Data centers are being rebuilt from the ground up to meet demands that didn't exist five years ago. Here's what's changing and why it matters.
Data centers have existed for decades, but the infrastructure being built today looks almost nothing like what came before. A combination of artificial intelligence workloads, edge computing demands, sustainability pressure, and power constraints is forcing a fundamental rethinking of how data centers are designed, built, and operated.
AI Is Rewriting the Power Equation
The single biggest driver of data center change right now is artificial intelligence. Training and running large AI models requires a density of compute power that traditional data center designs simply weren't built to handle.
A standard server rack in a conventional data center might draw 5 to 10 kilowatts. AI compute racks — packed with GPUs and specialized accelerators — can draw 50 to 100 kilowatts or more. That's not a minor upgrade challenge. It requires completely rethinking power distribution, cooling infrastructure, and physical layout from the ground up.
This is why hyperscalers like Microsoft, Google, Amazon, and Meta are committing to multi-billion dollar data center construction programs. The existing global inventory of data center capacity is largely unsuitable for the power densities AI workloads demand.
Cooling Is the New Bottleneck
When rack densities increase tenfold, air cooling — the standard approach for conventional data centers — stops being viable. Moving enough air to cool a 100kW rack is physically impractical at scale.
The industry is rapidly shifting toward liquid cooling. Direct liquid cooling runs coolant directly to the chip level, removing heat orders of magnitude more efficiently than air. Immersion cooling submerges entire servers in non-conductive liquid, achieving even higher efficiency.
Neither of these approaches is new — both have existed in specialized applications for years — but AI is forcing them into mainstream data center design at scale and speed the industry has never seen before.
The Power Problem
Data centers already consume roughly 1-2% of global electricity. AI is expected to push that significantly higher over the next decade. This creates a collision between data center growth and sustainability commitments that every major technology company has made.
The response is coming on multiple fronts. Nuclear power is seeing renewed interest specifically because it provides reliable, carbon-free baseload power at the scale data centers need. Several major technology companies have signed agreements with nuclear operators or invested directly in small modular reactor development.
Renewable energy procurement is accelerating, though the intermittent nature of wind and solar creates challenges for facilities that must run continuously. Battery storage, hydrogen fuel cells, and grid modernization are all part of the long-term answer.
Edge Computing and Distributed Infrastructure
Not everything is moving toward larger, more centralized facilities. A parallel trend is pushing compute capacity closer to where data is generated and consumed.
Autonomous vehicles, industrial automation, remote healthcare, and real-time financial systems all have latency requirements that centralized cloud infrastructure can't meet. Processing data at the edge — in smaller facilities closer to the end user — reduces the round-trip time between request and response.
This is driving investment in a new category of smaller, modular data centers that can be deployed quickly in locations that traditional large-scale facilities would never consider. Prefabricated, containerized data center modules can be shipped and operational in weeks rather than the years required for traditional construction.
What This Means Long Term
The data center industry is in the middle of a capital investment cycle unlike anything it has seen before. Estimates for global data center construction spending over the next five years run into the trillions of dollars.
The facilities being built today will define the digital infrastructure of the next two decades. The decisions being made now — about power sources, cooling technology, location strategy, and compute architecture — will have consequences that extend far beyond the technology industry into energy grids, real estate markets, water consumption, and regional economic development.
For anyone involved in industrial instrumentation, process control, or facility infrastructure, the data center build-out represents one of the largest concentrations of capital spending in a single sector in modern history. The monitoring, control, and measurement requirements for facilities operating at this scale and power density are substantial — and growing.
Not all sensors speak the same language. Understanding the differences between 4-20mA, 0-10V, and ratiometric outputs helps you select the right instrument for your application and avoid costly compatibility mismatches.
Walk into any industrial facility and you'll find sensors and transmitters outputting signals in several different formats. The three most common are 4-20mA current loops, 0-10V voltage outputs, and ratiometric voltage outputs. Each has its own strengths, limitations, and ideal applications. Knowing the difference saves time during design and prevents headaches during commissioning.
4-20mA Current Loop
The 4-20mA current loop is the standard for industrial process measurement and control. It transmits measurement data as a variable current — 4mA at the low end of the range, 20mA at the full scale.
The primary advantage is immunity to electrical noise and resistance to signal degradation over long cable runs. Because current is the signal carrier rather than voltage, resistance in the cable doesn't affect accuracy. This makes 4-20mA ideal for field instruments located hundreds or thousands of feet from the control room.
The live zero at 4mA is another significant advantage. Because a healthy loop never drops to zero current, a reading of 0mA immediately indicates a fault condition — a broken wire, a failed transmitter, or a loss of power. This makes loop integrity monitoring straightforward.
4-20mA is the right choice for: pressure, temperature, level, and flow transmitters in process industries; any installation with long cable runs; applications where noise immunity is critical.
0-10V Voltage Output
The 0-10V voltage output is common in building automation, HVAC, light industrial applications, and motion control. It transmits measurement data as a variable voltage — 0V at minimum, 10V at full scale. Some variants use 0-5V or 1-5V ranges.
The main advantage of voltage outputs is simplicity. No loop power supply is required — the sensor just needs a DC supply voltage and a signal wire back to the receiver. Many PLCs and building automation controllers have voltage inputs as standard.
The limitation is sensitivity to cable resistance and electrical noise. Over long cable runs, voltage drop across the cable resistance introduces measurement error. Voltage signals are also more susceptible to interference from nearby motors, drives, and power wiring.
0-10V is the right choice for: short cable runs in clean electrical environments; HVAC sensors and building automation; applications where the receiving equipment only accepts voltage inputs; situations where loop power supply is not available.
When you have a 0-10V sensor but need a 4-20mA input — or vice versa — a signal converter like the ISOCON handles the translation cleanly without any programming or complex wiring.
Ratiometric Output
Ratiometric output is less commonly discussed but widely used in pressure sensing, position sensing, and load cell applications. A ratiometric sensor outputs a voltage that is proportional not to a fixed reference, but to its supply voltage.
For example, a ratiometric pressure sensor powered by 5V might output 0.5V at zero pressure and 4.5V at full scale. If the supply voltage changes to 4.9V, the output shifts proportionally — 0.49V at zero and 4.41V at full scale. The ratio between output and supply voltage stays constant regardless of supply variations.
This is actually a significant advantage in certain applications. If your supply voltage drifts slightly, a standard voltage output sensor would show a measurement error. A ratiometric sensor compensates automatically because the receiving device (typically a microcontroller or specialized analog input) measures the ratio of output to supply, not the absolute voltage.
Ratiometric outputs are common in automotive sensors, industrial pressure transducers, and any application where the sensor and receiver share the same power supply and supply stability is a concern.
Ratiometric is the right choice for: pressure sensors interfacing directly with microcontrollers or specialized analog inputs; applications where supply voltage stability is a concern; position and force sensing in embedded systems.
Choosing the Right Output for Your Application
The decision usually comes down to three factors: cable length, electrical environment, and what your receiving equipment accepts.
For long runs in noisy industrial environments, 4-20mA wins every time. For short runs in clean environments with standard controller inputs, 0-10V is simpler. For direct sensor-to-microcontroller interfaces where supply variations are a concern, ratiometric is the purpose-built solution.
When your sensor and your controller don't speak the same signal language, a signal conditioner or converter bridges the gap without requiring you to replace either device.
The 4-20mA current loop is the backbone of industrial process measurement. Here's what you need to know to connect a sensor correctly and avoid the most common wiring mistakes.
The 4-20mA current loop has been the dominant signal standard in industrial process control for decades. It's reliable, noise-resistant over long cable runs, and simple to troubleshoot. But connecting a 4-20mA sensor correctly requires understanding a few fundamentals — get them wrong and you'll spend hours chasing signal problems that should never have happened.
What Is a 4-20mA Loop?
A 4-20mA current loop transmits a measurement signal as a variable current rather than a voltage. The transmitter (your sensor) regulates current flow through the loop in proportion to what it's measuring. At the low end of the measurement range, the loop carries 4mA. At the high end, it carries 20mA. Everything in between is proportional.
The reason current is used instead of voltage is simple: current doesn't drop across long cable runs the way voltage does. A 4-20mA signal from a pressure transmitter 2,000 feet away arrives at your control panel just as accurate as one 10 feet away.
Two-Wire vs. Four-Wire Transmitters
Before wiring anything, identify what type of transmitter you have.
A two-wire transmitter is loop-powered — it draws its operating power directly from the loop itself. The same two wires that carry the signal also power the device. This is the most common configuration for field instruments like pressure, temperature, and level transmitters.
A four-wire transmitter has a separate power supply for the device and uses a dedicated pair of wires for the signal output. These are common on analyzers and more complex instruments that require more power than a loop can provide.
Wiring a Two-Wire Transmitter
For a two-wire loop-powered transmitter, the wiring is straightforward:
The positive terminal of your DC power supply connects to the positive terminal of the transmitter. The negative terminal of the transmitter connects to the positive input of your receiver (PLC analog input, indicator, or recorder). The negative terminal of your receiver connects back to the negative terminal of the power supply, completing the loop.
Loop supply voltage is typically 24VDC. Check your transmitter's datasheet for minimum and maximum loop voltage requirements — most two-wire transmitters operate between 12VDC and 36VDC.
Load Resistance
Every device in the loop adds resistance. Your power supply must provide enough voltage to drive the full 20mA through the total loop resistance. A simple way to check: add up the resistance of all devices in the loop (input resistors on your PLC card, barriers, converters) and multiply by 0.02A. The result must be less than your available loop voltage minus the transmitter's minimum operating voltage.
If you're running tight on loop voltage, a signal isolator or loop-powered repeater can help — and also provides the benefit of electrical isolation between the field device and your control system.
Grounding and Shielding
Ground loops are the most common source of noise and measurement errors in 4-20mA installations. The fix is to ground the cable shield at one end only — typically at the control panel end. Grounding at both ends creates a ground loop that induces noise into the signal.
Keep signal wiring away from high-voltage power cables wherever possible. If they must cross, cross at 90 degrees rather than running parallel.
Verifying the Loop
Once wired, verify the loop with a clamp meter or loop calibrator before commissioning. At zero input (minimum process), you should read 4mA. At full scale, you should read 20mA. If you're seeing less than 4mA, check for wiring errors or insufficient loop voltage. If you're seeing noise or instability, check your shielding and grounding.
A properly installed 4-20mA loop should be stable, repeatable, and virtually maintenance-free for years of operation.
