Today’s AI software is fragmented. Every accelerator has its own compiler, kernel libraries, runtime, and development path. Developers absorb the cost of this fragmentation.

In his ModCon 2026 tech talk, Abdul Dakkak, Chief Scientist at Modular, presents our alternative: a unified compute layer, flexible enough to extend to new models, modalities, and hardware.

Abdul walks through our recipe for bringing up new hardware, demonstrated with three examples:

1. AWS Trainium, via our own team. Gemma 4 31B end to end.

2. Google TPU v6e, through our partner @HTECgroup. Their team brought it up with no LLVM backend to lower to, no prior knowledge of Mojo or MAX internals, and minimal support from us.

3. d-Matrix Corsair, implemented by @dMatrix_AI’s own team on their own stack. Repo access Tuesday, working matmul by the following Monday.

Thanks to HTEC and d-Matrix for their collaboration, and to Mihailo for presenting HTEC’s learnings.

Read about the HTEC collaboration: https://htec.com/insights/media-coverage/htec-and-modular-extend-max-and-mojo-to-google-tpu-cutting-ai-hardware-enablement-from-years-to-months/…

Watch the full talk: https://youtube.com/watch?v=hYjKbTAGOo0…

HTEC and Modular Extend MAX and Mojo to Google TPU, Cutting AI Hardware Enablement from Years to Months | HTEC

Source: https://htec.com/insights/media-coverage/htec-and-modular-extend-max-and-mojo-to-google-tpu-cutting-ai-hardware-enablement-from-years-to-months/ Advanced compiler and runtime engineering validates a faster, reusable path for bringing AI workloads to new silicon without rebuilding the software stack

PALO ALTO, CA,September 8 2026– HTEC, a global AI-first engineering and technology services company, and Modular have demonstrated Modular’s AI software stack running on Google’s TPU architecture. Presented at ModCon 2026, the work shows that support for a new silicon target can be delivered by extending an existing software foundation rather than rebuilding it for each accelerator.

The project brought Modular’s MAX and Mojo ecosystem to Google TPU in months, a class of hardware enablement effort that has traditionally taken years. It provides practical validation of an alternative model for AI infrastructure: a portable software layer, combined with specialized compiler and runtime engineering, that can make new compute architectures usable faster.

As demand for AI compute grows, organizations are looking beyond a single hardware ecosystem to improve performance and scalability, while reducing exposure to cost, capacity and supply constraints. The barrier is often software. Each new accelerator can require extensive work across compilers, runtimes and inference systems before AI workloads can be used effectively.

Modular is addressing that barrier with technology designed to make AI workloads portable across silicon. HTEC applied its specialized compiler engineering and runtime systems expertise to extend that foundation to Google TPU, working from ambitious technical requirements and with significant engineering autonomy.

Darko Todorovic, CTO at HTECcommented: “We proved that Modular’s MAX and Mojo ecosystem could reach a new silicon target in months rather than years. That validates both Modular’s architecture and a practical path for organizations to adopt new AI hardware faster.”

Chris Lattner, CEO, Modular, added: “AI cannot afford a full software rebuild for every new accelerator. HTEC’s work on Google TPU shows how a common software foundation can bring new silicon into the AI ecosystem faster and at scale. That fundamentally changes the equation for silicon companies: instead of rebuilding an entire software stack, they can focus their engineering on what makes their hardware unique.”

Beyond the Google TPU milestone, the project validates HTEC’s long-term investment in AI hardware enablement and Modular’s approach to hardware-agnostic AI software. The experience further strengthens HTEC’s ability to support future semiconductor and AI inference initiatives across the industry.

About HTEC

HTEC Group Inc.is a global technology and engineering firm headquartered in Silicon Valley, providing AI-first, full-stack solutions that move organizations from strategy to production in a repeatable way. Combining premium engineering talent with digital strategy, design, and venture-building capabilities, HTEC supports the entire product lifecycle—from ideation to deployment.

With more than 15 years of experience applying AI and machine learning in enterprise environments, HTEC brings deep expertise across the technology stack—from silicon to application, including data engineering, model optimization, and AI application development. This ensures that emerging technologies translate into measurable ROI, scalable growth, and enduring business impact. By bridging the gap between vision and execution, HTEC helps the world’s leading companies define what’s possible—and scale it successfully.

Media contact: [email protected]