AMD has unveiled a powerful new workstation in Berlin that it describes as a personal supercomputer, designed to bring data-centre level artificial intelligence capabilities to desktop users.

The American chipmaker introduced the Threadripper Halo Station at the Internationale Funkausstellung technology event on Friday. AMD Senior Vice President Jack Huynh presented the system as part of a new generation of computing hardware built around the growing demand for AI.

The workstation is designed for demanding AI workloads and can be configured with up to two terabytes of system memory and 576 gigabytes of high-bandwidth HBM3E memory. It also features 96 Threadripper Pro CPU cores and two MI350P data-centre accelerators, with liquid cooling used to manage the system’s high performance requirements.

Huynh described the machine as the world’s most powerful workstation and said it would be capable of running artificial intelligence models containing more than one trillion parameters.

During a demonstration in Berlin, Huynh showed the system generating an entire 3D environment and flight simulator from a single prompt. The presentation was intended to demonstrate how advanced AI tasks could be performed locally without depending entirely on remote cloud infrastructure.

AMD plans to release the Threadripper Halo Station early next year, although the company has not announced a final price. Analysts expect the system could cost more than €100,000, placing it firmly in the corporate, research and institutional market rather than the consumer PC sector.

The workstation is aimed at organisations that want to operate large language models and AI agent systems on their own hardware. Running these workloads locally could reduce dependence on cloud services, particularly for businesses and institutions facing rising AI subscription and computing costs.

AMD’s closest rival in this market is Nvidia’s DGX Station, which also brings powerful AI computing capabilities into a desktop-sized system. The two machines take different approaches to local AI performance.

AMD’s system has a major advantage in memory capacity and bandwidth. Its larger memory pool is designed to allow trillion-parameter models to run locally, while Nvidia’s machine offers tighter integration between local systems and cloud-based computing environments.

Nvidia’s DGX Station is already available and sells for around €100,000, giving it an early advantage in the emerging market.

AMD’s entry could intensify competition as companies seek alternatives to cloud-based AI computing. The growing ability to run increasingly large models on local hardware could also change how businesses manage AI development and deployment.

By using standardised components while targeting data-centre class performance, AMD is betting that powerful local AI systems will become an increasingly important part of the next generation of computing.

Artificial intelligence is becoming an increasingly important tool in agriculture, helping farmers monitor crops, forecast yields, automate machinery and manage water and livestock more efficiently.

The technology has developed rapidly in recent years. At an agricultural expo in Izmir, Turkey, three years ago, artificial intelligence applications were largely limited to inventory calculations and specialist products displayed by smaller technology companies. Today, AI-based systems are moving closer to the centre of modern farming.

One of the fastest-growing applications is crop monitoring. Farmers traditionally rely on experience and regular field inspections to identify pests, diseases and other threats. Drones and satellites can now collect detailed images that are analysed by computer-vision systems.

These systems can identify signs of fungal infections and pest activity before they become obvious to farmers. Combined with soil sensors, the technology can provide additional information about field conditions and help determine where crop protection treatments may be required.

Yield forecasting is another area where AI is gaining ground. Sensor data and advanced models can help farmers estimate how much produce they are likely to harvest rather than simply measuring output after crops have been collected.

Syngenta has developed a generative AI system that it says can forecast yields with 95 percent accuracy while also providing recommendations on seed placement. Better forecasts can help farmers organise transportation, storage and sales in advance, potentially reducing losses and improving supply planning.

Robots take on more farm work

Agricultural machinery is also becoming more automated. Companies including John Deere are developing autonomous tractors, while Carbon Robotics has developed the LaserWeeder, which uses cameras, lasers and computing technology to identify and remove weeds.

The system uses 36 cameras, 24 lasers and 24 NVIDIA GPUs and can reportedly eliminate up to 10,000 weeds per minute across more than 100 crop types.

AI is also changing irrigation. Modern sensor systems can combine soil moisture readings, crop temperature and weather forecasts to determine how much water different parts of a field require.

Some deployments have reported water savings of around 30 percent while maintaining or increasing yields. Variable-rate spraying systems use similar principles to apply pesticides only where they are needed, reducing chemical use and creating records that can support regulatory requirements.

Livestock farming is another emerging area. Computer vision systems and wearable sensors can monitor animals for changes in movement, behaviour and vital signs.

These systems can identify potential health problems before visible symptoms develop, allowing farmers to respond earlier. Earlier detection can improve animal welfare while reducing unnecessary antibiotic use.As these technologies become more accessible, AI is increasingly moving from specialised agricultural applications into everyday farm operations, giving producers new ways to manage resources, reduce waste and improve productivity.

Nearly six terabytes of data allegedly stolen from Berlin’s state administration have been published on the dark web, exposing a large amount of sensitive information and raising concerns over the security of government systems.

The hacker group Rhysida is reported to have released about 1.44 million files containing information taken during a cyberattack on Berlin’s administration. The leaked material reportedly includes personal details belonging to state employees, such as birth certificates, telephone numbers, home addresses and documents that appear to contain employee absence records.

The scale of the disclosure has prompted warnings that the incident could have serious consequences for public officials and government security.

Among the files is a folder labelled “AG CBRN-Rahmenplanung.” CBRN refers to chemical, biological, radiological and nuclear threats. The presence of documents connected to this area has raised particular concern because the material could potentially contain information about how authorities prepare for major security incidents.

Investigative journalist Lars Winkelsdorf warned on social media that the cyberattack was of a scale that could threaten the functioning of the state. The potential exposure of sensitive government information has also raised concerns about whether the material could be exploited by criminal groups, terrorists or foreign intelligence agencies.

Rhysida had previously threatened to release the stolen information unless Berlin’s administration paid a ransom of 30 Bitcoin, which was estimated at around €2 million. The hackers established a countdown as part of their demand, with the deadline expiring at about 3:35 p.m. on Friday.

Berlin’s Senate had stated before the deadline that it would not give in to the ransom demand, maintaining its position that the administration does not negotiate with criminals making such threats.

The hackers followed through shortly after the countdown ended, publishing the data package on the dark web. Members of the group have also begun distributing lists of filenames through social media, potentially making it easier for others to identify and access specific documents.

The release has intensified questions over the extent of the breach and whether additional confidential information could have been compromised. The reported presence of personal records also creates the possibility of identity theft, fraud and targeted attacks against affected employees.

Berlin authorities and political leaders have so far faced criticism over what observers describe as a limited public response to the incident.

Officials are expected to assess the leaked material, determine how much information was compromised and examine whether the stolen files contain further security-sensitive documents. The incident is likely to renew scrutiny of cybersecurity protections across government institutions as authorities work to understand the full impact of the breach.