As AI clusters push data center networks from 800G toward 1.6T and eventually 3.2T connectivity, semiconductor companies are competing to solve the bandwidth, power and signal-integrity problems underneath the AI boom. Semtech is preparing to put its data center strategy under the spotlight with an October teach-in for investors and industry participants, highlighting the technologies it believes can capture the next phase of AI networking demand.
Semtech is preparing to make a detailed case for its role in the AI data center buildout.
The semiconductor company will host a Data Center business teach-in on October 15, 2026, in San Jose, California, where CEO Hong Hou, Signal Integrity Products Group general manager Imran Sherazi and CFO Mark Lin are scheduled to outline the company’s technology portfolio, growth opportunities, strategy and long-term financial objectives.
The event itself is not a product launch. Its significance is that Semtech is using the investor-focused event to frame its data center business around one of the fastest-moving parts of AI infrastructure: high-speed electrical and optical interconnects.
That market is becoming increasingly important as AI systems distribute workloads across large clusters of GPUs and accelerators. The processors may receive most of the attention, but those chips are only useful at scale if enormous volumes of data can move between them with sufficiently low latency and power consumption.
Semtech is positioned around that connectivity layer.
The company’s Signal Integrity portfolio includes analog and mixed-signal components for optical and copper interconnects, including FiberEdge, DirectEdge and CopperEdge. Its products are designed for 800G and 1.6T data center connectivity, while the company is also developing technologies aimed at 3.2T architectures.
The shift toward higher bandwidth is happening quickly. Semtech said at OFC 2026 that Dell’Oro Group expects 2026 to be the first year of volume deployments for 1.6 Tbps switches, with deployments projected to surpass 5 million ports within one to two years of shipments. Dell’Oro also expects strong double-digit growth in AI networking spending during 2026.
For Semtech, that creates an opportunity beyond simply selling more chips.
Its portfolio increasingly spans multiple approaches to the interconnect problem. FiberEdge provides optical physical-medium-dependent components such as transimpedance amplifiers and laser drivers. DirectEdge targets lower-power linear optical architectures, including linear pluggable optics (LPO). CopperEdge focuses on active copper cables and equalization for shorter-reach connections inside AI clusters.
That breadth matters because there is no single connectivity architecture for every AI data center.
Optical links become increasingly attractive as distance and bandwidth increase, while copper can remain useful for shorter connections where cost, latency and power efficiency are priorities. Meanwhile, LPO, co-packaged optics (CPO), near-packaged optics (NPO) and other architectures are being evaluated as operators attempt to push bandwidth higher without allowing networking power consumption to overwhelm the data center.
Semtech has been positioning itself for that transition. In March 2026, it introduced a family of 224 Gbps-per-lane TIAs and Mach-Zehnder modulator drivers supporting LPO, LRO, XPO, NPO and CPO applications at 800G, 1.6T and 3.2T.
The company also demonstrated 1.6T optical and copper connectivity at OFC 2026, including active copper cables using its GN8234 redriver ICs and optical systems built around Semtech components. It demonstrated a 3.2T active copper cable using a next-generation 448G-per-channel redriver as well.
For enterprise infrastructure teams, the implications are larger than a semiconductor product roadmap.
AI networking increasingly affects total system cost, power budgets, cooling requirements and cluster performance. A network link that consumes less power can indirectly create more thermal headroom for compute. Lower latency can improve communication between distributed accelerators. Higher bandwidth can reduce network bottlenecks during model training and inference.
McKinsey estimates that 800G optical transceiver production could fall 40% to 60% short of demand through 2027, while 1.6T transceiver supply could remain 30% to 40% below demand through 2029 under its analysis. That points to an infrastructure market where component availability could become a constraint on AI data center expansion.
The wider spending environment is equally striking. IDC estimates that global AI infrastructure spending reached $318 billion in 2025 and forecasts $487 billion in 2026, with spending expected to surpass $1 trillion by 2029.
Semtech’s opportunity sits inside that much larger capital cycle.
The competitive landscape includes established semiconductor and networking suppliers across optical DSPs, laser components, Ethernet connectivity and signal conditioning. Companies such as Broadcom, Marvell and Credo are also pursuing parts of the high-speed AI interconnect market, while major hyperscalers including Google, Amazon and Microsoft are increasingly designing specialized infrastructure and silicon.
The market is consequently moving toward a more integrated signal chain.
Semtech has made that strategy increasingly explicit. Its acquisition of HieFo expanded the company’s access to indium phosphide laser and gain-chip technology, which Semtech says strengthens its position in optical platforms including CPO and NPO.
The October teach-in gives investors a chance to hear how management intends to turn that technology breadth into sustained growth.
For data center architects, the more interesting question is technological: which combination of optical and copper interconnects will deliver the best balance of bandwidth, latency, power and cost as AI clusters scale?
That question is still unsettled.
AI infrastructure is moving faster than traditional data center upgrade cycles, and networking architectures are being redesigned accordingly. Semtech’s own recent demonstrations suggest the company is preparing for an environment where 800G becomes mainstream while 1.6T and 3.2T systems move rapidly from roadmaps toward deployment.
The October event should therefore be read less as a routine investor presentation and more as a strategic checkpoint.
Semtech has spent the past several years expanding its position across high-speed optical and copper signal integrity. The next challenge is proving that its portfolio can translate into durable share and financial growth as AI networking becomes one of the most consequential semiconductor markets in the data center.
Market Landscape
AI infrastructure spending is rapidly moving toward a scale where networking components are becoming strategic infrastructure rather than supporting hardware.
IDC forecasts $487 billion in global AI infrastructure spending in 2026, up from $318 billion in 2025, and expects the market to exceed $1 trillion by 2029.
At the networking layer, the transition is being driven by:
- 800G Ethernet and optical connectivity moving into broader deployment.
- 1.6T switching and interconnects beginning to enter volume deployment.
- 3.2T architectures emerging on longer-term roadmaps.
- LPO, CPO and NPO gaining attention as operators seek lower-power optical architectures.
- Active copper remaining relevant for short-reach AI scale-up connections.
- Signal-integrity semiconductors becoming increasingly important as electrical channels become harder to manage at higher data rates.
McKinsey’s supply analysis suggests optical transceivers could become a significant bottleneck if production capacity does not keep pace with AI data center demand.
For buyers, that makes vendor roadmaps, manufacturing capacity, standards compatibility and multi-generation product support increasingly important alongside headline bandwidth.
Top Insights
- Semtech will detail its data center strategy in October as AI networking shifts toward 1.6T and future 3.2T architectures.
- The company’s FiberEdge, DirectEdge and CopperEdge portfolios address optical, linear optics and active-copper connectivity across AI data center networks.
- McKinsey warns optical transceiver supply could lag demand, making networking components a potential constraint on the broader AI infrastructure buildout.
- Semtech’s 224G-per-lane components target emerging LPO, CPO, NPO and 1.6T-to-3.2T optical architectures for AI clusters.
- Enterprise infrastructure teams increasingly need to evaluate networking power, latency, bandwidth and supply availability alongside GPUs when planning AI clusters.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
