Inside the Big-Money Business of Protecting Data Centers

Dow Jones
昨天

After Iranian drones struck three Amazon.com data centers in the Middle East this March, Austin Storms, the co-head of data centers at Galaxy Digital, got a call from the C-suite. Erin Brown, Galaxy's chief operating officer, wanted to know what the company was doing to prepare for similar attacks.

Galaxy is developing a 2,200-acre artificial-intelligence campus in Dickens County, Texas -- not exactly within Iran's range. But with trillions of dollars at stake in today's AI race, people like Storms and Brown can't leave anything to chance.

Developers are spending big on physical security as data centers become larger and more valuable, turning once-sleepy facilities into fortresses. The investments go beyond barbed wire and part-time guards. Developers are using drones and facial-recognition technology, coordinating with local law enforcement, and hiring veterans and former police officers.

That trend has created a new market for security companies more accustomed to selling to police departments or the Pentagon. With a construction rush under way and the backlash against AI potentially keeping some new projects from getting approved, these vendors' window of opportunity may be short -- so too investors' chances to profit.

Security Costs

Galaxy expects to spend eight figures in annual security costs once its Texas campus is fully constructed. Data-center developer Applied Digital, which has three facilities in development, likewise allocates $6 million to $10 million a year to security at a typical two-building site. That is just a fraction of total operating costs, but bear in mind there are 1,318 data centers operating in the U.S. and 2,299 planned projects, according to market researcher Cleanview.

Among those, the newest AI facilities present a different security challenge than the storage or web-hosting data centers dotted across the country. To start, they can be enormous, making foot patrols less feasible than in a more contained environment. Galaxy pointed to its facial-recognition system, which identifies and logs every figure that crosses a camera at the site. Applied Digital, meanwhile, uses an array of cameras and drones at its facilities; it even ran a trial with -- but hasn't deployed -- robot patrol dogs.

The value of the data centers' contents is an even more important reason for all the fancy security measures. The chips and servers inside the facilities, and the AI model running on them, are worth enormous sums of money. In discussions with potential customers, security has become "table stakes," says Applied Digital CEO Wes Cummins.

While the drone strikes against Amazon made clear that data centers have become "critical infrastructure," to use security jargon, widespread opposition to AI data centers has brought developers' worst fears closer to home.

The pushback against data centers has remained peaceful in nearly all cases, and no one Barron's spoke to suggested some kind of violent incident is imminent. But developers are still watching online chatter for threats or signs of trouble.

"We take everything very seriously," says Cummins. "The issue is if you have something happen, it can have a very large impact."

A Utah data center being developed by QTS was evacuated in April after a bomb threat. Earlier that month, a gunman shot 13 bullets into the house of Indianapolis councilman Ron Gibson, who had expressed support for a proposed data center in the city.

Robots Don't Need Breaks

Ondas, which has contracts with the U.S. and Israeli militaries, is pitching its drone surveillance and counterdrone technologies to commercial customers.

"Robots work 24 hours a day, they're always on demand, they don't call in sick, they don't get hurt," explains Ondas CEO Eric Brock.

A full counterdrone system, which can identify and sometimes neutralize unwanted drones flying overhead, could cost anywhere from the low six figures to "multiples of that," Brock adds. Ondas doesn't have contracts with the largest data-center companies yet, but it has struck security deals with semiconductor manufacturers.

Birdstop, which makes drones equipped with thermal infrared cameras and laser range-finders, started to count data centers as part of its core business this year, says CEO Keith Miao. The venture-backed start-up exclusively serves nonmilitary sites like power plants and telecom infrastructure, so securing AI sites was a natural fit.

Axon Enterprise is jumping on the opportunity, too, pitching its Dedrone airspace-security solution to data centers and other private-sector facilities, such as stadiums, corporate campuses, and even CEO's houses. The company is best known for making Tasers, body cameras, and license-plate recognition monitors for police.

"Every data center understands, especially with now some of the rising controversy around data centers, that any disgruntled person with a 3-D printer, access to a drone, and the internet could build something that could cause real problems," Axon CEO Rick Smith said on an earnings call in August.

Security systems, much like hard drives or copper wire, are cheap relative to the chips and servers they enable. What seems like a small amount of money to tech companies can therefore be substantial for some suppliers. Data-center spending isn't even close to peaking, so the argument goes, and each new construction project that breaks ground could mean millions of dollars in security money.

For the first time in nearly a decade, ransomware -- malicious software that locks a victim's data or device and demands a payment to restore access -- isn't the biggest fear for data-center professionals. In an annual survey released earlier this year by AFCOM, a professional association for IT workers, 61% of respondents cited human threats as one of their top five security threats. That is up from just 40% in 2024.

 

應版權方要求,你需要登入查看該內容

免責聲明:投資有風險,本文並非投資建議,以上內容不應被視為任何金融產品的購買或出售要約、建議或邀請,作者或其他用戶的任何相關討論、評論或帖子也不應被視為此類內容。本文僅供一般參考,不考慮您的個人投資目標、財務狀況或需求。TTM對信息的準確性和完整性不承擔任何責任或保證,投資者應自行研究並在投資前尋求專業建議。

熱議股票

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10