Co's (TW:2330) smallest two-nanometer chips.
The problem is that, for now, AI remains a small part of its business – 59 per cent of its revenue last quarter came from its client computing group, which produces processors for PCs. This market boomed during the pandemic, but has dried up since. Intel's first-quarter revenue was 31 per cent below the same period in 2022.
Its share price has remained resilient, though, as the company has been earmarked for billions in US government subsidies to fund a $100bn spend on local manufacturing facilities.
However, construction plans have been pushed back, with plants in Arizona and Ohio at least a year behind schedule, and, based on the share price movements, it appears investors have lost some confidence in its AI timeline.
“Investors are wary of the foundry initiative despite reported progress on the two-nanometer 18A node and a path to 1.4nm 14A later in the decade,” wrote Melius Research analyst Ben Reitzes.
AMD nearer to breakthrough
Although AMD has been taking market share from Intel, especially in data centres, it has also been hit by the downturn in demand for PCs. Its share price also proved resilient, though, boosted in recent months by the news that the company has designed a parallel computing chip that could rival Nvidia’s GPUs. Chief executive Lisa Su said the Instinct MI300X chip was the “highest performing accelerator in the world”.
AMD’s revenue was up 2 per cent year on year to $5.47bn but down 11 per cent on the previous quarter and below analysts' expectations. This was driven by poor performance in its gaming and embedded segments, both of which reported revenue declines of almost 50 per cent year on year.
Unlike Intel, AMD does not manufacture its chips so can quickly step up supply. It argued that revenue from AI would come in at $4bn this year, up $500mn on its previous estimate. Although this should have been seen as a positive, the market was expecting more.
This would equate to less than a fifth of last year's total revenue and just 4 per cent of the AI revenue that Nvidia is forecast to make from data centres this year. In addition, there is concern “about a potential AI order cut at a big customer and a high-bandwidth memory issue”, according to Reitzes.
Su tried to put demand concerns at ease on the company's earnings call, arguing that supply was its main constraint. She also pointed to the growing inferencing market as an area of opportunity. “MI300X is in a sweet spot for inference,” she said. “As we bring in additional products later this year into 2025, that will continue to be a strong spot for us.”
Reitzes sees the drop in AMD’s share price as a chance to buy. He is forecasting that AI revenue will increase to $9bn in 2025, driven by “an explosion in demand from inferencing”. The rising capex figures at computing companies are also seen as structural drivers. “Winners like Nvidia and Broadcom (US:AVGO) are obvious – increased capex at the big cloud should benefit AMD as well, as Microsoft and Meta place orders for inferencing workloads in particular,” he said.
In both cases, the share price declines demonstrate that the market is becoming more discriminate about which companies deserve the growth multiples attached to AI stocks, with missed expectations likely to be heavily punished.
