AC vs DC Again: Why the Future Grid Will Be Bilingual

In June 2026, Elon Musk revived one of electrical history’s oldest arguments. He wrote that “AC was the right choice back then, but DC is the right choice today”, pointing to solar panels, batteries, electric cars and computers before predicting that relatively little alternating current would remain in the distant future.

It is a perfect internet claim: technically suggestive, historically provocative and blunt enough to irritate almost everyone who knows the subject. The observation beneath it is real. Photovoltaic panels produce direct current. Batteries store it. Computers, phones, servers and LEDs operate internally at DC voltages. Electric vehicles are organised around DC battery packs, even though power electronics convert that energy for other parts of the drivetrain. Modern life contains far more DC than the wall socket reveals.

But “DC is becoming more important” is not the same claim as “AC is going away.” The better question is not whether Thomas Edison was right after all, or whether Nikola Tesla and George Westinghouse merely won a temporary victory. That framing belongs to the nineteenth century. The emerging system is layered: AC remains the public grammar of most transmission and distribution networks, while DC expands inside storage, photovoltaics, electronics, charging systems, selected buildings and long-distance links. Power electronics translate between them.

The next electrical age will not choose one current. It will become bilingual.

The future is not AC defeating DC, or DC defeating AC. It is the grid becoming bilingual. Editorial image generated by the author.

What Musk Gets Right

The electricity system is becoming more electronic, more storage-heavy and less strictly one-directional. The classical grid was organised around large generators sending power outward through transmission and distribution networks. Electricity had to be generated at almost the same moment it was consumed, while large rotating machines remained synchronised to the grid’s frequency.

That system still exists, but it increasingly shares the network with technologies that handle electricity differently. Solar panels generate DC before inverters convert it into AC for the grid. Batteries store DC and use converters when charging from or discharging into an AC network. Electric vehicles store energy in DC packs. Computers and telecommunications equipment use several internal DC voltage levels. The public grid may remain AC even as more of the equipment attached to it passes through a DC stage.

Power electronics have therefore become ordinary grid machinery. Inverters, rectifiers and DC/DC converters regulate solar farms, battery installations, motor drives, vehicle chargers and transmission links. They do not merely translate one voltage or current into another; they control power flow, respond to faults, regulate motors and allow equipment with different electrical requirements to participate in one system.

This is the genuine shift beneath Musk’s prediction. DC did not return by rebuilding Edison’s local generating stations. It returned inside devices that can transform and control power electronically.

Why AC Won—and Why Modern DC Is Different

The War of the Currents is often remembered as a conflict of personalities: Edison against Westinghouse and Tesla, commercial entrenchment against technical elegance, the incumbent against the future. The engineering reason for AC’s success mattered more. For a given amount of transmitted power, raising the voltage allows the current to fall, and resistive losses increase with the square of the current. Long-distance transmission therefore rewards high voltage.

Alternating current could be transformed economically with electromagnetic transformers. Voltage could be raised for transmission and lowered again for local distribution and use. As the US Department of Energy explains in its account of why AC gained its historical advantage, this allowed electricity to travel much farther with lower losses. Early low-voltage DC systems lacked a comparable method. Extending them over distance required very large conductors or many local generating stations.

AC was not a historical error waiting for semiconductor engineers to correct. It was a systems victory. It matched the central problem of the first electrical age: carrying electricity from increasingly large generators to distant consumers before efficient electronic conversion existed.

The direct current expanding today is not Edison’s system preserved in amber. Modern DC depends on semiconductor switches, digital controls, high-efficiency converters, battery-management systems and photovoltaic cells. Voltage can be changed electronically. An inverter can synchronise a DC source with an AC grid. A battery can move from charging to discharging within fractions of a second. A high-voltage converter can determine how much power crosses a transmission link and in which direction.

The old DC network was constrained by its inability to change voltage economically. The modern DC network is built around technologies whose principal function is transformation. Calling this “Edison’s revenge” is appealing, but it mistakes a shared current direction for a shared system.

Conversion Efficiency and the Building Case

A modern building can sound electrically absurd when described only as a series of conversions. Rooftop panels generate DC. An inverter converts it to AC for the building or grid. A battery may convert AC back into DC for storage. A laptop charger then rectifies AC into DC again. It seems obvious that retaining DC throughout should remove waste.

Sometimes it can. The conclusion does not apply automatically to every architecture. A grid-connected DC building still requires a central AC-to-DC converter. Devices require different DC voltages, so endpoint DC/DC conversion may remain necessary. Cable losses, protection equipment, converter design and the load at which equipment normally operates all affect the result.

A peer-reviewed study of matched AC/DC and DC/DC endpoint converters illustrates the danger of assuming that one converter type must always be more efficient. Using realistic time-series loads rather than rated-load figures, the researchers found no systematic advantage for the commercially available DC/DC converters they tested. DC performed better for some devices and AC for others, with low-load behaviour often deciding the result.

That finding should not be stretched into a universal verdict against DC buildings. The study examined endpoint conversion, not every benefit of integrating solar panels, batteries and loads on a common DC bus. It does show that counting the number of times the letters change from AC to DC is not enough. The complete architecture must be assessed under the conditions in which it will actually operate.

Buildings with substantial on-site solar generation, battery storage, LED lighting, electronics and variable-speed drives create a plausible case for local DC distribution. A shared DC bus may remove conversion stages between generation, storage and compatible loads while simplifying some forms of control and backup power. A new building designed around those components can consider hybrid AC and DC circuits as one system.

Retrofitting an ordinary house is a weaker proposition. Its supply, wiring, protective devices, appliances, standards and maintenance practices already assume AC. The market for DC-ready appliances remains limited, and devices still require different voltages. Avoiding several efficient wall adapters may not justify replacing functioning infrastructure. The opportunity belongs to careful design, not to ideological consistency.

HVDC: Where Direct Current Has Its Strongest Case

The clearest large-scale case for DC is not household wiring but high-voltage direct-current transmission. HVDC transmission can be especially attractive for long point-to-point routes, submarine or underground cables and connections between AC systems that are not synchronised with one another.

An HVDC link converts AC into DC at one terminal, carries the power across the route and converts it back into AC at the other end. The converter stations are expensive, but their cost can be outweighed when the line is long enough or when an AC cable would encounter more severe technical limitations. DC also allows the operator to control the quantity and direction of power crossing the link rather than leaving it to redistribute through a meshed AC network according to the network’s impedances.

These properties matter in a system where renewable resources may be far from population centres, offshore wind requires submarine connections and neighbouring regions want to exchange electricity without joining the same synchronous network. They also demonstrate why the energy transition cannot be understood by comparing generators alone. Cheap power has limited value if transmission cannot carry it to consumers. As discussed in the Journal’s essay on why the energy transition is a system-design problem, generation, storage, transmission, balancing and demand must work together.

HVDC nevertheless remains a specialised answer rather than proof of universal DC superiority. DC faults are difficult to interrupt because the current does not naturally pass through zero as AC does. Converter stations require complex equipment. Standards and multi-vendor interoperability are less mature, and a meshed multi-terminal DC grid is much harder to plan than a single point-to-point connection.

The choice changes with distance, topology and purpose. The appropriate current for a submarine interconnector need not be the appropriate current for a suburban feeder or an industrial motor.

Data Centres and Electric Vehicles Already Use Both

Data centres provide a concentrated version of the building argument. Servers ultimately operate on low-voltage DC, while conventional facility designs may pass incoming electricity through transformers, uninterruptible power systems and server power supplies before it reaches the processors. Berkeley Lab has demonstrated direct DC distribution to server racks as a way of removing some of those stages and reducing associated cooling demand.

The architecture has not become universal. Berkeley Lab notes that most racks are not currently powered through facility-wide DC distribution, and modern high-efficiency AC equipment has improved since the earliest demonstrations. Reliability, protection, equipment availability and operating practice matter alongside theoretical conversion savings.

Even modest differences become consequential when applied continuously to very large loads. That makes the issue more important as AI increases data-centre electricity demand. Data centres may become substantially more DC-rich behind the meter while the surrounding city continues to deliver AC. The external network does not have to change currents merely because the computational machinery connected to it does.

Electric vehicles make the hybrid arrangement even more visible. Their traction batteries store DC, while an inverter converts battery power into AC for the traction motor. Regenerative braking reverses the flow and returns energy to the battery. When a vehicle uses conventional AC charging, its onboard charger converts the incoming electricity into DC; a DC fast charger performs much of that conversion outside the vehicle.

The vehicle is therefore not evidence that one current has defeated the other. DC suits electrochemical storage. Inverter-controlled AC suits common traction-motor designs. Power electronics make the two behave as one drivetrain.

Charging hubs may use the same logic at a larger scale. A site with several fast chargers, solar panels and stationary batteries can organise some equipment around a shared DC bus rather than duplicating every conversion stage. The hub will usually remain attached to a wider AC network. Its internal DC architecture supplements that network rather than abolishing it.

Why AC Will Not Vanish

The weakest version of the DC argument treats existing AC infrastructure as obsolete clutter. It is nothing of the kind. Homes, factories, substations, transformers, protection systems, motors, appliances, grid codes and professional practices have been built around AC for more than a century, and much of that equipment performs its job extremely well.

Transformers remain simple, robust and highly efficient. AC protection can exploit the current’s repeated zero crossings when interrupting faults. Industrial motors and machines are widely designed for AC supplies. Utilities possess mature standards and extensive experience in operating synchronised networks. Replacing this system would require a benefit much larger than avoiding a conversion stage near a few endpoints.

Infrastructure also changes on a different timescale from consumer electronics. A phone or computer may be replaced after several years. Building wiring, transmission corridors, substations and generating equipment may remain in service for decades. The grid is not a collection of gadgets awaiting the next product cycle. It is capital stock, regulation, skilled labour, safety practice and institutional memory.

New technology usually enters such systems by attaching itself to what already works. Solar inverters connect DC generation to AC networks. Batteries use bidirectional converters. HVDC corridors begin and end at AC systems. Data centres and charging hubs can adopt internal DC distribution without asking the surrounding city to rewire itself.

DC growth may therefore depend partly on AC remaining the stable public network into which specialised DC layers can be inserted.

The Grid Becomes Bilingual

The modern electricity system is becoming more decentralised, controllable and dependent on power electronics. That does not mean every layer is converging on the same current. Large rotating generators still produce AC. Most public transmission and distribution networks remain AC. Solar panels and batteries contribute DC. Electronic loads consume it internally. HVDC carries selected blocks of power across particular routes. Converters negotiate the boundaries.

This is the part of Musk’s argument worth preserving. The technological centre of gravity has shifted. DC is no longer confined to batteries, railways and specialist installations; it is central to several of the fastest-growing parts of the electrical system. The prediction that little AC will remain is much less convincing. Expansion is not replacement.

There is an appealing symmetry in saying that Edison lost the first contest but will win the second. The symmetry is false. AC enabled electrification at a scale early DC systems could not economically support. Modern DC does not reverse that judgement. It responds to a world of semiconductor switches, photovoltaic generation, electrochemical storage and digitally controlled loads that Edison could not have built.

Technologies are not correct forever in the abstract. They are appropriate to particular problems, costs, materials and neighbouring inventions. Change the surrounding system and an old disadvantage can become manageable; a technology once limited to a niche can return in a form its original advocates would barely recognise.

The future may contain more HVDC corridors, battery-buffered charging hubs, DC server racks and hybrid buildings. Ordinary households may still receive AC. Factories may continue operating AC motors, and distribution transformers may remain on the same streets for decades. That is not inconsistency. It is infrastructure assigning each current the jobs it performs well.

The second electrical age will not end with Edison defeating Tesla from beyond the grave. It will end the need to declare a winner.

The grid is becoming bilingual.