Copper Fights Back: How Linear ACCs Are Taming the AI Power Pinch

SAN JOSE, Calif. – October 12, 2026 – For the past three years, the dominant narrative in data center architecture has been a story of optical inevitability. As artificial intelligence models grew exponentially in size, so did the bandwidth requirements of the clusters training them. Industry consensus dictated that copper interconnects would soon hit a physical wall at 200 gigabits per second (Gbps) per lane, forcing a wholesale, expensive transition to optical connections even for the shortest runs inside the server rack. Today, that narrative is being fundamentally rewritten.

At the OCP Global Summit in San Jose, the Active Copper Cable Multi-Source Agreement (ACC-MSA) group announced the successful completion of its first industry plug fest. The consortium successfully demonstrated multi-vendor interoperability for linear Active Copper Cables (ACCs) operating at 200 Gbps per lane. With nine companies participating—spanning cable manufacturers, semiconductor suppliers, networking switch makers, and test equipment platforms—the event marks a critical milestone. It proves that a robust, open ecosystem for next-generation copper connectivity is not just a theoretical concept, but a commercially viable reality ready to support the next wave of AI deployments.

For business leaders and infrastructure architects attempting to scale AI operations, the implications of this development are profound. It represents a strategic pivot point that could alter the total cost of ownership (TCO) and thermal management strategies for hyperscale data centers over the next decade.

The Power Pinch: Stripping Out the DSP

To understand why this interoperability milestone is generating such intense interest among cloud providers, one must first understand the severe energy crisis gripping modern data centers. Analyst firm 650 Group projects that AI hardware infrastructure spending will exceed $1 trillion by 2029, with networking infrastructure alone approaching $200 billion. But building these massive AI factories is only half the battle; powering and cooling them is the true bottleneck.

Historically, to push copper cables to higher speeds and longer reaches, the industry relied on retimed Active Electrical Cables (AECs). These cables incorporate digital signal processors (DSPs) at each end to clean up and retime the high-speed data signals. While effective, DSPs are notoriously power-hungry. In a dense AI cluster containing thousands of connections, the power consumed by DSPs alone generates massive amounts of heat, compounding the already severe cooling challenges presented by next-generation GPUs.

Linear ACCs take a radically different approach. By utilizing advanced linear equalizers integrated directly into the cable connectors, ACCs can extend the reach of copper without the need for a power-hungry DSP. The result is a connection that leverages the capabilities of each segment of the link to form a power, cost, and latency-optimized pathway.

"The successful completion of our first plug fest represents a major milestone for the ACC-MSA and the broader industry," said Marek Tlalka, Co-Chair of the ACC-MSA and Vice President, High Performance Connectivity Product Marketing at MACOM. "Interoperability is fundamental to ecosystem growth. By bringing together cable manufacturers, semiconductor suppliers, networking companies and test equipment providers, we are demonstrating that a robust multi-vendor ACC ecosystem is both achievable and ready to support next-generation AI and data center deployments."

By eliminating the DSP, linear ACCs offer a highly attractive value proposition: significantly lower power consumption, reduced latency, and a lower bill of materials compared to both retimed AECs and optical alternatives. For a hyperscaler deploying tens of thousands of ports, these incremental per-port savings translate into megawatts of conserved power and millions of dollars in TCO reduction.

Copper Fights Back: Challenging the Optical Consensus

This resurgence of copper challenges the long-held assumption that optical interconnects—specifically Linear Pluggable Optics (LPO) or Co-Packaged Optics (CPO)—were the only viable solutions for 1.6T (8x200G) rack architectures. While optics remain essential for longer-reach connections between racks and across the data center floor, they are inherently more expensive and delicate than copper.

Optical transceivers require lasers and complex photonics to convert electrical signals into light and back again. This conversion process introduces its own power penalties and reliability concerns. By proving that copper can reliably handle 200 Gbps per lane over short distances (typically up to 3 meters, sufficient for intra-rack connectivity), the ACC-MSA is effectively defending copper's home turf.

Interestingly, the push for linear solutions is not isolated to copper. Semtech, a co-chair of the ACC-MSA, has also been deeply involved in developing 200G per lane Linear Pluggable Optical channels. This suggests a broader industry convergence around "linear" architectures—stripping out DSPs wherever possible, whether the medium is copper or glass, to maximize power efficiency.

Ecosystem Economics and Multi-Vendor Realities

The success of any new hardware standard hinges not just on technical merit, but on supply chain resilience. Hyperscale cloud providers like Google, Meta, and Microsoft—several of which are heavily involved in the ACC-MSA—refuse to be locked into single-source vendor relationships. They demand a thriving, competitive ecosystem that can guarantee high-volume manufacturing and drive down costs.

This is where the ACC-MSA's recent plug fest carries the most weight. By demonstrating seamless operation across products from nine different vendors, the consortium has validated its open specifications. Furthermore, the group announced the addition of seven new member companies: Accton, ACES, Foxconn Interconnect Technologies (FIT), Lintes, Marvell, TeraSignal, and Zhaolong Interconnect.

The inclusion of heavyweights like Marvell and global manufacturing giants like FIT signals that the supply chain is rapidly tooling up for volume production.

"ACCs continue to gain traction in the market as we welcome an expanding group of industry leaders to the consortium," said Scott Schube, Co-Chair of the ACC-MSA and Vice President, Applications and Business Development, Signal Integrity Products at Semtech. "The addition of these new members strengthens our collective expertise and reinforces our shared commitment to advancing next-generation active copper cable technologies. Together, we are building a broader ecosystem that will accelerate innovation, interoperability, and adoption across the AI and data center markets."

Deployment Timelines and Hidden Pitfalls

Despite the optimism surrounding the OCP Global Summit demonstrations, enterprise leaders must approach the transition to 200G per lane linear ACCs with analytical rigor. The physics of pushing 200 Gbps over copper are unforgiving, and signal integrity remains a monumental engineering challenge.

While the plug fest proved interoperability in a controlled testing environment, real-world data center deployments introduce variables such as extreme thermal fluctuations, physical cable bending, and electromagnetic interference. Validation and testing will be more critical than ever. Keysight Technologies, a founding member of the ACC-MSA, is prominently featured at the OCP Summit, showcasing advanced methodologies for validating digital-layer bit error ratios (BER) and forward error correction (FEC) performance for 1.6T AI data center deployments.

One infrastructure engineering lead from a major cloud provider, speaking on the condition of anonymity, noted that while the power savings of linear ACCs are impossible to ignore, the burden of signal equalization now shifts partially back to the host switch silicon. This requires incredibly tight coordination between the switch SerDes (Serializer/Deserializer) and the linear equalizer in the cable.

Organizations planning their next-generation AI infrastructure should view linear ACCs as a highly promising, TCO-optimizing technology that is rapidly approaching commercial maturity. However, they must also budget for rigorous, multi-vendor validation testing within their specific rack architectures before committing to at-scale deployments. The ACC-MSA has proven that copper is far from dead; in fact, by shedding its power-hungry DSPs, it may have just secured its place as the vital nervous system of the AI era.