Getty Images

North America added a whole Silicon Valley’s worth of data center inventory this year. It’s not enough.

Four-year delays aren’t dampening demand.

8/20/24 12:26PM

North America’s eight primary data center markets added 515 megawatts (MW) of new supply in the first half of 2024 — the equivalent of Silicon Valley’s entire existing inventory — according to a new report real-estate services firm CBRE.

All of Silicon Valley has 459 MW of data center supply, while those main markets have a total of 5,689 MW. That’s up 10% from a year ago and about double what it was five years ago.

Data center space under construction is up nearly 70% from a year ago and is currently at a record high. But the vast majority of that is already leased, and vacancy rates have shrunk to a record low of 2.8%. In other words, developers are building an insane amount of data center capacity, but it’s still not enough to meet the growing demands of cloud computing and artificial intelligence providers.

A shortage of available power and necessary equipment, like transformers, switches and generators, is contributing to years-long delays, but that hasn’t dampened demand, as companies secure future data center capacity anyway.

“We’re signing leases that some of these clients won’t occupy for three or four years,” Pat Lynch, executive managing director and global head of CBRE Data Center Solutions, told Sherwood. Additionally, Lynch said enterprises are renewing existing data center leases even if they’d prefer newer data centers that can better handle their increasingly demanding workloads.

“They have no other choice,” he said. “It just shows that their capacity need is not going anywhere, and they just want to get in on it.”

That demand has sent national rental rates up 6.5% on average and much higher for newer premium spaces in premium markets. The imbalance makes it a data center landlord/owner-operator’s market, which Lynch expects to continue for the next few years.

Rani Molla1h

Google employees are now competing with Anthropic and Meta for access to Google compute

Google built its reputation as a paradise for ambitious researchers: a place where smart people got massive resources and freedom to experiment.

But in the AI era, the physical infrastructure that powers those breakthroughs is maxed out, and even Google’s own employees are reportedly struggling to get enough computing power.

According to Bloomberg, the bottleneck comes down to hardware. Google’s custom-built AI chips — tensor processing units, or TPUs — are in such high demand that internal researchers say they’re effectively competing for rack space against massive, paying cloud customers like Anthropic and Meta. Frustrated by the bureaucracy of fighting for server time, top engineers are jumping ship to launch their own startups, arguing they can secure more reliable access to infrastructure on the open market than inside the company that actually builds it.

In other words: Google became so successful at selling AI infrastructure that its own researchers now have to justify experimental projects against revenue-generating workloads and a more than $460 billion backlog of paying tenants.

Google’s Own AI Researchers Jockey for Access to Its Computing

$420

Rani Molla2h

Elon Musk once promised to take Tesla private at $420. More recently, he’s been offering xAI employees $420 to hand over their private tax returns as training data for Grok, Bloomberg reports, citing internal chats. In an effort to boost the chatbot’s tax-prep capabilities, the company asked employees — as well as friends and family — to submit completed tax returns in exchange for cash that, two months later, still hasn’t materialized. xAI is owned by the soon-to-be-public SpaceX.

Nate Becker3h

EY retracts report with apparent AI hallucinations

Consulting firm EY has retracted a report on travel loyalty points that an AI watchdog had found was full of hallucinations.

AI-detection firm GPTZero alleged that the report was “riddled with hallucinations,” including citing numerous sources that didn’t appear to exist. Sherwood News exclusively reported on GPTZero’s findings about the report on Thursday. EY didn’t respond to multiple requests for comment.

The firm later told the Financial Times that it had retracted the report, saying it was “reviewing the circumstances that led to this article’s publication.” It said the study wasn’t connected to work for any of its clients.

“EY Canada takes the accuracy of all the content we publish seriously and we have an organization-wide commitment to the responsible use of AI,” EY said, according to the FT.

A link to the report on EY’s site now displays an error: “Oops! We couldn’t find the page you were looking for.”

AI hallucinations appear to be creeping into consulting reports

“EY Canada takes the accuracy of all the content we publish seriously and we have an organization-wide commitment to the responsible use of AI,” EY said, according to the FT.

A link to the report on EY’s site now displays an error: “Oops! We couldn’t find the page you were looking for.”

Rani Molla3h

Elon Musk expects Tesla Robotaxis to be “widespread” in the US by the end of the year

After an uncharacteristically clear-eyed earnings call where Elon Musk was cautious about the timing of the company’s many ambitious goals, the Tesla CEO is back to making his usual unlikely predictions:

“We already have some vehicles operating with no people inside and no safety monitors in three cities in Texas, and it probably will be widespread in the US by the end of this year,” Musk said by video at the Smart Mobility Summit in Tel Aviv on Monday. It’s a prediction Musk has made before, but that doesn’t mean it’s going to happen.

Tesla’s expansion of its Robotaxi service, which launched nearly a year ago, has been painstakingly slow. The vast majority of the Robotaxis — more than 500 in the Bay Area — have a person behind the wheel using a version of Supervised Full Self-Driving. In Austin, 12 of the 40 Robotaxis have been spotted driving unsupervised in the last week, according to Robotaxi Tracker. There are two more each in Dallas and Houston. Alphabet’s Waymo, by comparison, is already operating more than 3,000 of its driverless vehicles in cities across the country.

“Initially, we’re taking a very cautious approach to the rollout here,” Musk had said on the last earnings call, estimating the service would be in a dozen states by the end of the year. Today he was more bullish, estimating that in 5 or 10 years, “90% of all distance driven will be driven by the AI in a self-driving car.”

Here is Elon Musk's full interview tonight from the Samson International Smart Mobility Summit in Israel.

Elon talks about Tesla's Unsupervised robotaxi rollout, Starship V3, Neuralink, and more.

He works nonstop. Elon started this live interview at 2 AM in Texas lol. pic.twitter.com/yKmqX5ueBP
— Sawyer Merritt (@SawyerMerritt) May 18, 2026

Anthropic really doesn’t want the US to help China with AI

Anthropic made its case for freezing China out of the AI race as much as possible in a new policy paper. The company warned that letting China catch up to US AI companies could risk AI-powered mass surveillance and huge risks to monitoring AI safety.

Jon Keegan5/15/26