- Intel intends to start rolling out three CPU platforms for each new process it builds instead of the two it did previously.
- The company is perhaps doing this to help it cut down on R&D costs.
- The move could also potentially allow Intel's semiconductor rivals to catch up with the company.
A peek into Intel's (NASDAQ:INTC) latest 10-K unearths some interesting findings, key among them that the company was finally kissing goodbye to its famous ‘‘tick-tock’’ chip manufacturing cycle in favor of longer processes. Under the tick-tock development model, Intel used to launch two CPU platforms for each manufacturing process (e.g. as 28nm, 20nm, 14nm) that it adopted. The initial platform (the tick) delivers power and performance benefits from higher transistor densities that it made possible by moving an existing architecture to a newer process. In the tock stage of the cycle, a brand-new architecture optimizes Intel’s use of the new process. A typical tick-tock cycle used to last two years before a new one kicked off.
But now Intel intends to roll out not two but three CPU platforms for every process that it builds. The new development model is already at work as evidenced by the company’s announcement in July of 2015 that it was planning to delay the launch of its first 10nm CPU platform dubbed Cannonlake to the second half of 2017 as opposed to 2016 as per previous expectations. Intel said that it will instead roll out a third 14nm cycle dubbed Kaby Lake to succeed the current Skylake and Broadwell cycles:
"We expect to lengthen the amount of time we will utilize our 14 [nanometer] and out next-generation 10 [nanometer] process technologies, further optimizing out products and process technologies while meeting the yearly market cadence for product introductions,"
Perhaps Intel’s radical shift in strategy has been driven by the company’s need to lower its capex. Developing a new manufacturing process gobbles up a huge amount of money. Indeed, Intel and fellow chipmakers including Taiwan Semiconductor (NYSE:TSM) are some of the biggest capex and R&D spenders in the tech world. Intel recently cut its 2016 capex by $500M to $9.5B. That’s considerably lower than more than $11B/year the company used to spend on capex a few years back. By launching three CPU platforms for every manufacturing process as opposed to two previously, Intel can manage to cut down on its capex considerably.
The new plan also points to a shift in the way Intel intends to be operating going forward. In the past, Intel relied heavily on developing almost everything in-house. But with the likes of Samsung Electronics (OTC:SSNLF) and TSMC constantly breathing down its neck, Intel has decided to change the way it plays the game. The company recently acquired Altera for $16.7B, the largest acquisition in the company’s history. But Intel said that it’s just getting started in the M&A game. Intel’s chief dealmaker Wendell Brooks recently said that the company was looking beyond its traditional approach of doing everything in-house saying:
"We're doing things very differently than we have in the past,"
By spending less on capex, Intel will free more cash to develop its more promising businesses and for M&A activity in areas where it’s been lagging behind.
Intel’s new modus operandi is not without risks though. By extending the lifecycle of each of its processes, Intel will be giving its rivals a chance to eliminate or even overtake the company’s historical manufacturing process edge. Intel has historically been able to maintain at least a 6-month lead over its rivals for any new manufacturing process. This gave the company a distinctive first mover advantage. But Intel now risks losing this lead. For instance, TSMC plans to commence production of its 10nm process in Q4 2016, then migrate to 7nm sometime in 2018. Although TSMC’s 10nm process is believed to feature lower transistor density than Intel’s 10nm process, this will be a significant win for TSMC because Intel probably won’t be launching its 10nm process till 2017.
In the short-term Intel’s new manufacturing strategy is likely to free up a lot more cash that the company can use to develop its more promising business segments as well as in M&A activity, which is a good thing. At the same time, such a move is likely to give the likes of Samsung and TSMC a chance to catch up with Intel in new processes. This is unchartered waters for Intel but investors probably won’t know the full effects till 3-4 years down the line.