Astera Labs Touts AI Connectivity Boom as Scorpio, Optics Fuel Growth Outlook

Key Points

• Astera Labs said demand tied to AI infrastructure is driving strong growth, with recent quarterly revenue of $308 million, up 93% year over year, and next-quarter guidance of $360 million at the midpoint.
• The company’s Scorpio product family is emerging as a major growth engine, with management expecting it to become Astera’s largest product line by year-end and to benefit from new hyperscaler and AI platform design wins.
• Executives also highlighted new opportunities in AI inference, custom silicon, and optics, including CXL memory offload, UALink, and near-/co-packaged optics as data center clusters scale beyond copper’s limits.

Astera Labs NASDAQ: ALAB executives said the company expects continued broad-based growth across its connectivity product portfolio, driven by rising demand for AI infrastructure, custom silicon deployments and emerging optical interconnect opportunities.

Speaking at J.P. Morgan’s 54th annual Technology and Media Communications Conference, Chief Executive Officer and co-founder Jitendra Mohan said Astera Labs reported $308 million in revenue in its most recent quarter, up 14% sequentially and 93% year over year. He said earnings per share were $0.61, compared with about $0.10 around the time of the company’s March 2024 IPO.

Mohan said Astera guided to $360 million in revenue for the next quarter at the midpoint, representing 17% sequential growth. He said the company expects all major product families — Aries, Taurus and Scorpio — to grow, with Scorpio expected to become the company’s largest product family by the end of the year.

Scorpio Products Drive Growth Outlook

Mohan highlighted Astera’s Scorpio family as a key contributor to the company’s expansion. He said the Scorpio P-Series, which targets scale-out applications, accounted for about 15% of revenue last year despite a partial-year ramp. The Scorpio X-Series, aimed at scale-up switching, began ramping into volume production in the first quarter of last year and is expected to overtake the P-Series within the broader Scorpio family.

Chief Financial Officer Desmond Lynch said Scorpio P-Series was Astera’s fastest-growing product line last year, driven largely by a lead customer across multiple programs. Lynch said the company is on track to ship the product to two hyperscaler customers in the second half of the year, with ramps expected to continue into 2027 and beyond.

For Scorpio X-Series, Lynch said Astera has been shipping lower-radix solutions and expects higher-radix products to begin shipping in the second quarter and continue into the back half of the year. He said the company has discussed 10 customer engagements across hyperscalers and AI platform providers and expects to convert “a couple” into design wins by year-end, creating revenue opportunities in 2027.

AI Inference Creates New Connectivity Needs

Mohan said Astera’s strategy remains focused on addressing connectivity bottlenecks in both AI training and inference systems. He said inference workloads, particularly those involving long context windows, are shifting bottlenecks toward memory, creating additional opportunities for Astera’s Leo CXL memory controller products.

He pointed to a hyperscaler opportunity involving KV Cache offload over CXL-attached memory as an example of a new use case. Mohan also said Mixture of Experts inference workloads require full peer-to-peer bandwidth among GPUs, a need Astera is targeting with its Scorpio switching portfolio.

Mohan said Scorpio features such as Hypercast and in-network compute can offload I/O and compute work from GPUs. He said Hypercast allows a GPU to send one message to a switch, which then distributes it to multiple GPUs, while in-network compute can combine results in the switch and send them back to GPUs.

Custom Silicon and UALink Opportunities

Mohan described Astera as “the Switzerland of connectivity,” saying the company aims to support both merchant GPU platforms and custom ASIC or XPU deployments. He said custom ASIC platforms can create more content opportunities for Astera because customers may deploy more ASICs than GPUs to address the same workload and may use PCI Express or PCI Express-like protocols for scale-up connectivity.

Mohan also discussed Astera’s custom silicon business, including its work with Nvidia on NVLink Fusion. He said Astera is collaborating with Nvidia and a hyperscaler customer to enable that customer’s compute to be deployed over the NVLink ecosystem, with a product ramp expected in 2027 alongside the customer’s ASIC ramp.

He also cited a custom version of Leo for KV Cache offload, noting that this did not require a new chip design but used software-heavy features in the existing chip architecture. Mohan emphasized that Astera’s custom programs are not traditional back-end ASIC engagements, saying the company handles concept-through-manufacturing designs and therefore can capture more value.

On UALink, Mohan said the open scale-up protocol is gaining traction because it was designed for high-throughput, low-latency scale-up applications and supports an open ecosystem. He said AWS and AMD have publicly supported UALink, with their platforms expected to ramp in 2027. Astera’s product development is on track to intercept those deployments, he said, adding that other customers have also decided to use UALink but have not publicly disclosed plans.

Optics Expected to Complement Copper

In response to an audience question, Mohan said copper will continue to be used at the rack level because of reliability, cost and power advantages. However, he said optics will be needed as clusters expand from one rack to multiple racks and copper reaches physical limits at higher speeds.

Mohan said some customers may move directly to co-packaged optics, with CPO for scale-out applications potentially arriving as early as 2027 and scale-up CPO in 2028 and beyond. Others may adopt near-packaged optics first. He said Astera expects initial NPO revenue in 2027 for rack-to-rack links.

He also discussed Astera’s aiXscale acquisition, saying its glass coupler technology helps address fiber coupling challenges, including alignment, mechanical attachment, detachability and high-density connections for large GPU clusters.

Margins and Long-Term Model

Lynch said Astera’s long-term operating model remains 70% gross margin and 40% operating income. He said recent gross margins have been slightly above the target but are expected to trend toward 70% due to warrant impact, product mix and broader switching use cases.

“Seventy percent gross margin is very rich for a semiconductor business,” Lynch said, adding that it reflects the value customers place on Astera’s solutions. He said the company will continue investing to support top-line growth as new products and customer diversification layer into the model.

About Astera Labs NASDAQ: ALAB

Astera Labs is a fabless semiconductor company that develops connectivity solutions for data center and cloud infrastructure. The firm focuses on addressing signal integrity and link management challenges that arise as server architectures incorporate higher-bandwidth processors and accelerators. Its technology is aimed at improving reliability and performance for high-speed interconnects used in servers, storage systems and compute accelerators.

The company's product portfolio centers on silicon devices and accompanying firmware and software that enhance and manage high-speed links.

This instant news alert was generated by narrative science technology and financial data from MarketBeat in order to provide readers with the fastest reporting and unbiased coverage. Please send any questions or comments about this story to contact@marketbeat.com.