For Tenstorrent's software stack, should development efforts focus on expanding compatibility with popular AI frameworks or optimizing performance for existing supported frameworks?

Introduction

The trade-off we're examining today is whether Tenstorrent's software stack development efforts should focus on expanding compatibility with popular AI frameworks or optimizing performance for existing supported frameworks. This decision is crucial for Tenstorrent's product strategy and market positioning in the competitive AI hardware landscape.

Step 1

Clarifying Questions (3 minutes)

• Based on Tenstorrent's market position, I'm thinking compatibility might be a key differentiator. Could you share our current market share and how it compares to competitors like NVIDIA or AMD in the AI hardware space?

Why it matters: Helps determine if we need to prioritize compatibility to gain market share or focus on performance to retain existing customers. Expected answer: Smaller market share compared to established players. Impact on approach: If confirmed, might lean towards expanding compatibility to attract more users.

• Considering our revenue model, I assume we primarily sell hardware with the software stack as a value-add. Is this correct, or do we have significant software licensing revenue?

Why it matters: Influences whether we should focus on software as a direct revenue driver or as support for hardware sales. Expected answer: Hardware sales are primary, software supports adoption. Impact on approach: If true, might prioritize compatibility to drive hardware adoption.

• Looking at user segments, are we seeing more demand from research institutions or enterprise customers? How does this split impact framework preferences?

Why it matters: Different user segments may have varying needs for compatibility vs. performance. Expected answer: Mix of both, with enterprises growing faster. Impact on approach: If enterprise is growing, might lean towards popular frameworks for easier integration.

• Regarding our engineering resources, what's our current split between compatibility and optimization teams? How quickly could we scale either effort?

Why it matters: Helps understand our capacity to execute on either strategy effectively. Expected answer: Smaller team on compatibility, larger on optimization. Impact on approach: If confirmed, might need to consider resource reallocation or hiring for compatibility expansion.

• Considering our product roadmap, are there any major hardware releases planned in the next 12-18 months that could impact this decision?

Why it matters: Aligning software strategy with hardware releases could maximize impact. Expected answer: Yes, new chip architecture planned. Impact on approach: If confirmed, might time framework expansion to coincide with new hardware launch for maximum market impact.