Altera 發佈 Agilex 2X，押注 AI 開發者拓展可編程運算

Altera unveils Agilex 2X, betting on AI developers to expand programmable computing

Altera Corp. today announced Agilex 2X, a portfolio of field-programmable gate arrays and system-on-chip devices aimed at artificial intelligence infrastructure, robotics, communications and intelligent edge systems, pairing new hardware architecture with an agentic software strategy designed to accelerate development.

The first family, Agilex 27, targets applications that demand high throughput and bandwidth. Altera claims a 1.8-times improvement in performance per watt over Agilex 7, alongside a new on-chip data network and support for next-generation connectivity.

Software support is available now in Quartus Prime 26.3, but Agilex 27 engineering samples are expected in the second half of 2027. QuartusIQ, the company’s AI development companion announced last week, is also available now through Quartus Prime Pro Edition 26.3.

The announcement puts a product roadmap behind a theme Chief Executive Raghib Hussain emphasized when I interviewed him for theCUBE’s NYSE Wired AI Factory Series: AI is driving demand for intelligent systems that can change as quickly as their workloads.

“AI is going to force a redesign of every equipment out there,” Hussain told me.

That is the context for this new platform launch. AI is changing the architecture of computing, from the data center to machines operating in the physical world. Altera is betting that programmable hardware will have a larger role in those systems and that AI-assisted engineering can help more customers use it.

The new products come as Altera is eyeing an initial public offering of stock following Intel Corp.’s sale of a 51% stake in Altera to Silver Lake in April 2025, 10 years after Intel acquired Altera. In September, it said its parent company filed confidentially to offer stock to the public, without providing further details.

A portfolio built around different system requirements

Agilex 2X spans four families. Agilex 27 addresses intensive data processing, beamforming and radio-frequency workloads. Agilex 25 targets applications such as high-resolution video, drones and low-latency sensors. Agilex 23 combines embedded computing and real-time processing for robotics, industrial automation and automotive systems. Agilex 22 focuses on input/output-intensive functions, including board control, motor control, security and interface bridging.

Altera says the families share a Quartus software experience, reusable IP and common security and embedded software foundations. That gives customers a framework for extending engineering investments as requirements change, although it does not mean every design moves between devices without modification.

In our conversation, Hussain described a similar spectrum with basic sensor interfaces at one end, sensor aggregation and synchronization in the middle, and more demanding vision processing at the other. The portfolio reflects the view that intelligence reaches the physical world through many different combinations of processing, connectivity, power and cost.

Moving data becomes a source of value

In our conversation, Hussain put data movement at the center of Altera’s AI opportunity. “Compute is one part, but data movement is the biggest problem to be solved,” he said.

Agilex 27 addresses that challenge through two related architectural advances. Third-generation HyperFlex embeds registers throughout the routing fabric, giving the design software more opportunities to retime and pipeline processing without consuming logic resources. New adaptive logic module modes improve how designs fit into the device.

HyperFlow, a distributed on-chip data network, is designed to move multiple terabits of data per second throughout the FPGA without using programmable logic for that transport or creating routing congestion, according to Altera.

My view is that the value lies in what customers can accomplish with the complete data path. Moving information efficiently leaves more capacity for application processing and can help address bottlenecks around accelerators.

That is consistent with the AI factory architecture I have been tracking: The system becomes the unit of value. Processing, memory, networking and software must work together to produce useful intelligence within a power and cost budget.

Altera’s big opportunity is to demonstrate where programmability improves that system. Its performance-per-watt claim is a starting point, but the customer results in sustained throughput, latency and total system cost will be more telling.

Physical AI needs coordinated sensing and action

Altera’s physical AI strategy starts with a fundamental engineering problem. Sensors produce information in different formats, at different rates and with different latencies. A machine or robot needs those inputs synchronized to form a useful view of its environment.

Altera describes FPGAs aggregating, cleaning, organizing and synchronizing sensor data before presenting it to other processors. The implication extends beyond humanoid robots. Industrial equipment, drones, surgical systems and autonomous machines all create opportunities for specialized data handling and control.

This is where I see a meaningful connection between AI factories and the edge. Intelligence developed using large computing systems must ultimately operate within local constraints: available power, physical size, response time and safety requirements.

Programmable hardware offers a way to adapt parts of those systems after deployment. But Hussain acknowleged that some functions will migrate into dedicated chips when volume justifies the investment. The opportunity for Altera depends on continuing to solve the functions that are changing, specialized or difficult to justify in a custom ASIC.

The developer opportunity may be the bigger expansion lever

Hardware flexibility only creates value when customers can build with it.  That makes QuartusIQ central to the strategy. The software combines Altera knowledge with the active project’s context and connects AI agents to engineering tools through Model Context Protocol. 

Altera reports up to 10-times faster design creation and up to 1.5-times faster total design cycles, based on internal benchmarks comparing Quartus Prime with and without QuartusIQ. Those figures are company measurements, not guarantees across customer projects.

Faster design creation is only part of the process. Engineers must still verify functionality, establish appropriate constraints, close timing and validate behavior in hardware.

My take is that the larger strategic opportunity is expanding who can use programmable computing. Embedded and robotics engineers understand their applications, but they may lack deep FPGA expertise. If agentic tools reduce the specialist work required, more projects could become economically practical.

Altera’s robotics example makes the opportunity tangible. An engineer starts with requirements for sensors, perception and control, then uses QuartusIQ and Nvidia Nemotron to help iterate toward a validated FPGA implementation. The commercial test is whether that process reduces the time, cost and specialist expertise needed to reach production.

Expanding who can build intelligent systems

Altera is building programmable hardware for a world of changing AI systems while using AI to change how that hardware is developed. The larger opportunity is to expand who can build those systems. 

If Altera can reduce the expertise and time required to turn an idea into working hardware, it could bring more applications and more developers into programmable computing.

Here’s Hussain on theCUBE NYSE Wired in June 2026:

Image: Altera

A message from John Furrier, co-founder of SiliconANGLE:

Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.

• 15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more
• 11.4k+ theCUBE alumni — Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network

SiliconANGLE Media is a recognized leader in digital media innovation, uniting breakthrough technology, strategic insights and real-time audience engagement. As the parent company of SiliconANGLE, theCUBE Network, theCUBE Research, CUBE365, theCUBE AI and theCUBE SuperStudios — with flagship locations in Silicon Valley and the New York Stock Exchange — SiliconANGLE Media operates at the intersection of media, technology and AI.

Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.