Texas Instruments ships AI inference chip

Texas Instruments began shipping its first dedicated AI inference chip this week, entering a market dominated by Nvidia and AMD with a chip focused on edge AI applications.

The new chip, the TI TDA5, is designed for AI inference at the edge — in cars, industrial machines, and consumer devices — rather than in data centres.

The TDA5’s key differentiators:

• Power efficiency. 8 TOPS per watt, roughly 2x more efficient than Nvidia’s Jetson line.
• Cost. $89 per unit in volume, vs $400+ for comparable Nvidia chips.
• Integration. Combines AI inference with traditional TI signal processing (audio, video, radar).

TI’s strategy is to focus on markets where Nvidia’s data centre pedigree is overkill. Cars, factory robots, and consumer devices don’t need a GPU optimised for billion-parameter models. They need an efficient, cheap, reliable inference chip.

Market context:

The edge AI inference market is estimated at $40B in 2026, growing to $120B by 2030. Nvidia dominates the high end ($500+ chips for autonomous vehicles). Qualcomm dominates the mid-range (smartphones, XR). The low end (under $100 per chip) is fragmented.

TI’s $89 chip is positioned for the volume middle — devices that need real performance but not flagship capability. The target markets include:

• Mid-tier automotive ADAS.
• Industrial machine vision.
• Robotics.
• Smart home devices.

Why this matters:

The TI launch is another data point that Nvidia’s AI dominance is being challenged at the edges. Frontier training (Nvidia) and high-end inference (Nvidia + AMD) remain Nvidia-dominated. Edge inference is fragmenting across TI, Qualcomm, MediaTek, and Apple Silicon.

For US AI strategy, the TI launch is good news. The CHIPS Act is funding US chip manufacturing, and TI’s launch shows that the funding is producing results. More US-designed, US-manufactured AI chips reduces dependence on Taiwan and South Korea.

The launch also signals that the AI chip market is bifurcating. Frontier chips (Nvidia, AMD, soon Intel) for training and high-end inference. Edge chips (TI, Qualcomm, Apple, MediaTek) for everything else. The two markets have different economics, different customers, and different competitive landscapes.

For now, TI is a small player in a huge market. But it’s a sign that the AI chip market is maturing — the same way the broader semiconductor market matured in the 1980s and 1990s, with specialised players in specialised segments.