Guess work and structural delays are a massive, multi-million-dollar problem in construction. With data centers in particular, compressed construction timelines, mission-critical/zero tolerance for structural failure, growing pressure to reduce embodied carbon, and community pushback for long-term projects, are all major stress points for developers and operators.

Schedule delays can cost data center operators between $100,000/week to more than $1 million/per week in idle labor, extended equipment rentals, and construction carrying costs. 

Today’s story, What is ‘smart concrete’ and why could it matter for data centers?“ gives a high-level view of how AI-optimized mixes and embedded IoT maturity sensors can take some of the guesswork out of data center construction, not only for hyperscalers, but also colocation providers and wholesale data center developers. Embedded sensors from companies like Giatec’s SmartRock are offering real-time data about the temperature, maturity, and strength inside concrete slabs – wirelessly transmitting information to the cloud to help construction teams accurately, and quickly, verify when a concrete slab is ready for heavy loading. 

Because concrete generates internal heat and cures at different rates, there traditionally were two major options for ensuring structural integrity: guess and wait.

Shoveling fresh concrete into plastic, cylindrical test cylinders to get crushed by a hydraulic press for break point measurements over different intervals of time, or shipping cylinders off to a labs to cure and get tested, are not only arduous, but extremely time consuming and expensive.

With “smart concrete,” data center operators are radically improving curing and strength tracking to ensure data center floor slabs will support massive AI server loads. Some examples include:

• Microsoft, which uses Azure IoT Hub to process real-time data collected by Giatec’s concrete sensors and AI engine for structural integrity and mix transit optimization;

• Meta, which has shifted heavily into “green concrete” to meet net-zero targets, even open-sourcing its own AI concrete optimization tools—like BOxCrete—to develop mixes that cure up to 43% faster while reducing embodied carbon by up to 40% at data center sites;

• Google, which is actively part of collaborative clean-tech efforts (like the Elemental Impact Data Center Innovation Initiative and OCP Green Concrete) to test and scale low-carbon concrete and smart monitoring tech across data center buildouts.

For a high-level view of smart concrete, read today’s top story, “What is ‘smart concrete’ and why could it matter for data centers?“

Also check out our deeper dive into the Nvidia RTX Spark, revealed at this week’sComputex event in Taipei.

‘Smart concrete’ in DC construction:  AI data centers’ zero tolerance for structural failures are a prime use case for smart concrete solutions that embed sensors and digital monitors for faster and accurate measures of strength, reliability, resilience.

Nvida ‘AI superchip’: Will the Nvidia RTX Spark shift AI from the cloud to the device? Native CUDA on Arm plus TensorRT, cuDNN, and Triton create a desk-to-laptop continuum for developers, deepening Nvidia’s lock-in across the stack.

O’Leary addresses backlash: Celebrity investor Kevin O’Leary says he’s “not walking away,” despite Utah Senate President J. Stuart Adams; demands that he cut his proposed 40,000 acre hyperscale data center back by 75%.

FERC weighing in on PJM: FERC is considering different ways to help PJM Interconnection keep up with growing electricity supply constraints amid data center builds and rapidly increasing consumer energy bills across the 13 states it serves.

Phoenix PSC proposes DC rate hike: In Phoenix, AZ, the PSC has proposed a 45% electricity rate increase on data centers, an attempt to counter APS’ 16% rate hikes for customers, and higher grid access charge to rooftop solar customers. 

AirTrunk $21B DC in India: Robin Khuda’s AirTrunk has signed a Letter of Intent with the Government of Maharashtra, India, for a $21 billion data center in Raigad, near Mumbai. Backed by Blackstone, the data center plans include 3GW capacity.

Asprofin $12B proposal in UAE: Asprofin Bank has unveiled a $12 billion proposal for a large-scale data center program in the United Arab Emirates, including three hyperscale campuses with capacity of 750MW to 1.28GW over 10 to 15 years.

Advanced packaging matters:  Samsung says that when it comes to semiconductors, advanced packaging and system co-optimization matter: shortening interconnect distances and bringing compute and memory closer.