UMC - Semiconductors Winners And Losers At The Start Of H2 2023

2023-07-06 00:23:31 ET

Summary

• Semis look poised for continued gains in Q3 after the rally continued in Q2, but there were several changes worth noting.
• The rally in semis is increasingly being driven by AI after other tailwinds lost strength, which could become problematic if AI disappoints.
• H2 2023 is facing a flare-up in export controls, which have the potential to seriously disrupt supply chains, depending on how governments approach the issue.
• The rally in semis can continue since tailwinds remain, but new buyers may want to think twice about getting in at this time.

Semiconductor stocks continue to build on the rally that started in late 2022. Semis got off to sluggish start in Q2 2023, but they took off in May with artificial intelligence or AI playing a key role. However, while the semiconductor sector as a whole looks poised to build on its gains with Q3 upon us, divergence seems to have popped up as an increasing number of semis failed to participate in the rally. Some may be better off and some may be worse off for various reasons. Why will be covered next.

The trend continues to favor higher prices for semis in Q3 2023

A previous article at the start of Q2 2023 concluded that semiconductor stocks as a group, and HPC names in particular, were likely to add to their strong gains in Q1 due to the existence of several tailwinds. Specifically, semis are the beneficiary of a less restrictive monetary policy from the Federal Reserve and other central banks, increased interest in semis from the investment community due to AI and an industry outlook which calls for the end of the downturn in the market for semiconductor chips and the start of a rebound that will pave the way for a year of strong growth in 2024.

As it turned out, Q2 was a good quarter, but not quite as good as Q1 for semis. The iShares PHLX Semiconductor ETF ( SOXX ), for instance, started Q2 with a YTD gain of 28% and finished with a YTD gain of 46% at the end of Q2. In comparison, the Invesco QQQ Trust ( QQQ ) gained 39% and the SPDR S&P500 ETF ( SPY ) gained 16% during the same period.

Furthermore, the trend argues in favor of continued gains at the start of H2 2023. Note the current trend with higher lows and higher highs in the chart below. It’s true trends can and do change, but there is also a saying that the trend is your friend and the trend suggests that sticking with semis is a bet worth taking heading into Q3.

Q2 saw double-digit gains, similar to Q1, but there were other differences worth noting

Still, it’s worth mentioning that unlike Q1, semis got off to a poor start in Q2. In fact, most of the gains came in the month of May. If May is excluded, then semis did not perform all that well as April and June did not add as much. Semis actually lost ground in April, but May came to the rescue after a number of high-profile companies showed how much of a difference AI could be.

Other companies that are not exposed as much to AI did not fare quite as well, creating a divergence in the process. A look at the individual stocks present in an ETF like SOXX makes this clear. The 30 companies included in SOXX includes Nvidia ( NVDA ), Broadcom ( AVGO ), Advanced Micro Devices ( AMD ), Intel ( INTC ), Texas Instruments ( TXN ), Microchip ( MCHP ), NXP Semiconductors ( NXPI ), ON Semiconductor ( ON ), Applied Materials ( AMAT ), KLA Corp ( KLAC ), Analog Devices ( ADI ), Qualcomm ( QCOM ), Lam Research ( LRCX ), Marvell ( MRVL ), Micron ( MU ), TSMC ( TSM ), ASML ( ASML ), Monolithic Power Systems ( MPWR ), Skyworks ( SWKS ), Teradyne ( TER ), Entegris ( ENTG ), STMicroelectronics ( STM ), Lattice Semiconductor ( LSCC ), Qorvo ( QRVO ), Wolfspeed ( WOLF ), MKS Instruments ( MKSI ), United Microelectronics Corporation ( UMC ), ASE Technology ( ASX ), Silicon Laboratories ( SLAB ) and Synaptics ( SYNA ). The table below shows the recent gains or losses for these stocks.

Source: iShares

Who was hot and who was not among semiconductor stocks

Semis posted strong gains in Q2, but it took some time for this to happen. SOXX, for instance, began Q2 with the stock priced at $444.67 as of March 31 and as late as May 24, SOXX was priced at $430.93. SOXX was flat to down for most of Q2, but a strong rally following May 24 helped SOXX end Q2 with strong gains.

This turnaround did not happen for no reason. May 24 happens to be the day NVDA released its most recent earnings report with guidance that blew part expectations thank to AI. NVDA soared higher, invigorating the sector and SOXX by extension. It also allowed NVDA to top all others in SOXX for the second consecutive quarter with a 54% gain in Q2, pushing its YTD gain to 189%, which is way ahead of everyone else.

MRVL and AVGO both made sure to mention AI in their quarterly reports, which likely helped them end up as distant runner-ups with Q2 gains of 38% and 37% respectively. However, not everyone fared as well. SOXX gained a respectable 15% in Q2, but 9 of the 30 stocks in SOXX actually finished with losses in Q2, unlike Q1 when just one stock did not post any gains.

This group was led by SYNA, which lost 22.5% and WOLF with 12.5%, but it also includes UMC, SLAB, QCOM, STM, SWKS, TXN and ASX. Another two, ADI and MU, posted gains in Q2, but only barely with gains of 0.31% and 0.03% respectively. LSCC and QRVO gained just 1-2%. In other words, the broad-based rally seen in Q1 became much less so in Q2.

The sector as a whole continued the rally that started in late 2022, but it was driven by a smaller number of companies than before, NVDA in particular and those who are perceived to stand to benefit from AI, whether in the form of increased demand for server GPUs or high-speed connectivity for hyperscale datacenters.

It’s not by accident that NVDA and AMD are the top two gainers on a YTD basis. AI has given HPC names a major boost, which helped mask weakness elsewhere. AI essentially bailed out stocks that probably should not have performed quite as well. INTC, for instance, is a prominent name in HPC and a competitor of NVDA and AMD, but one could argue that INTC should have done significantly worse than its 27% YTD gain, if not for AI. INTC as a company has not done as well as the 27% gain in the stock suggests, whether it is regaining lost market share, technological leadership or getting its foundry business off the ground, especially not in comparison to what was expected when its much heralded CEO was brought in.

In contrast, SYNA and WOLF are the only two who have yet to gain six months into 2023. Neither has any AI exposure worth noting to offset their company-specific problems like INTC, which might help explain why they are in the position they are in. Overall, Q2 was a productive quarter for semis, but there was divergence, which was masked to a certain extent by strong headline numbers for the sector such as the 15% Q2 gain for SOXX.

What could cause further divergence among semis

If not for AI, semis as represented by ETFs like SOXX would probably not have gained as much as they did. AI will need to live up to expectations, since there may not be much else to keep the rally going. On the contrary, while some segments continue to hold up, the semiconductor industry as a whole is confronted with weaker-than-expected demand, the memory market in particular.

Some like NVDA managed to blow past expectations, but NVDA was more like the exception than the norm. Many semis actually acknowledged facing demand that was weaker than anticipated, which led them to downgrade their outlook. This includes a bellwether like TSM, which lowered its FY2023 revenue outlook from an increase in the low single digits to a decrease in the low single digits. As a consequence, the outlook for the industry was revised lower as well.

For instance, WSTS predicted at the start of the year that the semiconductor market would decline by 4.1% YoY in 2023, but the most recent update has increased this to a decline of 10.3% after incorporating all the recent updates from various industry players. MU, for instance, was the latest to report on June 28 and its quarterly guidance was well short of expectations, which is not a positive sign heading into the upcoming earnings season that will start in a few weeks.

Still, most industry forecasts continue to expect a year of strong growth in 2024. WSTS, for instance, calls for the semiconductor market to expand by 11.8% YoY to $576B in 2024, just ahead of the record $574B achieved in 2022. Nevertheless, the latest industry projections are significantly lower than earlier in the year when some called for the market to grow to well over $600B in 2024.

What will be the impact of the recent flare-up in the U.S.-China tech war?

There is another reason behind the weak guidance from MU. MU became the first semiconductor company to be explicitly targeted by China in the ongoing tussle between the U.S. and China in the field of semiconductors. This struggle has been going on for several years, but until fairly recently, China was relatively restrained in its actions, unlike the U.S. government, which has been much more active with export controls and other restrictions/sanctions on Chinese companies.

This may have changed with the recent flare-up in the ongoing back and forth between China and the U.S., the latter with the assistance of the Netherlands/Europe and Japan. The latest chapter in this saga deals with export controls on gallium and germanium imposed by China’s Ministry of Commerce. Gallium and germanium are used to make compound semiconductors like gallium arsenide or GaAs, gallium nitride or GaN and silicon germanium or SiGe.

On paper, a whole of bunch companies could be affected. For example, NXPI, STM, QCOM, WOLF and QRVO. In fact, QRVO could be the one most affected. While QRVO’s core business revolves around radio frequency chips like power amplifiers for mobile communications, QRVO is also a supplier to defense contractors like Raytheon ( RTX ) and Lockheed Martin ( LMT ), which is something China may take issue with.

It is not clear what exactly China hopes to achieve with these export controls. China may want to use their own export controls as leverage against export controls from the U.S. and its allies or China may have decided that withholding gallium in particular can have strategic consequences as it is an important element in military applications, including AESA radar systems for combat aircraft, naval combatants and air/missile defense systems.

Export controls do add a degree of uncertainty to affected companies. If China allows everything to go through, the impact could be negligible. If not, then the impact could be more serious. If the export of gallium and germanium are reduced by a limited amount, then workarounds are possible. But if everything gets blocked, then there is a serious problem as finding alternative sources for gallium in particular will take many years and be very expensive.

The risk is that these new export controls could trigger a domino-effect with more and more export controls. It depends on how China and others go about implementing export controls, but this is not a positive development for semis as they have the potential to seriously disrupt supply chains for a long time.

Will the Fed be forced to become more hawkish?

Another tailwind for semis has been Fed policy. The Fed has been taking its foot off the gas pedal and the prospect of a more dovish monetary policy has been good for tech stocks, semis included. Q2 saw the end of the Fed rate-hike cycle with the Fed opting to keep the Fed Funds Rate or FFR unchanged at 5-5.25% at the June meeting, making it the first time there was no rate hike since early 2022.

However, the Fed tacitly admitted that inflation was proving to be more stickier than anticipated. The dot plots thus left open the possibility of another 50 basis points in hikes in 2023. The Fed still expects lower rates in the coming years, but it also hiked interest rate forecasts for 2024 and 2025 to 4.6% and 3.6% respectively, up from 4.3% and 3.1% respectively.

In light of this, Fed Funds futures believe there is a 86.6% chance the FFR will be raised by another 25bps to 5.25-5.50% at the July meeting. But futures do not anticipate interest rates higher than 5.25-5.50%. The next change is likely to be a rate cut and not a hike, which futures estimate is likely to come in May 2024 when the Fed is projected to lower the FFR to 5.00-5.25% to start the cycle of rate cuts.

This is more hawkish than earlier in the year. Nevertheless, the overall trajectory has not changed very much despite the more hawkish Fed. Most of the rate hikes are in the rear view and the FFR is more likely to be lower than higher in the next few years. Overall, current Fed policy remains a tailwind for semis, although sticky inflation remains a wildcard that could force more rate hikes.

Could AI cause the semiconductor rally to fizzle out?

There is a lot riding on AI. Of all the tailwinds, AI was by far the main driver behind the gains in Q2. Other tailwinds were still present, but they lost some of their strength as they suffered some setbacks. AI, though, is still standing strong and recent reports from the likes of NVDA support the notion that AI can be a potent driver of growth for the industry.

Yet there is no denying that there is a fairly widespread view that AI is being overhyped. This is in part due to different people having different perceptions of what AI stands for. Many within the general public still think of AI as portrayed in blockbuster movies, but mankind is not any closer to that kind of AI today than it was back in the fifties when the term AI first came into being.

Adding to the confusion is that people are being bombarded by sensational claims that true AI experts would take issue with. For instance, the notion that the world is getting closer to being taken over by sentient machines. It doesn’t help that OpenAI, the company behind ChatGPT, has made a number of contentious claims as to how far along AI is.

These claims from OpenAI have been given additional credence due to ChatGPT’s success, but this does not take anything away from the fact that AI applications based on large language models or LLMs like ChatGPT suffer from unresolved problems like “hallucinations”. No one, and that includes OpenAI, has provided the scientific proof that these problems associated with LLMs can be resolved

As long as these problems remain, ChatGPT and other LLM-based AI applications will remain limited in the types of use cases they are suitable for. They may be okay for say the novice programmer who asks the same old questions that have been asked many times before, but they’re much less suitable when faced with proprietary programming code or text data that is not included in the training of the LLM.

Most will agree that AI can be of great benefit, especially if it is limited to addressing a specific problem that is limited in scope, but by how much remains to be determined. It’s not out of the question that AI, at least the one based on machine/deep learning, will not live up to everyone’s expectations. This could trigger a reset in expectations and thus the price people are willing to pay for certain semis, perhaps once the general public realizes that the world is not really any closer to the holy grail of artificial intelligence despite all the recent developments. Semis will be impacted if AI is not there to keep things going and nothing else is there to take its place.

Investor takeaways

Semis have posted strong gains in H1 2023 and they look poised to add to their gains in H2 2023. The tailwinds that enabled the rally in H1 are still around and likely to remain in H2, suggesting additional gains for semis. The Fed still holds out the prospect of easier monetary policy, investor interest in AI plays remains very high and the semiconductor market is still believed to be on the verge of a strong expansion despite the current slump. As long as these conditions remain, semis are likely to continue to do well.

Having said that, the outlook for semis has deteriorated in several ways at the midpoint of 2023, at least in comparison to the start of 2023. The Fed has been less dovish than expected and weakness in semiconductor demand has been greater than anticipated. AI is the one pillar that has remained the steadiest, which explains why it was mainly responsible for the gains seen in Q2.

However, this also means AI is shouldering more of the load. This is expressed in ETFs like SOXX where huge gains by a small numbers of stocks offset the fact that a large number of stocks lost ground. Stocks like QCOM, SWKS, QRVO, SLAB and SYNA have lost some or all their gains from earlier in the year.

NVDA has almost tripled in value after two quarters, but that’s very much the exception. If not for AI, SOXX could very well have finished flat to slightly down in Q2 instead of the 15% gain it ended up with. In other words, the headline gains are masking weakness in the sector. Semis are not in as good a shape as perceived by many.

If an increasing number of stocks sit out the rally and the rally is driven by a decreasing number of stocks, then that is cause for worry. It suggests that the rally in semis, which has lasted for three quarters and is entering its fourth quarter in Q3 2023, is getting long in the tooth. Lower returns in H2 than those seen in H1 is likely. A correction in Q3 or Q4 is possible.

The latter could become more likely if export controls by China, U.S. and its allies trigger a tit-for-tat series of disruptions to existing supply chains. Keep in mind that export controls do not necessarily mean the export of things like gallium or germanium is prohibited. It means government permission is required to export. So depending on how governments go about export controls, the impact could be more or less. It does mean greater uncertainty for companies, which is generally speaking no good for their stocks.

Bottom line, semis can go higher, but it would probably not be prudent to be a buyer of semis at this time as the rally may already be running on fumes. HPC names like NVDA are still the best bet if one wants to open a new position in semis. Just be sure to remember that no matter what, nothing goes up forever.

For further details see: