How Intel Lost the Mobile Market, Part 2: The Rise and Neglect of Atom

Update (6/1/2020): The posting under could have been penned in 2016, but it however stands up as a postmortem of what went mistaken with Intel’s cell endeavours — with a single pretty important omission. Back again in 2016, we did not know Qualcomm experienced been ruthlessly imposing licensing and acquiring phrases that manufactured it properly unattainable for companies to present Intel-based mostly cell units. I recall questioning why Intel couldn’t locate a single US firm to generate a mobile phone about its hardware system for really like or cash when the primary Xolo X900 compared nicely more than enough towards a then-recent Iphone.

Intel however manufactured a range of errors with Atom, as this posting discusses, but the reality that Qualcomm experienced a stranglehold on the current market behind the scenes obviously experienced an influence on what variety of achievement Intel was ever likely to obtain.

I truly favored the Xolo X900 machine I tested all all those several years back, and the Bay Trail tablets I experienced circa 2013 were good units. Atom’s cell endeavours will always stay an attractive may possibly-have-been.

First story under.

In Part 1 of this two-component series, we reviewed the distinction involving Intel and TSMC’s foundry designs and how these dissimilarities manufactured it exceptionally tricky for Intel to contend in cell. In Section 2 we’ll investigate the precise decisions Intel manufactured, the increase and neglect of Atom, and why the company’s outstanding foundry technologies wasn’t more than enough to conquer the current market.

The frequent explanation for why Intel lost the cell current market is that its x86 cell processors possibly drew also a lot power or weren’t strong more than enough compared with their ARM counterparts. Intel’s choice to provide its ARM division and XScale processor line in 2006 has been commonly derided as a important error. It is a basic, frequent-perception explanation with just a single flaw: It errors signs for lead to.

All of This Has Transpired Ahead of

Intel’s struggles in the cell current market did not commence with Medfield, Moorestown, or even the choice to provide its ARM business and XScale chip division ten several years back. As EETimes described in 2006:

Intel (Santa Clara, Calif.) expended more than $10 billion to enter the communications business above the several years, but the microprocessor huge lost its shirt — if not millions of pounds in the arena. The described communications-chip sale is said to be component of Intel’s plan to overhaul the firm. Intel is also set to include things like the layoff or redeployment of 16,000 staff, according to speculation from a single Internet internet site.

Improve “communications” to cell, change the range of fired staff, and that paragraph could’ve been penned today. Intel’s challenges in cell are not new Santa Clara has been struggling to enter new markets for virtually 20 several years. Other articles from 2006 emphasize that XScale sales experienced been rather minimal, as experienced profits from Intel’s networking and communications division.

From Intel’s perspective, marketing XScale manufactured perception. Creating a cell processor business about ARM cores would have minimal Intel’s capability to leverage its personal IP and knowledge in x86 production, while simultaneously slicing into its earnings (Intel would have owed significant royalties to ARM if these types of a style and design ever turned preferred). Atom was now nicely into advancement in 2006 and Intel decided to bet on its personal hardware knowledge and computer software advancement competencies.

Atom and the Increase of x86 Just about everywhere

Contrary to preferred belief, Intel wasn’t caught totally off-guard by the increase of smartphones or the recognition of tiny, Web-related units. Atom advancement began in 2004 the Silverthorne core that Intel debuted in 2008 experienced a TDP of just 2-3W at a time when most cell Core 2 Duo processors were stuck in 35W territory.


Intel’s primary Atom.

Most persons recall Atom as the chip that released a thousand netbooks, but that wasn’t Intel’s primary plan. The firm considered nettops and netbooks would be a niche current market for Atom, not the chip’s principal system. Atom and its successors were supposed to start an armada of Mobile Web Units, acknowledged as MIDs.


Units like the Gigabyte MID M528, demonstrated over, appear hopelessly quaint today, but Intel was evidently pondering about the upcoming of cell computing. The firm envisioned an ecosystem of netbooks and MIDs pushed by its personal personalized x86 architecture, a aim the push dubbed “x86 almost everywhere.”

Conflicting Priorities

There were evidently executives at Intel who recognized how important cell would be to the company’s prolonged-phrase upcoming and pushed for aggressive positioning and products ramps. Regrettably, all those endeavours were stymied by other folks who were concerned about the influence Atom and the minimal-cost units it was supposed to help would have on Intel’s principal business. MIDs and later on netbooks were supposed to be bare-bones, minimal-cost units, practical as secondary equipment and for basic jobs, but no more.


Intel was doing the job on Atom SoCs in 2008, but its 1st completely unified chip would not ship till 2012.

Intel’s submit-start frame of mind in direction of Atom is greatest summarized as benign neglect. Although the chip went by way of quite a few revisions to combine parts and lower expenses, Intel refused to commit the means that would have manufactured Atom a greatest-in-class player in the cell current market. From 2008 to 2013, Intel released a cost-lessened model of its Nehalem architecture, the Westmere 32nm die shrink, a new architecture with integrated graphics (Sandy Bridge), a high-end fanatic system (Sandy Bridge-E), a new 22nm CPU with FinFET technologies (Ivy Bridge), yet another architectural refresh (Haswell), and a second-era fanatic system (Ivy Bridge-E). That is two whole tick-tock cadences for Intel’s big-core business, while Atom did not even make the jump to 32nm till 2012. Its single architectural refresh to date arrived in 2013, just following the start of Ivy Bridge-E.

In spite of currently being initially starved for means, 32nm Atom chips were competitive in the midrange cell current market. With Medfield, Intel seemed to have turned a corner, but the company’s styles typically unsuccessful to locate a lot traction in the current market. Only Intel’s contra-profits method gained the firm significant tablet current market share, and all those gains were only sustained by way of weighty economical losses.


Intel’s cell and communications profits and losses from Q1 2013 by way of Q2 2014.

Atom wasn’t the problem — Atom was the resolution Intel did not have the guts to chase.

The Tricky Selections Intel Did not Make

Intel unsuccessful to acquire traction in cell mainly because it wasn’t eager to threat upsetting the financial design that experienced reworked it into a titan of computing. The company’s fabs, production approaches, and means were geared in direction of large, expensive processors, not churning out large quantities of minimal-cost cell cores. Prioritizing Atom above Core would’ve essential the firm to retool at the very least some of its fabs to emphasize throughput and decrease expenses in order to contend with the ARM processors designed at Samsung and TSMC. It would’ve intended decrease gross margins and fewer financial gain for each unit marketed.

Intel's Atom Roadmap

Intel tried out to pace matters up ultimately, but both equally its 14nm method and the subsequent-era Goldmont CPU core were delayed.

Intel did take methods to strengthen its competitive standing vis-a-vis ARM and ARM’s foundry partners, but it seldom took them promptly and normally unsuccessful to observe by way of. Intel purchased Infineon Wireless in 2011 for $1.4 billion, but to this working day all of its publicly declared wireless goods, like the XMM 7480 modem, are however designed on 28nm at TSMC. Smartphones and tablets have always used SoCs, but Intel did not start its 1st Atom-based mostly SoC till 2012 — 5 several years following the Iphone released and four several years following Atom’s personal debut.

SoFIA integrated Atom SoC, details

The SoFIA partnership with TSMC raised eyebrows, but not profits.

One matter we want to worry listed here is that Intel’s choice to defend its core (Core) business and products margins could have been mistaken, but it wasn’t insane. Refitting fabs, constructing knowledge in SoC style and design, and porting modems from TSMC would have essential large dollars infusions and take significant amounts of time. If Intel experienced released Atom with an aggressive plan to set the chip in smartphones by 2010, matters may possibly have played out pretty in different ways. By the time the firm woke up to the danger it confronted from ARM and merchant foundries, it was also late to make up the gap.

Why Intel’s Foundry Tech Couldn’t Conserve Its Mobile Enterprise

Intel’s method technologies management couldn’t preserve the company’s cell division mainly because it wasn’t made to do so. Smartphone and tablet OEMs preferred units with integrated LTE radios Intel did not have them. Even the SoFIA partnership with TSMC by no means came to current market, seemingly mainly because Intel couldn’t secure more than enough quantity to kickstart generation.

Intel’s 14nm challenges delayed its subsequent-era tablet processors from 2014 to 2015. Its 10nm node, when predicted to secure huge economies of scale above TSMC, has been pushed to 2017 as nicely. I never consider these delays played a large part in Intel’s choice to leave the cell current market, but they could have influenced it. In 2012, Intel however predicted to be on 10nm by 2016 with EUV (extreme ultraviolet lithography) ramping in direction of whole generation. The now-canceled 450mm wafers weren’t predicted in-current market pretty this quickly, but Chipzilla predicted to recognize significant cost discounts from transferring to the greater wafers in the 2018 to 2020 timeframe — cost discounts that could’ve even more improved its standing towards Samsung, TSMC, and GlobalFoundries.

If Intel experienced started reorienting in direction of Atom when it released the chip in 2008, it might’ve weathered these delays and cancellations without having a lot hassle. Failing to do so remaining its beleaguered cell business struggling with bigger-than-predicted expenses and small profits.

Intel did not reduce the cell current market mainly because Atom’s efficiency and power usage did not contend with ARM research and analysis confirmed that Atom was able of matching ARM efficiency in several current market segments. It lost the cell current market mainly because it did not make the modifications that would have permitted it to contend on cost with goods produced at TSMC and Samsung. The exacting rules and distinctive layouts that drove Intel to the prime of a single current market could not be effortlessly adapted to other folks, and Intel was unwilling to threat its situation at the prime of the standard x86 current market for a risky payoff in cell. There’s no proof that retaining XScale or acquiring ARM goods would have changed that — if nearly anything, the ARM division would’ve been under even more tension to ensure it by no means turned a danger to the x86 business.

Now read:

Leave a Reply

Your email address will not be published. Required fields are marked *