Each new generation of mobile networks transitions to new frequency bands, and equipment for operating in these bands appears even before the standards are finalized. Researchers working on 6G technologies require radio systems capable of transmitting and receiving signals in bands inaccessible to current equipment, and they need these systems now, years before their actual deployment.
This concerns the FR3 band, which is located approximately between 7 and 24 gigahertz. This band sits between today's frequencies and very high millimeter waves. This is where a significant portion of 6G work is expected to concentrate, as it offers greater bandwidth than low bands while avoiding the coverage issues inherent in high frequencies.
The company was established in Chennai in July 2023 with the goal of creating equipment to solve this problem. The company's directors are Lakshman Bhaskaran and Ravi Teja Devara. The company emphasizes its roots in the research ecosystem around IIT Madras.
The equipment that this company competes against is almost entirely imported, with the research group in India acquiring software-defined radio for the new band typically purchasing American or European tools.
Software-defined radio moves as many functions of a radio system into software as possible. Instead of fixed circuits tuned to a specific band, the hardware digitizes a wide spectrum, allowing the code to determine what to do with it. This means one device can be reconfigured for various experiments and standards.
The company's platform covers the FR3 band and performs data processing directly on the board. It uses Field-Programmable Gate Arrays (FPGAs) for real-time beamforming, spectrum probing, and signal conditioning, without the need for external GPUs, which means less surrounding radio equipment for the lab.
The platform's architecture is modular and synchronized, which is critical for experiments conducted by its clients. Massive MIMO technology, which involves using multiple antennas simultaneously, cannot be tested on a single receiver; several devices working in coordination are required for this. The company states that its platform scales from a single unit to multi-node testbeds.
The company's clients include research groups, telecommunications equipment manufacturers, defense electronics companies, and drone manufacturers. The company positions itself based on affordable cost and open architecture.
REIO Systems has revenue but very little external funding. According to corporate reports, turnover for the year ending March 2025 was less than 10 crore rupees, and the company reports seed funding of about $52,000 and paid-up capital of 1 lakh rupees.
For a three-year-old hardware company, sales before attracting investment is unusual. Testing and measurement equipment is purchased by institutions with budgets, not end consumers, making early revenue more attainable than in most hardware categories.
The company has joined the Bharat Innovates 2026 program and was presented at the Nice exhibition in June, where it demonstrated seeking European research institutions for collaboration on FR3 and Massive MIMO testbeds. European universities and telecommunications research groups are among the primary users of such equipment, and a cheaper platform sells more easily to a laboratory than to a network operator.
Information regarding client names, order volumes, or number of units sold is not publicly available. On its own website, the company mentions defense research groups and wireless startups among its users without naming them specifically.
In an agricultural region approximately 120 km northwest of Houston, contractors have begun clearing land for construction that, according to Elon Musk, will be the largest building in history. This project, named Terafab, is a joint venture between Tesla and SpaceX.
The main goal of the project is narrow, despite its scale: to produce artificial intelligence chips required by Musk's companies, which they cannot acquire in sufficient quantities from other suppliers. According to Musk's vision, the single facility must provide one petawatt of computing power annually.
Musk stated on X on August 6th, accompanying initial visualizations, that 'Terafab Texas will undoubtedly be the largest and most valuable building on Earth.' He also noted that it would be 'strikingly beautiful.'
Over four planned phases, this complex is expected to cover about 9.3 million square meters of production space, distributed across four enormous rectangular buildings, with the largest reportedly measuring about 4 km in length.
For comparison, the area planned for Terafab would cover the center of Cape Town almost six times (the central improvement district covers 1.6 km²). Other analogies include seven and a half V&A waterfronts or roughly two-thirds of the entire Or Tambo airport territory. When compared to sports facilities, it is equivalent to the Loftus Versfeld field laid out approximately 1300 times.
The financial investment is also colossal. Tesla and SpaceX have allocated $16.8 billion for the first phase, which amounts to about 271 billion Rands at the current exchange rate. According to available data, the total cost of all four phases could reach $119 billion, approximately equivalent to 1.9 trillion Rands. Meanwhile, the entire consolidated South African government budget for 2026/2027 is 2.67 trillion Rands. Thus, one project on one site in Texas, spread over a decade of construction, could cost approximately 70% of the annual expenditure of the South African government.
Terafab is expected to employ at least 3000 people, but the company's commitment is less stringent. As part of a tax agreement with Grimes County, the company must spend no less than $5 billion by the end of 2030 and create 1800 full-time jobs by the end of 2035. In return, it will receive ten years of county property tax exemptions and a state grant of $30 million.
Musk's justification is that Tesla, SpaceX, and xAI are already utilizing a significant portion of the world's available contract chip manufacturing capacity. According to SpaceX's own projections, the combined future demand from these companies will exceed the capabilities of the global industry. Musk noted that production rates are 'much lower than we would like' and concluded: 'Either we build Terafab, or we won't have chips, and we need chips, so we build Terafab.'
Musk, known for his tendency towards exaggeration, called it 'the most epic chip-making exercise in history.' The goal is to produce chips for Tesla's Optimus robots and Cybercab, as well as radiation-hardened processors for planned SpaceX orbital data centers.
Terafab is also designed to house logical fabrication, memory production, advanced packaging, testing, and photolithography manufacturing on a single site. This will eliminate the need to send design iterations across continents—a combination Musk mentioned at the launch event that is absent globally.
When asked about the ultimate purpose of all this, he went beyond the balance sheet. Quoting Isaac Asimov and Iain Banks, as well as the concept of humanity as a galactic civilization, he stated: 'We want to make it real. Not just science fiction, but turn science fiction into scientific fact.'
Intel is the technological partner, and it is reported that Tesla was the first major client for this company's 14A process in April. Intel committed to ensuring high-volume production of 14A (1.4-nanometer) in 2028, although it did not publicly confirm the mandatory Terafab contract.
The plant is not expected to draw power from the Texas grid. SpaceX will build its own gas power plants on-site, which will operate alongside 'very large battery banks,' as described by the company's head of energy, Riley Trettel. Water will be sourced from the Gibbons Creek dam, not local groundwater.
Some points remain uncertain. The start date of construction has not been confirmed, and the project price fluctuated between $20 and $25 billion at the outset over five months, then $55 billion in SpaceX's May prospectus, and subsequently $16.8 billion for the first phase with a ceiling of $119 billion. This prospectus described Terafab as a general structure without commitments. Semiconductor analyst Mark Lapidus estimates the probability that the project will never happen at 80%, noting that none of the companies have previously managed a factory.
Twitch has implemented a change that allows the use of content generated by its users for training Amazon's generative artificial intelligence models. This feature is active by default, requiring users who do not wish to participate to manually disable it in their account settings.
The news gained prominence on Wednesday, the 12th, after content creator and streaming journalist Zach Bussey published a screenshot demonstrating the new option. It is important to note that this policy covers not only live broadcasts but practically all material from a channel, including VODs, clips, chat messages, images, and text.
The new setting authorizes the use of this content to 'train generative AI content models on Amazon.' Twitch itself cited the use of broadcast audio to improve speech-to-text conversion systems as an example, which could benefit the platform's automatic captions or be applied to other Amazon services.
The scope of the training goes beyond audio; the company clarifies that streams, saved videos, clips, chats, images, and text associated with the channel can be used. There is a specific caveat regarding chat: even if a user disables training on their account, messages sent in another person's channel with the option enabled may still contribute to the training data.
One of the main points of controversy raised by users is the adoption of an opt-out system. This implies that creators are automatically included in the training process and must take the initiative to access settings to express their refusal.
During a dedicated stream on the topic, Twitch Product Director Mike Minton justified the choice of this operational model. According to the executive, if participation depended on prior authorization (opt-in), adoption would be very low. He stated: 'If it were opt-in, nobody would opt-in,' thus explaining the maintenance of the default active option.
For those who wish to prevent their channel content from being used in training Amazon's generative models, the procedure is to disable the option in the account security settings, following the path: Twitch > Settings > Security > Generative AI Training. However, disabling this function does not mean losing all of Twitch's artificial intelligence features; tools such as chat moderation and ad selection will continue to operate normally.
Mary Kish, Twitch Community Head, commented during the stream that certain AI applications are already part of the platform's operation. She expressed skepticism about the possibility of Twitch operating without these technologies and confirmed the current use of AI in various areas of the service. Interestingly, Kish revealed that she personally decided to disable the use of her content for training purposes.
In contrast, Minton stated that if he were a streamer, he would keep the training option enabled. Another issue raised by users was the way the change was introduced, given that the option appeared in the settings without comprehensive prior communication to creators. Executives demonstrated where to find the feature only after receiving user complaints about it being hidden in the settings. Kish acknowledged the criticism regarding the lack of email notification but argued that creators do not always read their mail, suggesting that direct messages and community outreach could help disseminate the information.
The company also guaranteed that, for now, it does not plan to create a public identification indicating which channels have consented to or refused AI training. This decision places Twitch in a broader debate about which entity has the right to use user-generated content to develop AI models. In the context of Twitch, the debate gains complexity due to the vast quantity and diversity of information contained in a single broadcast, which can aggregate hours of video and voice, chat interactions, images, and text produced by both the streamer and the community. For now, users retain the ability to block the use of their own content in Amazon's training, although the default implementation has already generated criticism within the community.
For those who wish to modernize their home and have a virtual assistant always available, this is a good time. Three options of the Echo Dot, Amazon's smart speaker with Alexa, have been selected and presented in various colors to match any decor.
Among the highlighted features of these devices are Wi-Fi and Bluetooth connectivity, robust sound, and priority access to the Alexa+ service.
The blue-toned Echo Dot model provides vivid and powerful sound for any residential room. It has native Wi-Fi and Bluetooth integration, facilitating connection to the home network and other devices. Alexa is ready to perform tasks such as playing music, answering questions, and managing smart devices, in addition to offering early access to Alexa+ features.
The black version of the Echo Dot is ideal for those seeking a more sober and elegant look. Its performance is identical to that of the other versions: it features high-quality audio, Wi-Fi and Bluetooth connectivity, and Alexa is ready to assist at any moment. The benefit of early access to Alexa+ allows users to experience the assistant's innovations before other users.
For environments that value a lighter and contemporary aesthetic, the white Echo Dot integrates perfectly into the design. In addition to its minimalist appearance, the device delivers all the qualities of the Echo Dot line: powerful sound, Wi-Fi and Bluetooth connection, and full integration with Alexa. With early access to Alexa+, the user can start exploring the latest functionalities of Amazon's assistant.
It is important to note that the stock for these promotions may run out quickly, so it is recommended to access the links to choose the desired color and bring Alexa technology home.
