AI interconnect technology

The 81st UN General Assembly general debate has concluded in New York, with speeches dominated by artificial intelligence, the situation in Palestine, and ongoing wars around the world. Delegates used the annual gathering to argue that these issues are increasingly interconnected, framing rapid technological change alongside prolonged conflicts as defining challenges for international cooperation.

DE-CIX, the German internet exchange operator, has released an open-source MCP (Model Context Protocol) server designed to enable AI systems to manage and interact with network infrastructure. The move aims to make interconnection environments more accessible to AI-driven automation and tooling. It signals DE-CIX's push to position itself at the intersection of AI and internet infrastructure, and it has been picked up by telecom and technology trade press.

Salience Labs has secured $125 million in funding to develop photonic networking chips, which use light rather than electrical signals to move data. The Oxford-born startup is part of a wave of companies betting that optical interconnects will ease the bandwidth and power bottlenecks facing AI data centers, and the round ranks among the larger raises in the sector this year.

Korean chip giants Samsung Electronics and SK Hynix are moving into co-packaged optics (CPO) technology, which links AI semiconductors using optical connections to boost data transfer speeds and energy efficiency. As AI computing demand surges, Korean media describe the shift as a new race in which chips compete 'at the speed of light', alongside global players developing similar interconnect solutions.

Apple is reported to be considering re-entering the server market, a move that would intensify competition among major technology companies in the interconnected server ecosystem. Commentators are weighing what such a return would mean for an industry currently dominated by the likes of Dell, HPE and cloud providers, and how Apple's chip expertise could reshape it.

Nvidia unveiled a prototype of its Optics-on-Interposer co-packaged optics technology at SEMICON West 2026. The approach moves optical interconnects onto an interposer alongside the chip package, aimed at cutting power and boosting bandwidth in large-scale AI data center networking. Coverage so far comes from industry press reporting on the demonstration at the semiconductor trade event.

Apple is reportedly considering a return to the commercial server market as competition over accelerator interconnect technology intensifies among chipmakers. The move would mark a notable shift for a company that exited the server business years ago, and it comes as rivals race to define standards for linking AI accelerators in data centers. Details about timing, products or partners remain scarce, and Apple has not publicly commented on the reports.

Investors are asking whether Credo Technology can turn its 1.6T optical transceivers into its next major growth driver as AI data-center demand accelerates. The company, known for high-speed connectivity solutions, is positioning next-generation optics to capture spending tied to increasingly powerful AI infrastructure. Discussion centers on whether optical interconnects can meaningfully expand Credo's revenue alongside its core connectivity products.

Anthropic is raising concerns about self-replicating 'AI worms', malicious programs that could spread automatically between AI systems. The company says successful AI worms would need more capability running on less powerful devices, but current technology trends make that feasible by the end of the year. The warning has prompted discussion of new security risks as AI agents become more widespread and interconnected.

Upscale AI has announced Token Fabric, a technology designed to connect and coordinate mixed chip systems in data centers. The product, covered by technology news outlets, targets heterogeneous computing environments where different types of processors work together, a growing need in AI infrastructure. Details on partners, performance, and availability have not been broadly reported yet, but the launch adds to a wave of startups building interconnect and networking layers for AI hardware.

Photon Bridge, a startup developing laser-based optical interconnects for linking AI chips, has raised €10 million in funding. The technology aims to speed up data transfer between processors in AI data centres, an area attracting growing investment as chip-to-chip bandwidth becomes a bottleneck for large-scale AI systems.

A legal commentary argues that competition law and intelligent transportation are increasingly interconnected, pointing to AI, connected and autonomous vehicles, digital ticketing and smart infrastructure as areas outpacing existing rules. The piece calls for coordinated global legislation, referencing frameworks involving the United Nations, the European Union and the United States to govern these converging technologies.

CScale, a startup building scale-up interconnect technology, has launched out of stealth with $188 million in funding. The company is targeting the growing demand for high-performance interconnects as AI and data-center workloads push the limits of existing chip-to-chip communication. The funding gives the young firm significant resources to compete in a market increasingly central to computing infrastructure.

Chip interconnect startup Eliyan has attracted takeover interest, holding early acquisition talks with British chip designer Arm, according to The Information. No deal has been confirmed, and the terms or status of the discussions remain unknown. The talks highlight growing demand for technology that speeds data movement between chips, a key bottleneck in AI data centers.

Upscale AI, a startup backed by Nvidia, has launched Token Fabric, a technology designed to connect AI chips from rival vendors. The interconnect system aims to let different makers' accelerators work together in large AI computing clusters, addressing bottlenecks as models grow and data centers mix hardware from multiple suppliers. The announcement draws attention because it challenges Nvidia's own proprietary networking stack while being backed by Nvidia itself.

Credo Technology Group Holding has unveiled new AI interconnect products, prompting renewed debate among investors about whether the chipmaker's share price has already run too far. The company supplies high-speed connectivity solutions that benefit directly from spending on artificial intelligence data centers. Commentators are weighing its growth prospects against a stretched valuation, with bulls pointing to AI demand and skeptics flagging the premium multiple the stock now commands.

Upscale AI, a startup backed by Nvidia, has launched a platform designed to connect AI chips from competing manufacturers. The move positions the company as a neutral interconnection layer in a market dominated by rival hardware ecosystems. The launch draws attention because Nvidia's own backing places it behind technology that could tie together competing semiconductor vendors.

Upscale has introduced Token Fabric, a technology aimed at letting different AI chips work together across an improved network. The company's pitch is that as AI workloads spread across diverse hardware, a common fabric layer can make incompatible accelerators cooperate more efficiently. Details on performance, supported chips, and customers are limited so far, but the announcement has drawn attention in semiconductor and AI infrastructure circles.

Analysis points to the rapid merging of artificial intelligence, robotics and the space industry into one large interconnected market. As AI advances power autonomous robots and satellite systems, companies and investors increasingly treat the three sectors as a single growth opportunity rather than separate industries, with governments and firms moving to secure positions in the converging field.

Networking startup Eliyan, which develops chip interconnect technology, has attracted takeover interest, with early discussions held with British chip designer Arm, according to a report. No deal has been confirmed, and terms or other suitors were not disclosed. The report puts the spotlight on consolidation in the semiconductor connectivity space, where larger chip designers are seeking technology to address bottlenecks in AI data centers.

Optalysys, a UK computing company, says its photonic chip technology can actively move and process data rather than merely transferring it between components. The claim, reported by Quantum Zeitgeist, positions optical computing as a potential alternative to electronic interconnects in data-heavy workloads. It could matter for AI and high-performance computing, where moving data is often the bottleneck.