Apple is, perhaps, the only company that has pushed to adopt newer and more advanced 3nm chips since Qualcomm and MediaTek. However, there is uncertainty over whether the company would follow this through, owing to ‘unclear sales prospects for Android handsets’.

Moreover, with reports of big-ticket customers like AMD, NVIDIA, and Intel having slashed orders with TSMC, its utilisation rate is likely to reduce significantly across all process technologies. This may include 7nm/6nm declining ~50% in early 2023, the mode advanced 5nm/4nm process, and more importantly, even 28nm class, which have been fully loaded since the beginning of the chip deficit in early 2021. 

Samsung reported a 69% decline in profits for its December quarter, citing weak demand and pricing factors. DRAM and NAND flash memory prices have declined alongside the slack demand, while chip inventories are piling up across the supply chain.  

However, despite the grim state of the current global chip supply, experts aren’t perturbed. They believe it to be a part of the natural cycle of the industry. The industry is currently experiencing a downturn—the upturn during the pandemic (which had a sudden rise in demand and thereby a chip shortage globally) has come back to normalcy—leaving chipmakers with large inventories but no demand for the same. 

Cost vs pricing: The struggle

Sravan Kundojjala, principal industry analyst at Strategy Analytics, told AIM that for the past two years, the semiconductor supply chain was contending with inflationary pricing driven by a sudden uptick in demand, shortage of key materials such as gases, the Russia-Ukraine War and increased labour, freight, and energy costs. These factors also contributed to the pricing across the semiconductor value chain. As a result, the suppliers started to pass along increased costs to their customers to cope with inflationary costs and to protect margins. 

“Unsurprisingly, most semi-companies enjoyed healthy pricing in the past two years. NXP, for example, saw three straight years of price increases. Analog Devices saw up to 50% of its growth driven by pricing alone. Even TSMC saw a 20%+ jump in its wafer pricing in 2022 vs 2021,” Kundojjala pointed out.

As of now, the entire semiconductor ecosystem is struggling to find the right balance between cost and pricing. If we speak of Apple alone, the company was trying to maintain a stable price point by releasing the non-Pro models of its flagship iPhone 14 (iPhone 14 and iPhone 14 Plus) with the A15 (5nm) bionic chip while powering the Pro models (iPhone 14 Pro and iPhone 14 Pro Max) with A16 (4nm). Now, with it moving to the 3nm class for the A17 chip, Apple will continue to be under test each time it chooses to employ a denser chip, as the cost-per-transistor increase will become higher as we move towards smaller process nodes.  

However, unlike Apple, Android companies will likely wait until 2024 for 3nm. This is because, in addition to the worsening demand situation, the timing has not been right for them. Besides TSMC, only Qualcomm and Samsung have been using the leading edge until recently. MediaTek joined last year. If Samsung had perfected its 3nm technology (which it developed using the GAA-FET architecture) last year, Qualcomm would have been the first to market. 

Android still making its mark

Android’s failure to bank on the 3nm ahead of Apple will affect its competitiveness in the market. Still, experts like Kundojjala believe it is a safer bet for them to wait until 2024 as Android companies are conscious of initial costs since their premium tier volumes are not comparable to Apple. There is simply not enough room for cost efficiencies in premium tier chips such as Snapdragon 8 Gen 2. And increasing prices to accommodate the logical scaling of chips will also affect their margins. 

However, this is not the end of the story for Android. Process technology only goes so far. Companies look to optimise chip architectures by seeing how hardware and software components work together, so they can efficiently solve a problem in a given set of constraints. This way, chip companies include all the IP blocks and work closely with customers to satisfy their needs. In addition, these companies offer software SDKs to OEMs, who take advantage of underlying capabilities.

For instance, if Samsung wants to include more AI functionality, it will ask Qualcomm to enhance that aspect in their designs. Qualcomm then introduces the necessary changes in its designs. Thus, there are a lot of ways to increase performance that include adopting new CPU/GPU cores and improving AI and camera performance. However, Kundojjala says, “The costs continue to go up for advanced chips due to increased functionality on the processing side (CPU, GPU, AI, multimedia, camera, gaming, and security, etc).” And so, ultimately, it is the balancing act of how to enjoy better pricing for the rising costs. He says, “We should expect Qualcomm to enjoy better pricing on 3nm in 2024, compared to 4nm in 2023.”  

Furthermore, Android is inching closer on certain benchmarks to Apple’s A chips. For example, in one demonstration pitching Snapdragon 8 Gen 2 against Apple A16 Bionic, the AI engine in Qualcomm’s chip ran 4.5 times faster than its last-gen SoC and is said to be ahead of Apple’s neural engine, albeit only by a margin. On the 5G and wireless technology end, Qualcomm has been the leader since the A16 also uses a Qualcomm modem. However, this might not stay for long as Apple plans to go in-house and reduce its reliance on third-party vendors. Besides, Qualcomm has also done an excellent job in improving power efficiency in the GPU department. 

The post 3nm Chips: Apple Innovates, Android Disrupts appeared first on Analytics India Magazine.