The rapid expansion of artificial intelligence is increasing demand for the data centres that power it. Across the United States, however, plans to build these energy-intensive facilities are encountering resistance from communities concerned about electricity bills, water consumption, noise and environmental impacts. Against this backdrop, OpenAI announced a National Community Compact outlining commitments to improve transparency and strengthen community participation in the development of its computing infrastructure.

The company has also released a Compute Campus Policy Blueprint, proposing measures for governments and regulators to consider when overseeing data centre development. The announcement comes amid growing public and political scrutiny of the technology industry’s infrastructure expansion.

Ending Secrecy Around Data Centre Negotiations

One of the most significant commitments concerns non-disclosure agreements (NDAs). OpenAI says it will stop requiring public officials to sign NDAs during negotiations over future facilities, except where legitimate security concerns, trade secrets or commercially sensitive information require protection.

Such agreements have attracted criticism because they can restrict the information elected representatives share with residents about proposed developments. Greater disclosure could allow communities to understand the scale and potential consequences of projects before construction begins.

OpenAI has also committed to engaging residents early in the planning process, appointing a local point of contact for each facility and publishing annual reports detailing its progress against the compact’s commitments. The agreement applies to future computing facilities owned and operated by OpenAI, rather than every data centre used by the company.

Who Pays for the Electricity?

Electricity demand is among the most contentious aspects of data centre expansion. These facilities require continuous power to operate servers and cooling systems, potentially placing additional pressure on electricity grids and increasing costs for other consumers.

OpenAI has pledged to pay the full cost of the electricity its facilities consume, along with the infrastructure upgrades required to supply them. It also proposes financial safeguards to prevent households and businesses from bearing costs if a planned project is delayed, scaled down or cancelled.

The company has proposed reducing electricity consumption during periods of peak demand and making power available to the grid when necessary. Its planned facility in Effingham County, Georgia, illustrates this approach. Under an agreement with Georgia Power, the utility can reduce electricity supplied to the facility by up to 1,000 megawatts during periods of peak demand.

Water Consumption and Environmental Disclosure

Cooling systems are another environmental concern. Data centres generate considerable heat, and some cooling technologies require substantial quantities of water to maintain operating temperatures. OpenAI says it will prioritise closed-loop and low-water cooling systems and develop water-management plans for individual facilities. It has also committed to publishing site-level information on electricity and water consumption.

The company plans to disclose construction details and commission independent noise studies before facilities begin operations. Annual reporting on relevant environmental impacts is also part of the compact. These disclosures could help communities assess the pressures associated with hosting large computing facilities. Their usefulness, however, will depend on the completeness of the information and the scope for independent scrutiny.

Growing resistance to data centre expansion

The announcement comes as opposition to data centre development gains momentum in the United States. Residents have questioned whether the economic benefits of hosting these facilities justify their demands on local electricity, water and infrastructure. According to Data Center Watch, at least 45 data centre projects worth 68 billion dollars faced disruption from local opposition during the second quarter of 2026.

OpenAI has also promised to publish local hiring goals and provide information about employment opportunities. Its commitments include helping local suppliers compete for contracts and encouraging partnerships with community colleges and trade organisations to prepare workers for construction and technical jobs.

However, voluntary corporate commitments do not carry the same force as legally enforceable regulations. OpenAI’s policy blueprint therefore calls on governments to establish stronger oversight of electricity costs, environmental disclosures and community benefits.

As AI computing demand continues to grow, the pressure on electricity networks, water resources and local infrastructure is likely to intensify. Whether OpenAI’s compact delivers meaningful accountability will depend on how consistently its commitments are implemented and independently scrutinised.

A sharp bend, a junction or a sudden change in a circuit’s structure can weaken an electromagnetic signal. When communication systems operate at increasingly high frequencies, even these seemingly small disruptions become significant engineering obstacles. For researchers working towards sixth-generation (6G) wireless technology, finding ways to guide signals through compact circuits without losing their strength is a growing challenge.

A research team in India has demonstrated an approach that could help address this problem. Using a specially engineered silicon chip, scientists at the Jawaharlal Nehru Centre for Advanced Scientific Research (JNCASR), Bengaluru, have shown that terahertz waves can cross artificial barriers with very little signal loss, even while carrying high-speed data.

The work draws on a phenomenon called photonic Klein tunnelling, which describes how waves can pass through certain barriers without undergoing the substantial reflection or attenuation commonly associated with such obstacles.

In one of the three devices developed by the researchers, data transmission reached 100 gigabits per second (Gbps). The findings, published in Advanced Materials, offer a potential design approach for the compact terahertz circuits that future communication technologies may require.

Borrowing a Principle From Quantum Physics

The idea behind the research has its origins in a phenomenon associated with quantum mechanics. In quantum physics, Klein tunnelling describes the unusual ability of certain particles to pass through a potential barrier even when classical physics would predict that they should be reflected. Researchers have explored analogous behaviour in photonic systems, where electromagnetic waves can exhibit similar transmission characteristics.

The JNCASR team applied this principle through topological photonics, a field concerned with controlling the movement of electromagnetic waves using specially designed structures. They created a silicon photonic crystal resembling the honeycomb lattice of graphene, with triangular holes arranged in a precise pattern across a silicon slab. This arrangement produces boundary channels along which terahertz waves can propagate. These channels are resistant to certain forms of backscattering, helping signals retain their direction of travel.

The researchers then introduced artificial obstacles into these pathways by changing the dimensions of the triangular holes over a short section of the crystal. The experiment tested whether the waves could retain their transmission characteristics even when their pathways were deliberately disrupted.

When a Barrier Barely Interrupts the Signal

The results showed that the engineered structures could transmit terahertz waves across the barriers with remarkably little loss, provided the strength of the barrier remained below a characteristic coupling threshold of the crystal. A single barrier produced only about 1–2% signal loss. Even with two barriers placed consecutively, the loss remained around 2–3%. For comparison, a similar barrier in a conventional, non-topological waveguide caused a considerably greater reduction in transmission.

This is significant for circuit design because communication signals must often navigate multiple pathways and changes in geometry within a limited physical space. Reducing the losses associated with these obstacles could give researchers greater flexibility when designing compact terahertz components.

The team demonstrated the principle in three silicon devices. One incorporated two barriers in sequence. Another used an omega-shaped pathway containing four sharp bends and a central barrier. The third was a power splitter, which divided an incoming signal into two branches, with one branch containing an engineered barrier.

The devices were fabricated using high-resistivity silicon, a material compatible with established semiconductor manufacturing processes. Measurements across frequencies of 310–360 GHz closely matched the researchers’ computer simulations. In the power splitter, the two output signals remained nearly identical, indicating that the barrier had little effect on signal transmission.

What does 100 Gbps Mean in Practice?

The researchers also examined whether the structures could carry high-speed data, rather than simply transmit terahertz waves. The double-barrier device achieved a data rate of 48 Gbps, while the omega-shaped configuration supported 60 Gbps. The power splitter recorded the highest rate, reaching 100 Gbps using QAM-32 modulation. This technique encodes information through variations in the amplitude and phase of a signal.

At 100 Gbps, a five-gigabyte high-definition film could theoretically be transferred in less than half a second. The different transmission rates are important because they show how the approach performed across three distinct circuit configurations. The 100 Gbps result was achieved by the power splitter, rather than all three devices. The demonstration suggests that engineered barriers need not substantially compromise data transmission, even in circuits incorporating multiple pathways and sharp bends.

From a Laboratory Demonstration to Future 6G Hardware

Terahertz frequencies, which lie between microwaves and infrared light in the electromagnetic spectrum, are attracting interest for their capacity to support high-bandwidth communication. Their potential lies in accommodating large volumes of data at speeds that conventional wireless technologies may struggle to sustain. However, generating and transmitting terahertz signals is only part of the challenge. Researchers must also develop components capable of directing, dividing and processing these signals within compact devices.

The JNCASR study addresses this less visible aspect of future communication infrastructure. Its findings could help in developing integrated terahertz circuits that route and split signals on a single chip without substantial transmission losses. The next step for the research team is to explore whether the engineered barriers can be adjusted electrically, thermally or mechanically. This could eventually allow a device to redirect signals between different pathways on demand, potentially enabling more adaptable terahertz circuits.

The work remains an experimental demonstration of signal transmission at the chip level; it does not establish a commercially deployable 6G system. Further research will be needed to integrate these components into larger communication architectures.

Led by Abhishek Kumar, Assistant Professor at JNCASR and head of its Meta Optics Laboratory (MetoLab), the study involved joint first authors Samina Mulla and Sangram Kumar Sadhu. Researchers from JNCASR, IISER Kolkata, IIT Bombay, the Institute of Microelectronics at A*STAR, Singapore, and the University of Lille and CNRS, France, also contributed. The research offers a new way of thinking about how high-frequency signals can be guided through increasingly intricate circuits. As work on 6G progresses, such advances in the underlying hardware will help determine how much of the promised speed and bandwidth can eventually be translated into practical communication systems.

India has released three indigenous artificial intelligence models under the IndiaAI Mission, with applications ranging from multilingual reasoning and coding to offline AI assistance in sensitive sectors such as defence, government and education. The models — EKA-1, developed by Soket Labs Technology and Research Private Limited, and Dhīr 2B and Dhīr 5B, developed by ZenteiQ AiTech Innovations, were unveiled in New Delhi under the IndiaAI Innovation Centre (Foundation Models) pillar.

The three models have different technical capabilities and deployment priorities. EKA-1 is designed as a large multilingual foundation model supporting all 22 scheduled Indian languages and English. The Dhīr models, meanwhile, focus on lightweight, locally deployable AI assistance, allowing organisations to process information without depending on continuous cloud connectivity.

EKA-1 Trained on 7.5 Trillion Tokens

Developed by Soket Labs, EKA-1 is a 26.3-billion-parameter foundation model trained on approximately 7.5 trillion tokens. It supports all 22 scheduled Indian languages alongside English and is designed for mathematical tasks, coding, reasoning and agentic applications.

The model uses an open, sparse hybrid architecture. Its intended applications include banking, financial services and cybersecurity, where understanding India’s digital ecosystem and sector-specific requirements is important. According to Abhishek Upperwal, CEO of Soket Labs, the model was developed with support from the IndiaAI Mission to address requirements within India’s banking, financial services and cybersecurity sectors.

Dhīr Models Target Offline and Secure AI Applications

ZenteiQ has introduced two open-source models, Dhīr 2B and Dhīr 5B, supporting English and 10 Indian languages. Their primary focus is secure deployment within organisations. The models can operate on-premise, enabling institutions to use AI systems within their own computing infrastructure rather than relying entirely on external cloud services.

They support document-grounded automation, knowledge retrieval and citizen-facing or enterprise services. The press release identifies defence, government, industrial and educational institutions as potential application areas, particularly in environments with limited internet connectivity or sensitive information.

Their lightweight architecture also allows deployment on mobile devices, according to ZenteiQ. The company says the models can run across different computing platforms, including graphics processing units (GPUs) and tensor processing units (TPUs). Prof. Sashikumar, CEO of ZenteiQ AiTech Innovations, said local deployment would allow users to retain data on their devices, improving privacy and security while reducing dependence on cloud connectivity.

Public Computing Infrastructure Supports Model Development

Both development teams were selected under the IndiaAI Innovation Centre pillar, which aims to build indigenous foundation models and large language models. The IndiaAI Mission provided access to subsidised national AI computing infrastructure and support for associated research expenses. These resources helped the selected teams develop their models.

Speaking at the launch, S. Krishnan, Secretary of the Ministry of Electronics and Information Technology, said indigenous AI models could strengthen India’s technological and strategic autonomy while creating opportunities to improve productivity across government, industry, research and education. He also pointed to AIKosh, the IndiaAI datasets platform, as a channel through which researchers, students, innovators and the wider public could access these resources and experiment with AI applications.

India’s Wider AI Development Programme

Approved by the Union Cabinet on March 7, 2024, the IndiaAI Mission operates through seven pillars: computing capacity, the Innovation Centre, the AIKosh datasets platform, application development, future skills, startup financing, and safe and trusted AI. The development of EKA-1 and the Dhīr models falls under the Innovation Centre pillar, which supports the creation of foundation models tailored to India’s requirements.

The models represent two different approaches to expanding domestic AI capabilities: large-scale multilingual processing and lightweight AI systems designed for secure local deployment. Their practical impact will depend on how effectively they perform across real-world applications and how widely organisations adopt them.