The question – actually, it was more of a plea – was to the point: How are fabricators handling first article inspection requirements?
Confusion, the writer pointed out, appeared to be growing throughout the supply chain, as it attempts to sort out just how often FAI needs to be performed.
It’s a fair question. Inspection is time-consuming (and thus expensive) and considered in many quarters to be non-value-added. And that’s not just for manufacturers but also customers, which face both additional evaluations and a data avalanche. That avalanche is the result of the surge in acceptable quality limit (AQL) samples of FAI measurements as supplied to the customer, which then must enter, sort and distribute (as needed) all that data. It is one thing to impose requirements on suppliers, but in the case of AS9100D, those requirements might be backflowing.
It took until the second business day of the new year for the chips to start falling in the US printed circuit laminate industry. On the same day, Isola changed hands, and Park Electrochemical announced it was reviewing options for its PCB unit – a move generally seen as a precursor to a sale.
As the East Coast braced for a winter blizzard of epic proportions, Park Electrochemical sent a cold shiver down the spines of more than a few industry observers with its announcement of a “strategic evaluation” of its core printed circuit materials business, one that could spell the end of one of the last domestic manufacturers of FR-4 in North America.
Park has been paring its PCB operations over the past few years amid declining revenues and tighter margins. Even as the firm’s aerospace revenues have grown, overall Park sales have fallen year-over-year in 10 of the past 11 quarters, more than half the time by double digits. And said PCB revenues have been falling despite a rebound in the overall bare board market.
Mount Pleasant is a village in Racine County, Wisconsin. It is located roughly 30 miles south of Milwaukee and 60 miles north of Chicago.
Until earlier this year, Mount Pleasant was perhaps best known for once being the home to the late Henry Herzog, who once owned a farm there. For those unfamiliar with Herzog, he was a Wisconsin State Assemblyman.
In 1915 and 1916.
Mount Pleasant is that kind of place.
Sorry, Mount Pleasant was that kind of place. As readers of this publication know by now, Foxconn plans to start construction on a $10 billion, 20 million sq. ft. LCD manufacturing plant in Mount Pleasant this year. And Racine County just gave its blessing to a deal to offer up to $764 million in financial incentives to support that campus.
As of the 2010 census, there were 26,197 residents of Mount Pleasant. Its population makes up about 13% of the 195,408 citizens in Racine County. That $764 million amounts to $3,910 of new debt for every citizen of Racine County. Worse, it represents $29,163 of new debt for every citizen of Mount Pleasant.
Foxconn says it will employ 3,000 people when the site opens, and if things break right, it could ultimately employ 13,000 at the site. That comes to $58,769 in new debt per job created, best case. Assuming 3,000 new jobs, it’s an outlay of $254,666 per job.
The impact is so blunt, Moody’s Investor Service immediately cut Racine County’s bond rating, meaning it will be more expensive for the county to borrow for future needs.
Wow.
As you are letting that sink in, consider the potentially even bigger wow around the corner. Andrew Ng, once Google’s top artificial intelligence researcher, just launched a new venture to bring AI to the factory floor. And who is his partner? Foxconn.
It’s not immediately clear Ng understands how the typical electronics manufacturing plant works. A company video (at landing.ai) demonstrates a human hand-inspecting a semiconductor package. On the video, Ng says the software can easily be taught to spot defects on PCBs, which he says is often done by “thousands of people at a single factory.” Except it’s not. Inline AOI overcame that issue years ago.
There’s an obvious disconnect between the technology and the need that must be sorted out.
What is true, however, is that work toward not just the smart factory (also known as Industry 4.0), but one dominated by AI, will accelerate. The value proposition is too great to ignore. And that’s where matters become, well, uncomfortable.
Ng says that while AI technology is likely to disrupt traditional human-run operations, his new company is seeking ways to prepare displaced employees for higher-skilled work. If he succeeds, that would be great. I wouldn’t bet on it. In a perverse way, capital flows to those who appear best-positioned to blow up old industries, which typically results in lower labor costs, or greater automation, or both. As a startup, Ng’s investors will drive his research.
As Michael Ford notes in our cover story this month, Industry 4.0 married with AI could spell the end of employment, not just for line operators but also higher-level workers such as supervisors, managers and engineers. “The question,” he asks, “is whether this in fact will be necessary, or whether there are ways in which it could be avoided, or at least reduced.”
During its recruitment of Foxconn, Wisconsin state transportation officials were surprised by the ODM’s request for driverless lanes from the new campus to key regional transportation hubs. If the Landing-Foxconn partnership succeeds, those driverless cars might be human-free as well.
We’ve come a long way from the early 2000s, when China pushed factories to buy semiautomatic equipment. The reasoning, as old friend Gene Weiner pointed out to me at the time, was simple: jobs. When you have 1.3 billion people to keep happy, the benefits of automation aren’t necessarily prioritized. Now we are spending hundreds of millions of dollars to underwrite corporations that might accelerate the obsolescence of humans in building electronics. In doing business with Foxconn, the citizens of Mount Pleasant might find it anything but.
Mike Buetow Editor-In-Chief
P.S. A couple of housekeeping notes. In our December issue, a column from Roy Akber borrowed heavily from previous work by Craig Armenti. Our sincere apologies to Craig for the unattributed use of his ideas.
And we continue to add interviews to our podcast series, PCB Chat. Stop by upmg.podbean.com to hear the latest, including my chat with Randall Sherman on the state of the EMS/ODM market.
Since man first walked, he has been scouring the skies for understanding of how he came to be. Big events are bound to enhance our learning, but capturing those moments and making sense of them is never a simple task. In August came one of those big moments astronomers had been waiting a lifetime to witness: the collision of two neutron stars.
While the cosmic event answered questions astrophysicists had long been contemplating, it also opened a new world of questions. The answers to at least some of those questions are buried in mountains of digital data that could take years to sort through.
A similar situation is unfolding in factories worldwide. We no longer look at machines in isolation, nor do the machines themselves act independently. Systems designed to check the work of other machines on the line are proliferating. Those machines generate immeasurable amounts of data, some of which are used to independently resolve ongoing or potential processing “events,” big and small. But the capture of all these data threatens to bury already overworked engineers.
Read more: Are Data Protocol Answers a Wish on Stars?
The question – actually, it was more of a plea – was to the point: How are fabricators handling first article inspection requirements?
