- Intel delays the deployment of its EUV lithography by 2 years.
- Non-cash depreciation expenses could cause a drag on Intel's bottom-line until 2019.
- This could also allow companies such as Samsung, GlobalFoundries and Taiwan Semiconductor to catch-up.
Intel Corporation (NASDAQ:INTC) has spent billions in capital expenditures for several years straight, to maintain an edge over other fab companies. Intel was planning to keep the trend alive; it was gearing to deploy capital intensive EUV lithography to manufacture its 10nm chips next year. This would have further bolstered its lead over the competition. But owing to delays in retrofitting manufacturing equipment, and probably due to yield issues, Intel has now reportedly delayed the deployment of EUV lithography to 7nm node, perhaps till 2019. This pretty much allows competitors such as Samsung, GlobalFoundries and Taiwan Semiconductor to catch-up.
What Is EUV Lithography?
We have seen several node shrinks over the past decade, where each iteration delivered promising performance, efficiency and power gains. But with each passing shrink, the laws of physics kick in. We have arrived at a point where shrinking nodes further, beyond 10nm, requires sizable investments on the part of chipmakers, with little performance gains to show for. So it's just not feasible anymore to keep up with node shrinks if chips are manufactured the conventional way. This is where EUV comes in.
Its worth noting that the industry currently uses a technique termed as “Deep Ultra Violet” lithography to fabricate chips. It’s similar to the technology found in cameras: You focus light through lenses to etch circuits onto silicon wafers. The problem here is that there’s a limit to how precise and diametrically small these focused light beams can be. The laws of physics play a spoiler here.
But EUV lithography, or Extreme Ultraviolet light has much smaller wavelengths that allow focusing and controlling of EUV lasers on a smaller scale, relatively easily. This, in turn, allows chipmakers to carve circuits onto silicon wafers on a much smaller scale and we can theoretically continue to shrink nodes down to 5nm without facing any issues. A few speculative reports suggest that chips manufactured using EUV lithography are up to 100 times faster than their conventional counterparts. Whether this claim is true or not, still remains to be seen. But the established fact here is that EUV lithography aids in continued node shrinks.
I may have oversimplified things here a bit. But the fact of the matter is that the next-gen EUV lithography won’t really defy the laws of physics, at least not for now. It pushes the boundary of conventional node shrinks a bit further and allows us to shrink nodes a bit further, in today’s time, with today’s technology.
What’s the fuss with the delay?
The concerning thing for Intel shareholders is that the chipzilla was earlier hoping to manufacture 10nm chips using EUV lithography sometime around next year. None of its competitors had such an aggressive release timeline. So the ramp of EUV was seen as a key factor that would have kept Intel ahead of its competition.
But Intel is now looking to deploy EUV lithography at the 7nm node. I believe that this would happen sometime around the 2019-20 timeframe, assuming everything goes as per plan and the chipzilla doesn’t face any ramp issues with its 10nm node. The key thing to note here is that Intel’s competitors, Samsung, Taiwan Semiconductor and GlobalFoundries plan to introduce EUV lithography along the same timeline as well.
This could be bad for Intel. I’ve mentioned in my previous article that all four companies will share the same node from 2018-19 onwards. And without the EUV advantage, Intel would pretty much be at par with its competitors in terms of chip fabrication technologies. Its manufacturing lead would shrink to nothing.
So this was the business impact, now let’s take a look at this delay from a financial standpoint.
First and foremost, deploying EUV lithography is an expensive activity, with each alpha tool costing as much as $100 million. Intel is believed to have already spent $1 billion as of April last year on these “tools”. So the cost would have only inflated by now, considering that it’s been more than a year since then.
But the thing to note here is that Intel ramped up these purchases in the hope of deploying the next-gen technology commercially next year. With the delay of 2 years, Intel would be sitting on the matter, without generating any returns out of its expensive equipment. It’s non-cash depreciation expenses would continue to pile up over the next 2 years, causing a drag on its profitability, without seeing a single dollar in revenue from its next-gen EUV tech. So this could particularly hurt Intel’s bottom-line.
Secondly, with Intel’s manufacturing lead shrinking over the next few years, ARM-based server offerings could pick up steam. Companies such as Qualcomm, AMD and Cavium would get access to same EUV lithography, at the same nodes, by 2019, which could theoretically bring the performance of their chips close to Intel Xeons. So Intel may struggle to curb the growth of ARM-based offerings in the server space, leading to a loss of the chipzilla’s revenue and profits.
Putting it all together
Intel has delayed the deployment of a technology that would have potentially aided in retaining the chipzilla’s manufacturing lead over competition. Now, all four companies will have similar chip fabrication technologies by 2019-20. I believe that this would be bad for Intel, both operationally and financially speaking. The chipzilla needs to get its house in order if it wants to curb the growth of its competition.