Meta to begin production of its own AI chips in September 2026

Meta announced on July 9, 2026, that its new AI chips will go into mass production in September 2026. The company employs a modular approach to development — this will allow it to adapt the architecture as AI requirements change.

Why Modularity Makes Sense

When a company designs a specialized chip, it must predict which computations will be critical within 1–2 years. With artificial intelligence, this is nearly impossible: bandwidth requirements, memory capacity, neural network architecture all change monthly. Meta chose a practical solution instead of creating a single monolithic crystal: chips are designed modularly, so that individual blocks (compute cores, caches, memory interfaces) can be redesigned or replaced without a complete overhaul of the entire architecture. This is like the difference between a monolithic building and a LEGO constructor — the latter is far easier to adapt to new conditions.

Why This Is Critical in 2026

AI models are growing exponentially, and compute requirements now are completely different than two years ago. Where the previous focus was on memory bandwidth, what's now needed is low latency and specific operation types for transformer architectures. Meta is investing billions in its own infrastructure for training large language models and recommendation systems. If the company had designed chips statically — based on 2024 requirements — by the end of 2026 it would be shipping obsolete hardware. Modularity enables rapid iteration: release the first batch in September, identify bottlenecks and hot spots in practice, then redesign the necessary blocks the following year without redoing the entire project.

Independence from Suppliers

For major AI labs (OpenAI, Google, Meta, Anthropic), access to powerful GPUs and specialized chips is a strategic resource. If you depend on NVIDIA or other suppliers, you're locked into their delivery schedules, pricing policies, and business decisions. Meta (like Google, which already uses its own TPUs) wants independence from external monopolies. Proprietary chips provide:

• Complete control over costs and delivery schedules
• Ability to optimize architecture for proprietary models
• Reduced vulnerability to export restrictions on AI hardware

Vertical Integration as New Standard

The start of production in September 2026 is a landmark moment. Meta is signaling that it competes with OpenAI, Google, and Anthropic not only at the level of models and services, but also at the level of hardware. The industry is moving toward vertical integration: companies that control the entire stack (models + chips + platforms + software) gain a decisive advantage in development speed and overall economics.

What This Means

When a megaplayer like Meta announces the start of proprietary chip production in September, it's not just an internal project — it's a signal that competition in AI is entering a new level. An invitation to the race for control over hardware, which is now often more important than control over algorithms.

*Meta has been recognized as an extremist organization and is banned in the Russian Federation.