Industries / Technology / Software and SaaS
Cloud and data centers
Cloud and data centers is the work of building and running the buildings, power and cooling systems, servers and networks that host computing, for a company's own use or for rent.
1958
On 1 July 1958 the first AN/FSQ-7 computer of SAGE, the US air defense network, was declared operational at McGuire Air Force Base, New Jersey, housed in a four-story direction center with a 3,000 kW power supply.
44
Milestones achieved across 7 domains.
14
To watch for: announced, not yet confirmed outside.
14
Open deliverables. Nobody has shipped them.
In one minute
Where it started
From 1958 the SAGE air defense network put each of its IBM computers in a windowless concrete building of its own, with a 3,000 kW power supply and a second machine on standby. IBM's large computers were cooled by water from the late 1960s until lower-power chips let air take over in the mid-1990s. In 1999 Google ran its search engine on racks of cheap PC parts, and in 2011 Facebook began publishing its server and data center designs for anyone to use.
Where it stands
Lawrence Berkeley National Laboratory puts US data centers at 176 TWh in 2023, 4.4% of US electricity, and projects 6.7% to 12% by 2028; in Ireland they took 23% of metered electricity in 2025, by the Central Statistics Office's count. AI has moved the work toward liquid cooling: NVIDIA puts the cooling one of its 72-GPU racks needs at 120 kW. Epoch AI, an independent group, estimates that the largest AI site it tracks draws 1,037 MW of IT power, and Uptime Institute finds that power failures, mainly in UPS units, transfer switches and generators, are still the leading cause of outages with an impact.
What is open
Much of the industry's evidence is private or typed in by hand: rack positions entered manually, water use rarely published site by site, coupling test reports given only to customers who ask. The open jobs are public tests and public records: cold plates and couplings tested side by side, second-hand memory and processors screened for errors, the 2024 Virginia load drop replayed in open models, and a ledger of promised data centers set against the dates they open.
Domains
AchievedWatch forOpen
Domain
Progress
Last achieved
Watch for
Open
Sites and construction
This is the work of choosing where data centers go, getting permission to build them and putting up the buildings.
5 · 2 · 2
Last achievedNo more building by right2025
Watch forServer halls in tents
2
openPower
This is the work of getting electricity into the building and on to the servers without a break: the grid connection, the batteries and generators that cover an outage, and the wiring inside the building and rack.
5 · 3 · 2
Last achievedIreland: bring your own power2025
Watch for800-volt DC to the rack
2
openCooling
This is the work of carrying heat away from the chips and out of the building, by air, water or other liquids, so that the machines stay within safe temperatures.
7 · 3 · 2
Last achievedAI racks need liquid2024
Watch forCoolant inside the chip
2
openServers and networking
This is the work of choosing, building and connecting the computers in a data center: the servers, the racks that hold them and the network that links them.
6 · 3 · 2
Last achievedMirrors replace patch panels2022
Watch forOld memory in new servers
2
openOperations and reliability
This is the daily work of keeping a data center running: maintenance and repair, watching for failing parts, and learning from outages.
7 · 1 · 2
Last achievedGPUs fail in big runs2024
Watch forRobots for data hall chores
2
openPlanning capacity
This is the work of deciding how much computing capacity to build, and getting the most use out of what is built.
7 · 1 · 2
Last achievedMissed dates counted2026
Watch forRacks sized to average power
2
openEnergy and water use
This is the work of measuring how much electricity and water data centers use, reporting it, and cutting it.
7 · 1 · 2
Last achievedAI ends the plateau2024
Watch forAn EU data center label
2
open
Narrower industries
Advances to build with
1 research note
Open in cloud and data centers
3 of 14 open deliverables
Open · Sites and construction
Measured noise at named sites
The first standards-grade public recordings of the sound around data centers. Log for a week at the property line and at the nearest homes of named facilities with calibrated Class 1 meters, recording A- and C-weighted levels, one-third-octave bands and narrowband spectra, the method that finds the low hum residents complain about. Pair each site with its cooling type.
Built with what exists
Class 1 sound level meters, such as an NTi Audio XL2 with a Class 1 microphone or a Brüel & Kjær 2250, rented or borrowed from a university; the narrowband method in ISO 1996-2; Python with NumPy and SciPy for the spectra; the measurement protocol the review proposes.
Open · Power
Open flexibility test for GPUs
A public test suite that sends grid-style requests ("cut to 70% in one second and hold for an hour") to AI training and inference jobs on rented GPU servers and records how fast power falls, how long the cut holds, how much work is lost and how the load comes back. The requests follow real program rules, such as the Texas emergency curtailment above. Results go in one public table, by GPU type and workload.
Built with what exists
Rented cloud GPU servers of two or three generations; NVIDIA's NVML library and nvidia-smi tool to set power limits and read power each second; vLLM for inference and PyTorch for training, running open-weight models; the event shapes in the published papers and grid programs.
Open · Cooling
Cold plates on one rig
A public test of at least six commercial direct-to-chip cold plates, the metal blocks with internal channels that clamp onto a chip and carry coolant over it, all run on one thermal test vehicle (a heater block the size of a processor) under the Open Compute Project's new cold plate test specification, publishing thermal resistance and pressure-drop curves for each.
Built with what exists
A resistive heater block with a copper top, built after the standard test vehicle of Martinez, Caceres and Ortega and kept under 500 W; a benchtop recirculating bath holding water at 45°C, the specification's typical test temperature; a turbine or Coriolis flow meter; a differential pressure sensor; type-T thermocouples; a data logger such as a LabJack; Python for the curve fits.
Where it sits
Broader
