What are the opening hours?

Monday 9am - 6pm Tuesday 9am - 6pm Wednesday 9am - 6pm Thursday 9am - 6pm Friday 9am - 6pm

The Networking Report

Home
India
Pune
The Networking Report

Find the Right Information for Your Work
Discover. Learn. Decide
Get the latest market studies, rese

19/09/2026

Tether AI Research has developed multilingual on-device translation models that support nine European languages through a single multilingual deployment.

The TranslatePsy-EuroNano family is designed to reduce the need for separate bilingual models for every language pair. Using English as a pivot, the deployment enables 90 translation directions across English and nine European languages.

Key details include:
- Smallest deployment tier is 17.6 times smaller than a comparable setup
- Model sizes range from 36MB to 89MB
- Firefox’s equivalent setup requires 18 separate bilingual models totaling 633MB
- Controlled CPU benchmarking showed faster first-sentence translation times for TinyQ and BaseQ

The weights are openly available, and researchers can fine-tune them for domains such as customer support chatbots, virtual assistants, educational tutors, and question-answering systems. The models are also available through QVAC SDK for Android, iOS, Linux, macOS, and Windows.

How do you see on-device translation changing app experiences?

18/09/2026

A new report says AI data center e-waste may be far larger than current estimates suggest, because many calculations focus too narrowly on servers and accelerators.

The report from Basel Action Network says servers and accelerators represent just 13% of a data center’s electromechanical infrastructure, while five equipment categories, including networking, power distribution, storage, backup, and cooling, make up the rest. It estimates about 70,000 metric tonnes per GW of capacity.

The report also says AI equipment lifespans are being compressed to 2.5 to 5 years, and projects global e-waste could reach 196 million to 211 million metric tonnes per year by 2050. Of that, 31 million to 46 million metric tonnes per year could come from AI waste alone.

Some analysts question parts of the math, but agree the broader concern is valid. The main takeaway for IT leaders is that lifecycle planning may need to include more than compute equipment.

How should organizations measure the full lifecycle impact of AI infrastructure?

17/09/2026

Axelera has introduced Europa, a new AI Processing Unit aimed at enterprise data centers that need more inference performance within tight power and cooling limits. The platform is designed to help organizations add AI capacity to existing server environments without building new high-power infrastructure.

Europa will debut in two PCIe card products, the Edge 232p and the Server 250p. The Server 250p is a dual-slot accelerator with 128GB or 256GB of LPDDR5 memory, and up to eight cards can be installed in a server.

Axelera says the Europa AIPU delivers 629 TOPS within a 45-watt power envelope. The company also says its software stack spans edge and server deployments, with tools for computer vision models, LLMs, VLMs, diffusion models, speech, and other AI workloads.

Could power-efficient AI hardware become the new standard for enterprise deployment?

16/09/2026

Cisco has released emergency patches for a critical zero-day in its Secure Email Gateway appliance after active exploitation was found in the wild.

The vulnerability, tracked as CVE-2026-76461, is an SQL injection issue in email parsing code that could let an attacker send a crafted email and gain root-level command ex*****on on affected devices.

Key details include:
• The flaw affects both physical and virtual versions of the product.
• Fixes are available in AsyncOS firmware releases 15.5.5-0141, 16.0.4-3021, and 16.5.0-780.
• CISA has added the issue to its Known Exploited Vulnerabilities catalog.

Cisco also warned that indicators of compromise may be missing because attackers with root access could alter logs. Organizations are advised to review network and firewall logs outside the device as well.

How should organizations prioritize response when a zero-day may already have been exploited?

15/09/2026

Arm is expanding its AI data center push with two chip announcements focused on agentic AI workloads.

The new Neoverse CSS N4 is the company’s most configurable Compute Subsystem to date. It is designed to help customers move faster from processor design to production silicon and supports customized silicon for AI compute platforms, data-processing units, networking, and other specialized infrastructure.

CSS N4 can include up to 128 processor cores on a die and supports LPDDR6 memory and PCIe Gen 7 connectivity. Arm says it can deliver up to twice the performance, 1.25 times the performance per watt, and 1.75 times the memory bandwidth of Neoverse CSS N3.

Arm is also highlighting adoption of its Arm AGI CPU, a production-ready processor for highly responsive AI workloads. Support is growing across the AI and cloud ecosystem, including OpenAI, Meta, Cloudflare, Oracle, SAP, Lenovo, Supermicro, and Verda.

What stands out most in this shift toward agentic AI infrastructure?

14/09/2026

AI infrastructure is creating more heat, and data centers are beginning to turn that heat into cooling capacity. Next-generation AI racks are projected to reach densities north of 300 kW, while liquid cooling and on-site power generation are producing waste-heat streams that are harder to ignore.

Absorption chillers use heat instead of electricity to produce chilled water. That makes them a strong fit for two trends already shaping data center design.

Key points:
Liquid cooling loops reject heat at higher, more stable temperatures.
On-site power systems, including gas turbines, engines, and fuel cells, produce usable exhaust heat.
Some techno-economic studies point to payback periods of roughly 2.5 to 2.8 years, with CO₂ reductions in the thousands of tons per year.

For infrastructure leaders, the question is shifting from how to remove heat to how to reuse it. Could recovered heat become a standard part of data center cooling strategy?

12/09/2026

1.6T networking is moving from specification work to live interoperability testing as AI-driven bandwidth demand accelerates. Work on 1.6T began in 2020 and is now in the final balloting and review stage of IEEE P802.3dj, while vendors are already bringing products to market.

Later this month at ECOC in Malaga, Spain, more than a dozen vendors will place 1.6T switches, modules, and cables on the same show floor to test interoperability. The Optical Internetworking Forum will also highlight 1600ZR, a coherent interface designed to carry a full 1.6 terabit Ethernet connection over a single wavelength across campus and metro distances.

Direct-detect 1.6T is already shipping, and coherent optics are extending reach even farther. New work is also underway for 1600ZR+ and 1600CL, while planning for 3.2 terabit Ethernet is expected to begin in the first quarter of 2027.

How soon do you think 1.6T will become common in production networks?

11/09/2026

Nvidia is investing $3.5 billion in MediaTek through convertible bonds, while MediaTek will adopt Nvidia’s NVLink Fusion technology. The deal gives MediaTek access to Nvidia’s data center infrastructure and expands collaboration across AI infrastructure, local AI computing, and automotive platforms.

The partnership is centered on connecting custom AI chips and third-party processors to Nvidia-based systems. MediaTek will help customers develop application-specific AI chips that can operate alongside Nvidia processors in large-scale AI systems.

The companies will also continue working on multiple generations of Nvidia RTX Spark and DGX Spark PC chips, along with platforms for AI-powered, software-defined vehicles.

What stands out most in this deal, the funding or the tighter technical integration?

10/09/2026

MikroTik has released patches for six RouterOS vulnerabilities, including two flaws that can be chained to take over devices without authentication over SSH. The exploit chain, called MikroTrick, is already being used by attackers in the wild.

The affected flaws were found in several firmware components, including the SSH server and client, the bandwidth-test service, X.509 certificate handling code, and the WebFig interface.

MikroTik released fixes in RouterOS 7.25 beta 3, 7.24.2, 7.23.4, and 6.49.21. The company also advises against enabling SSH on the internet interface.

More than 122,500 MikroTik devices have SSH reachable from the internet, with the highest numbers in Brazil, the US, and Indonesia. The company says devices may be marked as 'Flagged' if compromise signs are detected.

How should organizations handle exposed router management services right now?

09/09/2026

ThousandEyes reported 687 global network outage events during the week of August 31 through September 6, up 6% from the prior week.

U.S. outage events totaled 537, a 15% increase from the week before.

Key category trends included:
• ISP outages rose globally from 297 to 331, and in the U.S. from 203 to 245.
• Public cloud network outages increased globally from 183 to 185, and in the U.S. from 158 to 166.
• Collaboration app network outages remained at zero globally and in the U.S.

Two notable outages were reported. Cogent Communications experienced an outage on September 2 that lasted two hours and 41 minutes over a period of three hours and five minutes. AT&T experienced an outage on September 6 that lasted 57 minutes over a one hour and five-minute period.

Network performance continued to shift across providers and regions. What trends stand out most to you?

Address

Pune
411014

Opening Hours

Monday	9am - 6pm
Tuesday	9am - 6pm
Wednesday	9am - 6pm
Thursday	9am - 6pm
Friday	9am - 6pm

Website

http://networking.report/

Alerts

Be the first to know and let us send you an email when The Networking Report posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to The Networking Report: