My local grocery store in rural Virginia now has a blinking, beeping robot that rolls up and down the aisles, scanning inventory to let someone in charge know when they’re low on stock.
Sentara Health, the Norfolk-based chain that opens hospitals across Virginia and North Carolina, now uses artificial intelligence programs to monitor patients and discharge timelines to make sure whatever services a patient needs after discharge (such as rehab or at-home care) is lined up.
My email provider now automatically provides an “AI inbox” to summarize email. I didn’t ask for it. I don’t want it. I’m quite capable of reading my own emails, but there it is anyway.
The connection for all these three otherwise unrelated events runs through “the cloud,” which is not really a cloud but a collection of data centers somewhere.
One of the unexpected developments in this year’s congressional campaigns has been the emergence of data centers as an issue — with Democrats being the ones most likely to push the issue. Tom Perriello, the Democratic candidate in the 5th District, opened his campaign with an ad in which he appears to smash a hammer against a wall that says, among other things, “data centers.” Beth Macy, the Democratic candidate in the 6th District, has made a data center moratorium a central part of her campaign. So has Joy Powers, the Democratic candidate in the 9th District.
Across Virginia, we see some localities already taking their own actions to restrict data centers. Suffolk and the town of Front Royal have enacted outright bans. Chesapeake and Virginia Beach have passed temporary moratoriums. At least six other localities have taken different actions to make it more difficult to build a data center.
Whether the opposition to data centers is well-founded is a matter of political opinion, but there’s no doubt that the opposition is quite real. Now, here’s the problem: Any restrictions on data centers are treating the symptom, not the problem. The problem is that we keep creating data that requires data centers.
A state moratorium, as some want, would simply send data centers elsewhere. That might solve our problems with increased energy and water demands, but in practice, it only transfers it elsewhere — most likely Texas. A national moratorium, as some want, would simply send that data overseas, which raises the question of whether you’re comfortable with your financial records or medical records sitting in some other country. I explored those questions in a previous column. Plot twist: Those records might already be overseas anyway.
For all the political clamor over data centers, there has been surprisingly little political push to ban, regulate or otherwise restrict the data-creation forces fueling them. To do so would require legislators to reach further into the business world than even the most regulation-happy politician has ever proposed.
There are some efforts underway. At the federal level, both Sen. Mark Warner, D-Virginia, and Rep. Eugene Vindman, D-Prince William County, have proposed to ban “surveillance pricing,” the practice of using computer data to charge some customers more than others. (That makes it a more personalized cousin of dynamic pricing, which relies on general market signals to change prices, such as hotels raising prices when there’s a big college football game in town). At the state level, state Sen. Russet Perry, D-Loudoun County, has also introduced a bill to ban surveillance pricing. As a practical matter, hers probably stands a better chance because almost every bill in Richmond actually gets acted on in some way, while bills in Congress are often little more than press releases until party leaders decide to do something with them.
Here’s how a statement from Perry describes the practice: “Companies now collect detailed information on their customers, including shopping history, browsing activity, location, and the device they shop on. Increasingly, businesses use that data to estimate how much each person is willing to pay and set prices accordingly. The practice, known as surveillance pricing, means a customer can be charged more simply because a company’s algorithm predicts they can afford it or won’t shop around.”
However, the bill doesn’t propose to ban companies from collecting and storing that data — just using it or combining it with other data “to raise or change the underlying price of a product,” Perry’s statement says.
The reality is that so much of our economy — and our personal lives — is digitized that little can be done to stop the demand for more data centers. All of our online banking, our direct deposits of paychecks, our online shopping, our social media activity — all of it requires data centers, and every day we create more data that has to be stored somewhere. In a previous column, I used estimates from one data center company to calculate that we each make use of anywhere from 58 to 87 different data centers every day. That’s because data is typically not stored in a single site; there are copies stored in multiple locations for redundancy.
Fun fact: A few weeks ago, the company that Cardinal contracts with for our web hosting sent out a notice that its data center in the Dallas-Fort Worth area had cooling problems and was overheating, but not to worry, because “we maintain a real-time copy of each site across multiple data centers for redundancy” and operations were being temporarily shifted to those other locations. This did not appear to affect us; the vendor, which works with lots of online news sites, said it keeps “the main version of your site in the region closest to you,” with copies elsewhere. That means you’re probably reading this from a server somewhere in Ashburn in Loudoun County. However, the company has 28 data centers around the world — North America, South America, Europe, Africa and Asia. As I write these words, the vendor’s site is showing its busiest data center to be in São Paulo, Brazil.
The arrival of AI promises (or threatens) to send our data demands into hyperdrive — because AI requires more energy than normal types of data. Estimates vary, but the website Programs calculates that 23% of “data center workloads” around the world are now for AI-related requests. By the end of the decade, that figure is expected to rise to 50% to 70%, Programs says. AI isn’t going to crowd out all that other data; it’s going to require more data centers, be they here or somewhere else. The McKinsey consulting company estimated earlier this year that by 2030 the power demands for data centers would be 220% more than they were just three years ago.
Logically, if we feel we as a society can’t or don’t want to deal with those kinds of energy requirements, we shouldn’t just restrict data centers — we should restrict the demands driving that kind of computing power. Is that already too late?
The U.S. Census Bureau reported earlier this year that between 17% and 20% of American businesses were already using artificial intelligence in some fashion, and that figure was likely to rise to between 20% and 23% in six months (which, based on when the original report was published, means, oh, about now). The larger the company, the more likely it was to be using AI. For firms with 250 or more workers, the AI adoption rate was 37%, the Census Bureau said. It’s not just private companies, either; it’s also government. Newport News and Norfolk police now use AI to handle some non-emergency calls. Fairfax County police use it for data analysis. Prince William County uses AI for other types of inquiries. Fairfax County and Roanoke have AI-powered chatbots on their websites to answer questions.
The state budget that Virginia legislators passed this year (the same one that imposes a new tax on the electricity that data centers use) specifically directs some state agencies to explore using AI, including two of the largest — the Department of Medical Assistance Services and the Department of Transportation. It also sets up a new state-run institute to “coordinate with other universities and stakeholders to develop a statewide ecosystem in AI.” It seems hypocritical to want to stop development of data centers while voting for a state budget that promotes a technology that will require more data centers. In the case of Roanoke, two members of city council have objected to the data center that wants to locate in the Wood Haven Technology Park in Roanoke County (in which the city owns a partial interest). That’s a legitimate policy viewpoint, but shouldn’t they also object to the city’s chatbot, which requires a data center somewhere?
Lots of politicians have called on AI to be regulated — Sen. Bernie Sanders, I-Vermont, wants a Department of Artificial Intelligence — but I’m not aware of any who have called for the technology to be banned. If a candidate wants a moratorium on data centers (as Sanders and others do), should they logically be for a moratorium on AI, as well?
Perriello, the Democratic candidate in the 5th District, wrote a commentary earlier this fall in The New Republic in which he talked about how often he hears voters bring up technology issues. “In my first week on the trail, voters asked me about AI at every event we held,” Perriello wrote. “At a town hall in Pittsylvania County this month, one-third of all questions were about AI — over half, if you include questions about data centers.”
The concern about AI comes in two categories — worries about an unregulated technology of unknown capabilities and its capacity to eliminate jobs. In his New Republic article, Perriello warned that AI could be “NAFTA on steroids.” Macy, the 6th District Democrat, has cited studies that show AI could eliminate half of entry-level white-collar jobs. There could be some truth to both claims, although the two Democrats are only citing one side of the ledger. Perriello’s NAFTA reference is interesting, because economists generally agree that the North American Free Trade Agreement probably created as many jobs as it eliminated — the catch is that they weren’t in the same place, or with the same workers.
There’s certainly some evidence that AI is eliminating some jobs. Stanford University released a report this summer that says “employment of young workers in AI-exposed occupations stands 19% below where it would be had it kept pace with that of their less-exposed peers, with no comparable gap for experienced workers,” which suggests that companies are using AI to automate some entry-level jobs.
However, Goldman Sachs reports that the number of jobs created by AI is pretty much the same as those being eliminated — they’re just different types of jobs. I’m thinking of that grocery store robot. That might be eliminating the job of a store clerk, but it’s creating jobs for whoever is making and selling those machines. In this case, those making those robots appear to be in Guadalajara, Mexico, so this robot might be creating sales and administrative jobs in the U.S. but manufacturing jobs south of the border.
By that measure, Perriello could be right — that is akin to NAFTA, in terms of jobs going out of the country, although in this case it’s a different type of job being created in Mexico than the one being done away with here. On the other hand, if that robot saves the store money, perhaps some of those savings are reflected in lower prices — or at least a slower rise of high prices. Shouldn’t I, as a customer, like that?
Meanwhile, Sentara says its use of AI in monitoring patients “could save nearly 90,000 hours — the equivalent of adding more than 40 full-time employees and nearly $5 million in labor costs.” These are real-life trade-offs; is it better to have the company hire those 40-plus employees, or is it better to shave some costs off healthcare? Also keep in mind that the size of the American workforce is going to decline as more baby boomers retire and a smaller generation — the consequence of declining birth rates — occupies the labor pool. There’s a ready solution to that — we could increase immigration — but that’s politically untenable at the moment. Those demographic pressures are only going to accelerate the push for companies to automate as much as they can. Demography isn’t just destiny; it’s also more data centers.
The McKinsey company issued a report this year that probably isn’t getting enough attention. On the plus side, it said that over time AI will create more jobs than it does away with — and that those new jobs will likely pay more than the ones they replace because they will require higher skills. But that’s the catch, McKinsey warned: “The change may require the largest and most sustained workforce transformation in U.S. history. We estimate that roughly 11 million workers in declining occupations may need to transition to new jobs, ranging from six million to 16 million depending on the pace of automation adoption and its impact on labor demand.”
Are we, as a society, prepared for that? Probably not, right? That appears to be where we’re headed, though — and our current focus on a data center moratorium doesn’t come anywhere close to addressing the real factors that are driving the need for more data centers.
We have more politics each Friday afternoon in West of the Capital, our weekly political newsletter. This week I’ll take a last look at the early voting trends and update what the congressional candidates in the 5th, 6th and 9th districts have done this last full week of the campaign. Sign up here:
